Amazing.

The big one: Gossip Goblin, the YouTube AI outfit run by Zack London, is putting an AI-generated feature in actual theaters this fall. Not a trailer. Not a proof of concept. A ticket, a seat, a run. And he’s not alone. There’s a whole wave of it right now, from indie creators you’ve never heard of, to Dave Clark’s Promise, to Darren Aronofsky.

This isn’t just momentum. It’s proof of concept.

When I started this Substack, the AI feature film was the brass ring. The moment somebody would reach up, grab it, and prove the whole thing was real. Now, it’s happening.

I’m chasing it a different way.

Subscribe now

When I started out making indie films back in the ‘90s, anything less than a full-length feature film felt like a failure. A betrayal, almost. Then television had its renaissance, premium episodic storytelling became the apex predator of the whole business, and the feature quietly stopped being the only mountain worth climbing.

I think we’re at another one of those moments.

My new project, The Kronos Conspiracy, is technically a feature. Or it will be. I’m telling it in two-to-six-minute episodes, one at a time, and when they’re all out I fully intend to assemble them into a feature-length film. I’m sure lots of creatives are approaching projects the same way. That’s not news.

What I find interesting here is that the form of the finished piece has begun to feel secondary. It’s almost beside the point whether it’s a feature, a TV series, or a run of webisodes. There are differences to each form, obviously, but none of them feel invalid to me anymore. None feel better or worse than the other. None of them has a monopoly on art (or on slop).

All of them are valid ways to create and consume high-quality media in 2026. And that’s cool. That’s another roadblock coming down. Another barrier to entry, broken. A step towards a world where the quality of the content is what matters, not where you get it or what shape it takes.

As for me, I’ve got a big story to tell, and I’m telling it scene by scene, moment by moment, and releasing it as I go.

It’s liberating, and feels every bit as valid as making a whole feature and showing nobody a single frame until it’s done.

Nothing against the people doing it the other way. It’s badass. I love that it’s happening. I just find it freeing to not think that way.

Which brings me to episode two, which I just published to YouTube.

Episode one was the prologue. The man, the masks, the setup. Episode two is called The Body, and it’s where the world cracks open.

A scientist is dead at a billionaire’s research facility. They said he fell. Our man doesn’t buy it. And to tell you why, he hides behind AI masks and uses AI “recreations” to tell his story.

Part of the story is how he tells it.

The tools aren’t the subject. They’re just how he, and I, tell the thing.

I’m having fun here… hope you come with me on the journey.

If you missed episode 1, start here:

And if you’re ready for episode 2:

#AI #Filmmaking #Storytelling #GenerativeAI #IndieFilm #YouTube

Share

Leave a comment

Subscribe now

I have a very clear memory of standing in the Boston University bookstore, seventeen years old, about to overpay for an Apple computer because I was starting film school in the fall and simply nothing else would do. And while I waited in line, I was reading American Cinematographer.

Subscribe now

That’s the memory, anyway.

In reality the magazine must have been Fangoria or Starlog. There’s a non-zero chance it was Gore Zone. But let’s pretend I was a precocious young film major reading a respectable trade publication and not a kid looking at latex werewolf transformations.

In any case, the cover story was Terminator 2.

If you weren’t there, it’s hard to explain what that movie did to people. The morphing. The liquid chrome. James Cameron reaching into the future and pulling out something nobody had seen before. I’d seen the movie three times in the theater, and the article made me realize why. It wasn’t just the spectacle. It was the strategy.

The 1990s technology gave Cameron a small window of what the technology could actually do flawlessly. So he built the whole film to live inside – and fully exploit – that window. The T-1000 isn’t chrome because chrome is cool. It’s chrome because chrome is what the effects could sell, perfectly, without ever tipping into fake. And god it was cool. The story was engineered around the capability. Not the other way around.

Seventeen-year-old me thought that was the smartest thing he’d ever heard.

I’m no James Cameron, but thirty-some years later, I’m kinda stealing the approach.

(And you should, too.)

Here’s the news

I’ve been quiet on here for a minute. Two reasons. One, I’m in pre-production on our AI-native animated series for Amazon, which is very happily eating a healthy percentage of my brain. And two, I’ve been building something of my own.

It’s a dark webseries called The Kronos Conspiracy.

Episode one is live on YouTube right now, and new episodes will be dropping every Thursday. The episodes are short and I promise they aren’t quite as creepy-clown-forward as the thumbnail might have you believe. Here’s the premise: a journalist sets out to debunk a conspiracy theory and gets pulled into one that’s bigger, and realer, than anything he imagined. The story is too big not to publish, but too dangerous to tell – so he’s using AI to hide his face.

Here’s a link: The Kronos Conspiracy | Chapter 1 | Why I Hide My Face

Episode one is a prologue. Setting the world. Establishing the vibe. The story proper starts next Thursday.

Why I built it this way

Because I did the Cameron thing.

I didn’t start with a story and then go wrestle the tools into serving it. I started by figuring out what generative AI is genuinely, reliably good at right now, today, this week. And then I built a world that lives inside those capabilities on purpose.

That was the plan: Find the window. Build inside it.

Next week’s episode expands the show’s visual language in a couple of ways that I chose specifically to live inside the window of what the robots are good at. Once that video is live I’ll be back here talking about the lessons I learned and applied in the making of these pieces.

Talk to you then.

And meanwhile, thanks for reading and checking out The Kronos Conspiracy.

#AI #Filmmaking #Storytelling #GenerativeAI #IndieFilm #YouTube

Leave a comment

Amazon broke the news this week: our film is going to series. I’ll be directing and executive producing with Peter Grumbine (BuzzFeed Studio’s VP of Video & AI).

Subscribe now

The news broke this week at AI on the Lot, where I had the pleasure of sitting on a panel with Pete and Kartik Garg (Amazon’s Head of Product and Tech, AI Studios), in front of an audience of journalists, executives, technologists, and creatives all trying to figure out what the next chapter of storytelling looks like.

It was a wonderful event, and I’ll have more to say about it next week. But one big takeaway is that by and large the community is approaching generative AI the right way. And Amazon’s strong “creators first” approach should be music to everyone’s ears. They’re leading the charge, and leading it responsibly. That’s not the sound of a creative blowing smoke at his corporate overlords. It’s a fact.

Now, the hard part: what comes next as we build a team of talented humans to create a full scripted series using AI as a tool.

The shift

Almost everything I’ve made with AI up to this point has been experimental. Trailers. A trippy claymation short. Projects deliberately designed to test the medium, find its edges, and figure out where the cheat codes live. They were short, mostly solo, and built around playing to whatever the models happened to do well that week.

The Amazon pilot was a different animal entirely. It’s an animal I don’t think you could have made even five or six months ago. We were working from a real script. It was dialogue-driven. Heavy on lip sync. The characters have non-traditional bodies and non-traditional faces, which the video models have strong opinions about and are not afraid to share.

So the pilot was already a big jump. And now the jump gets bigger again: from one six-and-a-half-minute proof of concept, made by a very small team; to eight episodes of broadcast-quality content built by a full production team. That’s a different operation entirely. And it’s the puzzle I’ve been thinking about every waking minute since the pickup.

Before I get to the team question, though, I want to talk about how the pilot actually worked. Because the principles that got us through the short may not be the principles that get us through the series.

Start with R&D, not with making stuff

Every instinct, on every new AI project, screams open the timeline, build a thing, make me feel something. The power of the tools urges you to leap before looking. It’s seductive. And I have been seduced. I always remind myself that the single most valuable thing you can do at the start of an AI project is sit with the tools, in low-stakes prep mode, and figure out the cheat codes.

What does this model do well, with these characters? What does it refuse to do? Which tool is right for which task? What prompt structure consistently produces usable output for this specific art style? Where are the points of failure that are going to bite you in week six if you don’t surface them in week one? And, most importantly, what projects should be abandoned before they get too far?

Lately I’ve jettisoned a bunch of good work on a photorealistic film because the uncanny valley is still very much a thing. And on a pre-teen adventure because the models don’t like to generate children (which, ok, fair play).

Both abandonments hurt, badly. But not nearly as badly as struggling with inadequate tools would have hurt. Storytelling is supposed to be fun. Wrestling with inadequate tools for the job is not fun at all.

The honest truth is that on this project we finished the first two weeks of an eight-week schedule with almost nothing tangible to show for it. No cut footage. No assembled scenes. Just a growing pile of tests, notes, and a slowly-building confidence that we knew how to execute. That confidence turned out to be the most valuable asset we produced in that time. And when we started building shots in earnest, we weren’t guessing anymore. We were running a play we knew the shape of. And that allowed us to truly run.

The R&D phase isn’t fun. It’s not sexy. But it’s where the film gets made.

Our alpha cheat code: manual lip sync

There are different ways of doing lip sync with AI. I can’t promise that we found the best solution, but we found the one that worked for this project. It’s the method that let us breathe and feel confident.

Lip sync in AI animation is notoriously brutal. You can’t just point the model at a recorded performance and expect it to match. The audio comes out close, or it comes out exact for half a sentence and then drifts, or it comes out nothing like what you asked for. One option we discussed was to fix this by hiring an After Effects artist to manually adjust mouth shapes frame by frame. I’m all for creating jobs, but for this it seemed counter-intuitive.

What we discovered – and this was the actual unlock – is that simple 2D animation is incredibly forgiving in the edit timeline. Unlike high-end CGI, where every frame is unique and any tampering shows immediately, simple 2D animation is full of repeated frames. A character holds a freeze for six frames. Her eyes flick left, then right, then she smiles. Those beats are loopable. Cuttable. Stretchable. You can double time a two-second “character looks around” beat into one second and nobody notices. You can slow a beat down by 300% and still get away with it. It’s enormously forgiving, and allows for an enormous amount of control and finesse.

You can tweak the performance of the voice actors AND the performance of the AI.

That’s the kind of control I’m always looking for.

The surgery was possible. It was repeatable. And once we had it down, lip sync stopped being the existential threat to the project that it looked like in week two.

That’s the kind of cheat code that changes what’s possible. Not because it’s clever, but because it converts something terrifying into something laborious.

Laborious is fine. Laborious is just work. Terrifying is what kills projects.

What I’m bringing into the series

Here’s where the real puzzle starts. Up to now, almost all of my AI work has been a one-man-band operation: me, the model, the timeline. A couple of gentle-handed collaborators tops. Lean is good because the work is improvisational by necessity: the model’s output reshapes the plan in real time. Shot A comes back with a surprise; and because you love the surprise, shot B has to change to accommodate it; so by the time you’re working on shot C, the storyboard you started with bears almost no resemblance to where the sequence actually went.

Before anyone accuses me of yielding the creative reins to a machine, that’s not what I’m describing. Collaboration is the same with humans: you can put twenty writers on the same problem and they’ll give you twenty different paths to the same destination. Working with humans, your job as the showrunner is to hold the destination tightly, but hold the path loosely. AI is the same. Some of the model’s ideas are bad. Some are offensively wrong. Some are better than mine. Some are just different in a way that opens a door I wouldn’t have found on my own. The job is recognizing which is which and steering accordingly. That’s how humans collaborate with each other. It’s also how humans collaborate with machines.

That’s the job.

The trouble is that this kind of improvisational, shot-by-shot rebuilding doesn’t scale logically to a team. I know a lot of AI filmmaking teams who have struggled to work as a team. I have too. It’s hard to collaborate when one of your collaborators is fundamentally unreliable.

So how do you do it?

The hundred-year-old answer

I went looking for inspiration from the greats. A colleague pointed me to The Illusion of Life: Disney Animation by Frank Thomas and Ollie Johnston. It’s one of the foundational books on the medium. (To the animators out there: yes, I know. It’s a classic. Allow me my late-arriving wonder.)

And what I found is that a hundred years ago, when Disney was first inventing animation as a medium, they were wrestling with our exact problems. How do you take a drawing – an inert line on a page – and make it feel? How do you make a viewer care about something that demonstrably isn’t alive? Those were genuinely open questions then, the same way “how do you direct an AI film” is a genuinely open question now.

The workflow Disney settled on was elegant. The director – or the lead animator – drew the keyframes. The big poses. The major moments. Then those keyframes got handed off to a team of animators who drew all the in-between frames that connected them. One person owned the vision. Other people owned the bridges.

The closest thing we have to keyframes in current AI parlance is “start frame” and “end frame.” But really, that’s not close at all. The start and end frames of a shot are often fundamentally the least interesting frames. They aren’t where the emotion live. They aren’t where the story is told. As the AI models are maturing there will be a shift towards models that are built to reference start, end, and key frames.

(Currently the work-around is to explain to the model what a key frame is and where it should happen in every prompt. It’s tiresome. It’s annoying. But it’s gonna get better.)

I think that’s the answer for us, too.

As we’re writing the scripts for the series, I’ll do all the keyframe work in parallel. Build the locations. Lock the character designs. Set the cameras. Block the scenes. Stage the visual gags. Set the style. Create the entire show in keyframes: a rigorous, fully-realized blueprint (possibly even a fully scored and edited animatic) before anyone generates a single moving shot.

Then the team takes those keyframes and makes them move. Lip sync. The micro-edits. The fills between A and B. And of course we leave plenty of room for them to improvise, to push past what I “drew,” to bring their own ideas. It’s a collaboration, not a paint-by-numbers. But the spine is locked, the destination is shared, and nobody has to rebuild shot three because shot two surprised everybody.

My gut tells me this is going to work, and that it’s going to be fast.

Of course, anyone who’s spent serious time with AI will tell you that when your gut tells you AI is going to be fast at something, there’s a fifty-fifty chance you’re about to get a swift kick in the ass.

So we’ll grasp onto this plan the same way we hold every other plan in this medium – holding our intention firmly, and our execution loosely. Stay flexible. Stay adaptable. Stay ready to throw the whole workflow out the window if the technology surprises us in either direction over the multi-month production.

Which it will. It always does.

And that, honestly, is the part I’m looking forward to most. Every challenge in this medium is an invitation to creativity. We’re inventing this process together, in real time, and I’m lucky enough to get to spend the next several months figuring out whether a hundred-year-old workflow holds up against the most advanced storytelling tools that have ever existed.

As much as I can, I’ll be sharing what we learn – the wins, the disasters, the cheat codes – as we go. Thanks for joining me on this wild ride.

Share

Leave a comment

If this is your first time here: welcome. Thanks for reading and sharing. The comment section is open!

I don’t think AI filmmaking is any fun at all.

The rendering. The consistency problems. The rejected prompts. And the typing. So much typing.

The long, tedious, copy-paste-from-another-document, JSON-prompt-as-long-as-your-arm, “wait, why is this supposed to be fun again?” typing.

Subscribe now

Character descriptions. Scene setups. Style references. Visual tone. All of it, hammered into a text field over and over again. Like it’s 1994 and you’re writing a term paper.

Or it’s 1980 and you’re Jack Torrance.

I kept thinking: this is not how creative people work. This is not how ideas actually move through a brain. Writing the perfect line of description in a screenplay is satisfying. Typing it over and over again in a form field is death.

As Truman Capote once said about Jack Kerouac’s work, “That’s not writing, that’s typing.”

A few weeks ago, a colleague pointed me to a workflow that fixed all of that.

It’s not revolutionary. It’s not some secret technique I invented. I didn’t write a line of code. I’m not “vibe coding,” whatever that means. (Okay, I know what it means. Still not doing it.)

It’s embarrassingly simple. And it’s made my creative process faster, more productive, and, crucially, genuinely fun again.

I’m going to walk you through the whole process I’ve been using from initial back-of-the-napkin ideation, to show Bible creation, prompt crafting, and beyond. The idea throughout is to use AI tools as a force-multiplier for your creativity – never, ever as a replacement.

Here’s how it works.

STEP ONE: STOP TYPING. START TALKING.

A few weeks back I discovered an app called Wispr Flow.

[I swear to God this isn’t a sponsored post… but it’s sure about to sound like one. Wispr hasn’t given me anything. I just love this tool and want you to use it. There’s an offer code at the bottom for a free month.]

Wispr is basically a transcription tool, but that description undersells it. Think of it less like dictation and more like a very patient assistant who follows you around, writes down everything you say, cleans it up, and hands it back in a form that’s actually readable. Actually makes sense. Actually organized.

Ever wanted to act like a brilliant novelist, pacing around the living room dictating ideas into a micro cassette recorder? Ever tried that? I have. It feels cool for about 5 seconds. Then you realize your train of thought is about as easy to follow as the plot of Tenet.

Wispr changes that because it corrects your work on the fly. You can tell it:

Wispr goes back, fixes it, and keeps going. No rewinding. No retyping. Just thinking out loud, the way creative people actually think.

So now, instead of staring at a blank word doc, you can walk laps around the Silver Lake Reservoir and talk into your phone like the guy nobody at the dog park is making eye contact with. You just... riff.

Stream of consciousness. Train of thought. The messy, chaotic, one-idea-bumping-into-the-next way that ideas actually come.

Then Wispr cleans it up. Then you drop it into Claude.

Then the real fun starts.

STEP TWO: LET CLAUDE ASK THE QUESTIONS.

This is where Claude comes in. Opus 4.7 is a bonafide credit hog, but it’s worth every token.

Once I’ve dumped my brain into a document via Wispr, I bring it to Claude and tell it (him?!) that we’re writing the bible for a TV series. But first, I want it (HIM?!?!) to ask me questions.

Just read what I’ve input, figure out what’s missing or underdeveloped or contradictory, and ask me about it. Clarify my thoughts. Expose inconsistencies. Point out opportunities. Shine a light on problems, cliches, mistakes. Make me question and explain myself.

Last week I spent a couple hours brain-dumping a new project. Claude came back with about 20 questions. Some I could answer immediately. Some were irrelevant, because Claude is just a machine. But others I sat with for a day because they were questions worth answering, and worth answering well. All of the questions were, in their own way, useful.

Then I spent another hour or two answering those questions, out loud, via Wispr, walking around the reservoir again, and fed the answers back to Claude. Rinse and repeat. All while getting my steps in. This did not feel like a typing session. This felt like a proper brainstorm.

And after all of this, it’s time for Claude to write the Bible. You haven’t typed a single word, and you’re not about to start now. Why would you? The typing is the part that sucks… and dammit, you deserve better.

Claude will spit out twenty, twenty-five, thirty pages of detailed, expansive creative documentation in highly structured outline format. Your characters’ histories. Where they start in episode 1. Where you think they might possibly be going in season 5. The look of the filmmaking. The tone. The editing style. The music. The feeling you’re trying to create. Everything that’s been living in your head, finally on paper. Consummately readable, whether your reader is a bot or a human collaborator.

And here’s the thing I want to be really clear about.

Neither Claude nor Wispr is writing your Bible. Wispr is organizing it. Claude is the writers’ room assistant who keeps the whiteboard legible. And Claude is the writing partner who asks useful questions while you pace and pontificate and get to feel generally impressed with yourself. As you should be, because the ideas are yours. The vision is yours. The AI tools are just doing the least fun part, so you can focus on the most fun part.

Could you type all of this up yourself? Sure. It’ll take ten times as long and be about 2 percent as fun. Your call.

STEP THREE: NEVER WRITE A PROMPT AGAIN.

(Okay, almost never.)

This is where the workflow really pays off. And it’s where I think it solves the single biggest barrier to entry in AI filmmaking.

When it’s time to actually generate character sheets, stills, video, the old way is brutal. You’re copying style references from one document, pasting character descriptions from another, hand-crafting JSON prompts longer than some screenplays. It’s grunt work. It kills the creative momentum right when you need it most.

The new way: you just talk to Claude.

Describe your scene. Where it fits in the story. Who’s in it. What the shots are. What you’re going for emotionally. Claude already knows everything about your project, it’s been living inside your Bible. It knows your character’s blue skin and pink hair (or was it green skin and a buzz cut?) and the visual language of your entire world. It knows the what. And more importantly it knows your why. For everything.

So it writes the prompts. The long ones. The JSON ones. The ones that suck all the fun out of the room.

And when it nails a prompt, you let it know. The color grade that you described, finally rendered in 4k glory by NanoBanana? Upload that frame to Claude and tell it: yes that. That was the prompt that worked for this scene. Remember that… so I don’t have to. Or: hey, here’s what your prompt did wrong. Fix it.

Claude handles it while you go back to thinking about story, rather than trying to remember what word doc you pasted that one JSON color prompt into.

That’s the whole workflow. Brain dump via Wispr. Writer’s room assistant via Claude. Bible via Claude. Prompts via Claude. Feedback loop to Claude so it continues to learn what good means to you, for this particular project.

You, meanwhile, get to be the director.

I’m currently using this process on a few very different projects. One is for a production company and a major studio streamer I’m not allowed to name yet. And one is my own project, which I’m keeping under wraps a little longer but am very excited to tell you about soon.

Both are running through the same workflow. Both have benefited from it in ways I didn’t expect.

The biggest surprise has been how much more I enjoy the process now. AI filmmaking has a reputation for being technically demanding and creatively draining. All tools, no soul.

This workflow flips that. The tools do the draining parts. You do the soul parts.

That’s how the future was always supposed to work.

Here’s the Wispr Flow offer code for a free month. Full disclosure: if you dictate 2,000 words then I get a free month, too. That’s not why I’m writing this post… but it doesn’t hurt.

Try the workflow. Pace around your living room. Talk into your phone like you’re dictating a manifesto. Make your neighbors nervous.

Then go make something impossible.

I remember exactly how I felt when I saw Being John Malkovich for the first time.

I was a broke indie filmmaker in my late twenties. I walked out of the theater and I couldn’t speak for about fifteen minutes. Not because it was the greatest film I’d ever seen. Because it felt like permission.

Permission to be strange. Permission to be specific. Permission to make the movie that only you could make, because no one else had your particular damage.

Subscribe now

For a few beautiful years in the late-90s and early 2000s – with films like Eternal Sunshine of The Spotless Mind, Donnie Darko, Ghost World – it really felt like the door had swung open. The indie film movement had stormed the castle. The weird kids were winning. And those of us who had been grinding away with our 16mm Arriflexes, our Canon XL1s, and our maxed-out credit cards thought: Finally. These are the films we get to make now.

You know what happened next.

Marvel happened. Then more Marvel. Then the sequel to the Marvel. The door didn’t just close – they bricked it up and replaced it with a wall.

Now look. I love a good blockbuster. I have cried during at least one Marvel movie. (True story.) But somewhere along the way, “safe” became the only strategy, and “original” became a liability. The polish stayed. The pulse faded.

This isn’t news.

The news is that the wall no longer exists.

Not because Hollywood changed its mind. They didn’t. They won’t. That’s fine. We don’t need them to.

The tools that used to live behind studio gates – the visual effects, the sound design, the ability to actually realize an ambitious vision without a $40 million greenlight and a roomful of executives telling you to make it more like the last thing – those tools are now sitting on your laptop.

They’re ready. It’s time.

Let me re-phrase that: they’re almost ready. And it’s almost time.

But we’re close enough that right now is when we all need to be in pre-production. Right now is the time to be building your characters, fleshing out your world, writing your story, bringing your vision to life in your mind and on the page.

Because in about… oh, I dunno, seven minutes… the AI tools are going to be good enough to make your vision a reality. Whether it’s a space opera or a quiet indie comedy… the technology to empower you to tell whatever story you want is about to arrive in force.

Last week I had the pleasure of speaking with Leon Silverman of MovieLabs. He’s one of the sharpest minds working at the intersection of technology and storytelling. He said something that I keep turning over in my head.

He said we need to use AI “to make a future worthy of our past.”

A future worthy of our past.

Not a future of cheaper content farms. Not a future of algorithmic slop dressed up to look like cinema. A future that honors what film has always been at its greatest – and then dares to go further.

That’s the assignment.

Here’s the historical pattern, and it never fails: every time a new technology democratizes filmmaking, the first wave is noise. Then someone comes along and makes something so specific, so personal, so theirs that it changes everything.

Godard did it with jump cuts. Cassavetes did it with handheld. Rodriguez did it for $7,000. Linklater did it with Slacker. Someone did it with every tool that everyone said would cheapen the art form.

Someone is going to do it with this.

And, no, even though they’re all quietly working on it: it isn’t going to be a studio. It’s going to be someone working alone, or nearly alone. Someone who has a story that no development executive would ever greenlight. A story too big, too strange, too personal, too specific to survive a notes process.

The Charlie Kaufman that doesn’t need a Spike Jonze. The Spike Jonze that doesn’t need a studio. The filmmaker with the story that only they could tell, finally with the tools to tell it.

That person is out there right now.

Struggling with the still-not-quite-good-enough-but-almost-good-enough tools and wondering if any of this is worth it.

It’s worth it.

The films that are going to matter in this new era won’t be the ones that look most like what Hollywood already makes. They’ll be the ones that could only exist because of this moment. Because the gates are open and the gatekeepers are confused and the tools are miraculous even when they’re maddening.

This is our moment.

Nobody’s minding the gate. And nobody’s handing out permission.

Now go make something impossible.

What’s the film you’ve always wanted to make that Hollywood would never touch? I mean it when I say “Let’s make something impossible.” So if you’re a creative with a vision, reach out: tom.danon@gmail.com

Subscribe now

In Part 1, we got into the weeds of the AI legal minefield. We talked about the monkey selfie that defines current copyright law, the exact threshold for “materially altering” an image, and how to build a defensible chain of title when a machine does the rendering. (If you missed it, go read it here).

Today, we are zooming out.

Everyone in the industry feels the tectonic plates shifting. The studios are looking to cut costs. The guilds are fighting to protect their members. And sitting right in the middle of this standoff is Generative AI. We all know AI isn’t just a new tool – it’s a wedge being driven straight into the heart of the traditional Hollywood system. Some would say it’s a stake.

In this climate of disruption, I asked Brent and Charlie to help us read the tea leaves on Fair Use, union politics, and the rise of a secondary, highly-profitable indie filmmaking ecosystem.

Let’s get back into it.

We all hear a lot of theory, but in your day-to-day practice right now, how are you actually seeing AI tools being incorporated into professional, monetized workflows?

Everywhere, even if it’s quietly.

Writers are using AI for brainstorming, outlining, and first-draft generation. Not necessarily to replace the writing, but to accelerate development. VFX and post-production teams are using generative tools for concepting: generating reference images, mood boards, and rough composites before committing to the expensive work. Music supervisors and composers are experimenting with AI-assisted scoring for temp tracks.

On the production company side, we’re seeing AI used in research, development, and pre-production: generating pitch visuals, analyzing comparable content, even drafting initial treatments. The tools are compressing the development timeline. The common thread is that the professionals using AI most effectively right now are using it as an accelerant, not a replacement.

When we were making True Crime docs at Lionsgate, we spent half our lives talking about Fair Use. With AI models scraping the internet for training data, can you give my readers a brief explainer on how Fair Use applies (or doesn’t) to generative AI right now? It seems to me that what AI does when generating an image is the very definition of “material alteration.” But as you’ve explained [in part one], that’s not at all the case in the eyes of the courts.

Your instinct – that what AI does when generating an image is itself a form of material alteration – is exactly the argument the AI companies are making. They say: the model takes in millions of data points and produces something new and transformative. That’s fair use. It’s a reasonable argument. And they might be right.

The early rulings, though, are going against them.

The Andy Warhol Foundation v. Goldsmith decision from the Supreme Court in 2023 was the biggest fair use ruling in decades, and it reframed the entire analysis. The old question was whether the new work was aesthetically transformative. The new question is whether it has a transformative purpose. That framework will be applied to AI, and it’s going to be difficult for AI companies. Courts will say: I don’t care how different this looks or how you accomplished it. Show me a real transformative purpose (e.g., commentary, criticism, education). Not just a shinier version of the same thing

Thomson Reuters v. Ross Intelligence is the case most people aren’t tracking, and it’s one of the more important ones. A federal court ruled that Ross’s use of Westlaw headnotes to train its AI legal research tool was not fair use, because the AI’s output was a direct market substitute for the original product.

Fair use has four factors, and “transformative use” is only one of them. Market substitution is another, and that’s where AI companies are getting hit the hardest. If an AI tool trained on copyrighted works produces output that competes with or replaces the need for the original, that factor weighs heavily against fair use, no matter how “transformative” the underlying technology is. So the courts so far are saying that even if the output is “new,” the training process itself may be infringing.

We all know that the studios are desperate to use AI to cut costs, but it seems like the artists’ unions are dead set against it. I know this may not be strictly true, but it feels like the unions are protecting the A-list at the cost of denying powerful storytelling tools to the rest of us. What do your more sober minds think about this standoff?

The framing of “studios want AI, unions want to ban it” is how this gets reported, but it’s not exactly accurate. No major guild is trying to ban AI. What they’re trying to do is make sure that when AI replaces a human job, someone still gets paid.

The most elegant example is what’s being called the “Tilly Tax,” a concept being discussed in SAG-AFTRA’s 2026 negotiations. The idea: whenever a studio uses a synthetic performer instead of a real one, they pay a royalty into the union’s pension and health fund. It wouldn’t ban AI. It wouldn’t restrict the technology. It would just remove the economic incentive to replace humans with it by making the AI option cost roughly the same.

Now, to your point about the unions protecting the A-list. There’s truth in the concern that established power structures don’t always serve emerging creators. But the 2023 guild agreements included some of the first real protections specifically targeting AI replacement of working performers. Those protections aren’t for the A-list. The A-list can negotiate their own protections. Those provisions protect the people who can’t.

Here’s what’s really interesting: the studios are playing both sides, and successfully. [Before OpenAI pulled the plug on their Sora AI platform] Disney invested $1 billion in OpenAI and licensed 200+ characters for the Sora AI platform. At the same time, Disney is sending cease-and-desist letters to ByteDance for generating unlicensed Disney characters. That’s the model: embrace AI, monetize it, and control the terms. “License or litigate.”

On a related note, we’ve all seen (or experienced) the DGA effectively kill a project with their good intentions. You scrape together the money to make a cool indie film on a shoestring budget, then the DGA forces you to go union, your costs explode, and suddenly a makeable film is dead in the water. To me this kind of nonsense is indicative of an industry setting itself up for massive disruption. Do you see a future where cheap, powerful AI creates a secondary, highly profitable entertainment ecosystem operating wholly outside of the studio and union systems? Or is that a fantasy?

It’s not a fantasy. It’s already happening.

YouTube, TikTok, Substack, Patreon-funded content – there’s already a massive, profitable content ecosystem that operates largely outside the traditional studio and union infrastructure. AI tools are accelerating that. The cost of producing watchable content is dropping fast. A solo creator with the right tools can now produce visuals, music, and effects that would have required a team and a meaningful budget five years ago.

But here’s where the “wholly outside” part gets complicated. The moment you try to scale (e.g., get on a major platform, attract real investment, secure distribution, hire recognizable talent) you run into the same infrastructure. Distributors require E&O insurance, which requires clearances, which requires a defensible chain of title. Talent with any profile is going to have union obligations.

The more likely future, in our view, is a hybrid. AI tools make mid-budget content viable again. But the legal and commercial infrastructure doesn’t disappear just because the tools changed. The creators who build within that infrastructure (even while using AI to push what’s possible) are the ones who’ll capture real, scalable value. The disruption is real. But it’s a restructuring, not an escape.

That said, here’s one version of where this could go. The guilds evolve to protect only A-list talent and cover a very small, concentrated number of “premium productions.” Meanwhile, the other ecosystem is AI-powered, global, and focused squarely on profit margins. The cost of failure drops to near zero, which means more shots get taken, which means more weird and original ideas get greenlit. That’s not a threat to creativity. That’s the best thing that could happen to it.

That’s the optimistic vision of the future I was hoping for! So let’s assume there actually IS a future for new creatives in this business. What advice do you have for creatives at the start of their careers, trying to navigate this massive technological shift?

Four things:

1. Learn the tools now. AI fluency is becoming the new baseline, the way editing software and digital cameras became prerequisites over the past two decades. The creators who learn to use it as a genuine creative tool rather than a shortcut are going to have an enormous advantage. Don’t wait. Build something with it this week.

2. But don’t skip the fundamentals. AI makes production cheaper and faster. It does not make storytelling better. The craft (voice, perspective, emotional truth, knowing what to cut and what to keep) is more valuable now, not less, because it’s the one thing AI can’t replicate.

3. Be original, perhaps to a fault. Remember that AI is a predictive model. Human beings are still the ultimate machine for originality. If you want to write the next Top Gun, you’ll be competing against models that have every screenplay ever memorized. But if you have a crazy idea that nobody has ever done before, you have a much better shot. So go do something weird, make it cheap, and let the audience decide if it’s any good. And please, do not help make any more Marvel movies.

4. Protect your rights from day one. Understand what you own and what you don’t. Document your creative process: save your iterations, your prompts, your editorial decisions. The deals that hurt people most aren’t the ones they lose. They’re the ones where they give away leverage they don’t realize they have.

This is one of the most exciting times to be starting a creative career. The tools are extraordinary. The opportunities are real. But so are the legal landmines. Get smart about both.

Share The AI Showrunner

Thanks for reading The AI Showrunner. And a massive thank you to Brent and Charlie at Evos Law for sharing their expertise. If you want to dive deeper into how to structure your projects to survive contact with the legal system, you can find them here.

Subscribe now

But who actually owns the film?

Working in traditional media as a showrunner and executive at Lionsgate, the rules were clear: a writer wrote something, a studio bought it, and an army of lawyers made sure the chain of title was airtight.

Subscribe now

But Generative AI has thrown a massive wrench into the gears of Hollywood’s legal machine. If you want to move beyond making YouTube videos for 1,000 views and actually start monetizing your AI-assisted work, you need to understand the new rules of the game.

To help us navigate this, I recently sat down for a conversation with two of my former Lionsgate colleagues: attorneys Brent Owens and Charlie Kelsey. They have since left the studio to found Evos Law, a firm specializing in the wild intersection of traditional entertainment and artificial intelligence.

In Part One of our conversation, we dive into the murky waters of copyright, the threshold for “material alteration,” and what happens when your AI protagonist accidentally looks exactly like Ryan Gosling.

Let’s dig in.

First off, give us the quick origin story. How did you both get to where you are, and what is the vision for your new firm?

BRENT: I started my career at WME, then spent nearly a decade at Pilgrim Media Group and Lionsgate negotiating entertainment deals from the production company side. By the time I left as VP of Business and Legal Affairs, I’d closed hundreds of deals across talent, producers, distribution, and IP. Charlie’s path was through scripted television, where he started as a manager for writers and directors before law school. He later joined the Los Angeles office of Simpson Thacher & Bartlett, where he worked on M&A deals in a rapidly consolidating media industry, before joining me at Lionsgate.

We started Evos Law because we kept seeing the same problem: the entertainment business is evolving faster than most of its lawyers. AI and emerging technology are creating entirely new categories of content, new ownership questions, and new deal structures. What excites us most is that for the first time in decades, the fundamental rules of who owns what, who gets paid, and how content gets made are all being renegotiated at once.

Let’s get into those fundamental rules. In the traditional model, we know exactly who holds the copyright. In the AI ecosystem the waters are far murkier.

The traditional system is fairly straight-forward: a human creates something, that human owns it (or assigns it to the entity that paid them). The entire chain of title flows from that simple principle.

AI presents new challenges in three primary ways:

First is Authorship: Copyright requires a human author. The Supreme Court just confirmed this when they refused to hear an appeal in Thaler v. Perlmutter on March 2^nd – AI cannot be an author under current U.S. copyright law. So if a machine generates an image, a song, or a script with no meaningful human creative input, nobody owns it.

NOBODY owns it?

That’s right: it immediately becomes public domain.

Next up is Training: Every major AI model was trained on copyrighted works. Whether that training is legal under fair use is the biggest copyright fight in a generation.

And finally, Outputs: Even when AI generates something “new,” it may contain elements recognizably derived from copyrighted works. That creates infringement risk that didn’t exist in the old model, because a human writer or artist knew what they were referencing. An AI model doesn’t inherently know.

Let’s talk about what you can do to actually copyright images created with AI. The other day you told me a story about a macaque monkey…

This is one of our favorite cases to bring up because it sounds absurd, but it established a principle that turned out to be incredibly important.

In 2011, a wildlife photographer named David Slater was in Indonesia and a crested macaque named Naruto grabbed his camera and took a series of selfies. The images went viral. PETA then sued on the monkey’s behalf, arguing that Naruto was the author of the photographs and owned the copyright.

The Ninth Circuit said no. The Copyright Act requires a human author. Naruto took the photo, but Naruto can’t own the photo.

Fast forward to AI, and the principle maps directly. In the Thaler case we mentioned earlier, Stephen Thaler tried to register a copyright for an image created entirely by his AI system, DABUS, naming the AI as the author. The Copyright Office refused. Thaler sued. He lost at the district court, lost on appeal, and two weeks ago, the Supreme Court declined to hear his case

The reason this matters for creators is that it establishes the baseline rule: if you want to own what an AI produces, you need to show that a human made the creative decisions that drove the output. The machine is the camera. You need to be the photographer, not the monkey.

So, pure prompt-to-output with no human intervention is unprotectable. But in my understanding, if a human artist uses AI as one tool among many, making creative decisions that “materially alter” the AI images, that’s potentially a different story. What actually qualifies as a material alteration? Does anyone know or is it like that old line from a Supreme Court Justice about pornography: “I know it when I see it”?

The honest answer is that nobody knows exactly where the line is yet. The Copyright Office hasn’t published a bright-line test, and the courts haven’t given us one either.

What the courts will be looking for is whether your creative decisions drove the expressive content of the work. Selection, arrangement, and creative judgment matter more than volume of changes. Did you composite multiple AI-generated elements together in a way that reflects your artistic vision? Did you use the AI-generated image as a starting point for significant human creative work? Those choices start to build a case for authorship.

One more thing to keep in mind: owning copyright in the new work does not automatically give you ownership of the underlying work. If you edit a Midjourney scene into something copyrightable, anyone can still use the raw Midjourney output. Your copyright covers what you added, not the AI-generated base.

Wow, ok. Let’s try to make a potential case for ownership. I may be biased, but as a former editor, my feeling is that editing is authorship. If I take raw AI-generated clips and cut them together – adding music, dialogue, pacing – does that final compiled sequence cross the threshold to claim full ownership of the work?

This is where the argument gets strongest, and your instinct as a filmmaker is exactly right: editing is authorship.

Copyright law has a well-established concept called “compilation copyright.” The selection, coordination, and arrangement of elements (even elements you don’t individually own) can be protectable authorship. Think of a mixtape, a curated anthology, or a documentary assembled from archival footage. You don’t own the underlying clips, but you own the creative work of assembling them.

If you’re taking raw AI-generated clips and making the editorial decisions – what stays, what goes, the sequence, the pacing, the juxtaposition, the emotional arc – and then layering in human-created elements like original music, recorded dialogue, sound design, and visual effects, that compiled sequence has a much stronger copyright claim than any individual AI-generated clip.

Let’s talk about the Ryan Gosling problem that I’ve written about before. Sometimes I’ll prompt a tool, and the output is a screamingly clear infringement – like an astronaut that looks exactly like Mr. Gosling, or a copyrighted character. Clearly all the major models are trained on real people and real protected work. And sadly the “copyright safe” models kind of suck. How can you protect yourself when using the major models?

This is actually two separate legal problems stacked on top of each other.

The first is federal law: potential copyright and/or trademark infringement. If your AI-generated astronaut is clearly derived from a specific copyrighted film, that’s a potential copyright claim by the studio. If the output looks like a real person, you’re also in Lanham Act territory: § 43(a) covers false endorsement, meaning if consumers could reasonably believe the person authorized or sponsored the content, you have a federal claim on top of the state right-of-publicity.

The second problem is state law: right of publicity, the legal right every person has to control commercial use of their name, image, and likeness. If your AI output looks recognizably like Ryan Gosling, you have a right-of-publicity issue regardless of whether any specific movie was copied.

Best practice for creators who want to monetize: treat AI output exactly like any other content that needs clearance. If it looks like someone, clear it. If it echoes a copyrighted work, clear it. If you can’t clear it, don’t monetize it. And know that E&O insurance carriers are now specifically asking about AI-generated content at renewal. If you can’t demonstrate clearances, you may not be able to insure your project. And if you can’t insure it, you can’t distribute it.

I’m starting to really appreciate why you guys are putting AI issues front and center at your new firm. Let’s talk about risk as it applies to an indie creator making AI YouTube videos that only get a thousand views. That’s a totally different risk factor than if you’re talking about a prod co. delivering a $20M feature. What’s the real risk of inadvertent infringement for an indie creator hoping to scale?

Let’s be honest about the risk spectrum. If you’re making a YouTube video that gets 1,000 views, the practical risk of enforcement is low. Studios aren’t sending legal teams after small creators for incidental AI-generated content.

But here’s the trap: if you’re building a library of content that you hope to eventually sell, license, or scale into something bigger, you’re building on a foundation you may not be able to defend. The risk isn’t getting sued today. The risk is that when your project gets real attention – when a distributor wants to pick it up, when a streamer makes an offer, when you need E&O insurance for a commercial release – you can’t prove you own what you made.

Every legitimate distributor, every platform, every insurance carrier is going to ask about your chain of title. “Was AI used in the creation of this content?” is now a standard question. If you can’t answer that question, the deal falls apart. Not because someone sued you, but because the buyer can’t take the legal risk.

Our guidance for indie creators is practical: use AI aggressively in development (concepting, brainstorming, reference material). That’s where the tools shine and the risk is lowest. But for the content you actually want to monetize and scale, make sure human-created elements are the backbone. Document your creative process. Build your chain of title from the ground up. The creators who do this now will be the ones who can actually capitalize when their projects take off.

Next week in Part 2: We dive into the impending clash between AI, the Hollywood Studios, and the Guilds. Are the unions protecting the working class, or just the A-list? And is a secondary, highly profitable, non-union entertainment ecosystem about to disrupt the entire industry?

Subscribe so you don’t miss it…

Subscribe now

Share The AI Showrunner

It felt like a disservice to write a 3-part course on AI Filmmaking and not talk about music at all, so let’s get into it. Today’s post will present two exercises that you can do right now to level up the way you use music in your work.

Unsurprisingly, it starts with leveling up the way you think about music.

Subscribe now

WARNING: This post is 0% about using AI to generate music. It’s about using music as a tool. If you’re here for prompt crafting or a hot take on Seedance, you’ll probably want to keep scrolling. I’ll see you next week.

Here’s the problem with music: it’s a cheater.

Take any sequence – a montage, or an action sequence, or a dialogue scene – and slap some music on it. It doesn’t matter what genre, or whether the music even fits the scene. You’re always going to find some nice moments of synchronicity: a swell in the cue that matches a dramatic head turn, or a beat drop that lands just as a door slams. These moments can help you “find” the emotion of a scene, but nine times out of 10 they’re a mirage.

Over my years as an editor and a showrunner I’ve been handed countless scenes where professional editors had just slapped a cue under the timeline. No sculpting. No choices. The relationship between the shape of the music and the shape of the scene was wholly inconsequential. And the thing is, if you just look at the cut quickly, those cuts often seem “good enough.”

In the AI world, we have a word for that: SLOP.

And whether you’re making AI films or traditional ones, Slop is the enemy. So today I’m going to talk about how to go deeper with your music choices. Some of this is going to be a bit woo-woo, and some of it may be a bit rudimentary, but I promise you that the techniques I’m writing about today will level up your scoring game.

Color and Shape

Way back in the silent film era, the Russian filmmaker Sergei Eisenstein was chasing a vision of “total cinema,” where the visuals and soundtrack were perfectly aligned to create what he thought would be an almost magical power.

He wrote about this extensively in his book The Film Sense. He posited that filmgoers “read” a frame the way they read a book: rigidly proceeding from left to right.

As seen in the below storyboard from his film Alexander Nevsky (scored by Prokofiev), Eisenstein extensively diagramed each frame, charting visual intensity from left to right, then used that as a roadmap for Prokofiev’s score. He aimed to make the soundtrack match the intensity of the visuals precisely.

This approach is as adorable as it is totally bonkers.

With 130+ years of movies behind us, we all know we don’t read a film frame rigidly from left to right. Our eyes wander around the frame wildly, only constrained by the filmmaker’s mastery of composition and flow. But sure, in the early days of cinema Eisenstein’s approach seemed valid.

And in fact, Eisenstein was onto something. Music does have a shape. Identifying that shape and calibrating it to (or against) your visuals, pacing, and performances is the editor’s superpower.

EXERCISE #1: Day In, Day Out

Here’s your first exercise. It’s of the woo-woo variety I warned you about. Let’s listen to some tunes. But not just any tunes – some great ones. “Great” being defined by a fantastic combination of talents coming together on one track:

• A writer with a strong POV.

• An orchestration with a strong POV.

• A performer with a strong POV.

(Shout out to my film school screenwriting professor who taught me this exercise approximately one million years ago.)

The Listening Playlist:

Here’s a link to a Spotify playlist featuring the American standard, “Day In, Day Out” (composed by Rube Bloom with lyrics by Johnny Mercer), performed by three very different artists:

1. Billie Holiday

2. Frank Sinatra

3. Sarah Vaughan

Listen to each track twice in a row.

On Your First Listen: Just feel the song. Understand the lyrics. Get a vibe.

On Your Second Listen: Think about colors and tactile sensations. Does the music make you see sparkles? Dirty violets or bright pastels? What shape does the music take? Is it poking at you with knitting needles or blanketing you in velvet? What about images? Does the drum break call to mind a subway rattling through a platform, or footsteps on dirty concrete? Is the vibe oily, or pristine?

The only wrong answer is a vague answer. Get specific.

Then move on to the next track and repeat.

I think you’ll find that the differences between the three versions are profound. Go do some listening before I spoil your fresh ears with my own takes below…

On Billie Holiday’s version, my inner ear hears desperate longing, and I see velvety purples and dusky reds. This is a woman who is rapt and in the thrall of her lover – addicted, you might say. “Addicted” is the safe-for-work word for what I hear on this track.

Sinatra (predictably enough) is the confident man about town, wrapped in chrome and in control. The images are sharp. Where Billie was singing from her bedroom, Frank is hunting… and he knows he’s not going home empty handed.

Then Sarah Vaughan comes at it from a wildly different angle: irrepressible excitement. Her tempo is way up. Sometimes she’s swept along by the music. Other times the music can just barely keep up with her. I see bright stabs of light, lens flares, gold, and silver.

Each version of this song knocks me on my keister in a very different way. And I would use each one to score a very different scene.

I loaded up that playlist with a couple of other songs – and their covers – that I think are useful:

• “There is a War”: The Leonard Cohen original vs. the Nathaniel Rateliff cover. Notice how the Rateliff track is a conversation, a dialogue. That’s a major shift to my ears that completely changes the track.

• “Gimme Shelter”: The Rolling Stones vs. Merry Clayton’s solo version.

The point of this exercise is to drill down and really hear the music. You don’t need any musical ability to figure it out; it’s all about emotion, and intention. Finding the cue that perfectly aligns with the intention of your scene – or sculpting it to create that alignment – that’s where the magic happens.

Shaping Scenes With Music

Now that you’ve learned to identify what a song is doing, it’s time to put it to work.

Sometimes (though rarely) you’ll play a track from beginning to end under your scene. When you do that, more often than not you’re ceding control… and that’s a recipe for slop. A great song will make it seem like it’s not slop, but that’s a trap. Don’t be seduced.

I like to think of music as a table of contents for the flow of a scene. If the emotions are dragging, what you might need is a cue change, a pause, the breath that music can create. Music can land a moment, it can pull the rug out, it can say “ok we’re done here, what’s next?” Sometimes just a key change at the right moment will make your actor’s internal life become crystal clear. It can save a bad performance.

Some of the best film actors do very little, because they know the power of two things: the close-up and the score.

When I assess a track in the edit bay, I have a checklist of questions I’m looking to answer, ranging from the coldly technical to the heartfelt:

• The Intro: Does it feather in so I can ease it under my scene, or do I need to crash in hard?

• The Outro: Does it fade out slowly or does it resolve in a dramatic fashion?

• The Arc: Does it develop over time? A track with a key or tempo change gives you exponential possible uses. You can place the shift at a moment that underscores or subverts your characters’ hopes. That’s power. We like power.

• The Variance: Does it have a quiet part to counter the loud part? A one-note track will only last so long before you need a new track. You need variance. You are the sculptor here, and you’re only as good as your clay.

• The Lyrics: The tone fits the emotion… but do the words? All the pieces matter.

Let’s assume you’ve found the perfect cue. Now it’s time to think about how to use it.

EXERCISE #2: Entourage

So how do you learn to control the cue? How do you bend the music to your indomitable will?

Here’s a fun exercise that I’ve used to train junior editors (and sometimes senior ones) on how to exert mastery over music.

Imagine you have a 3-minute, 58-second cue. It’s a great song. Perfect for the vibe of your scene. And let’s say that scene is the main titles of an HBO show. Crucially, the sequence has a hard time limit of exactly 60 seconds. So you’ve got to lose 75% of the cue. But the song is expensive to license, so you want to make sure to get all the bang for your buck by making sure all the best parts are on screen: the intro, the creeping build, the blasting chorus, the attention-grabbing bridge, and the final resolve.

That’s a lot to cram into a minute. But just playing the chorus is a disservice to the artist, the viewer, and the budget. You gotta make the most of it…

The specific track and show I’m thinking of is Jane’s Addiction’s “Superhero” as heard in the opening credits of Entourage.

Don’t you dare look the credit sequence up on YouTube. Don’t do it.

Don’t!

(Not yet.)

Instead: Spend a buck-ninety-nine, download the song, drag it into your editing software, and get to work. You’ve got a hard 60 seconds. Lose 75% of the song while maintaining its general awesomeness. Make it work. Know that if you sell the song short, Perry Farrell will egg your house.

The goal of this exercise is not to perfectly match the HBO version. It’s to identify the key elements of the song, figure out how much screen time they require to be “satisfying” to the ear and the heart, what can be lost, and what absolutely must be kept.

Perhaps you start with the intro and the outro and figure out how to fill in the middle. Perhaps you drop the song in your timeline and whittle down increasingly shorter versions to discover how much can go. Or perhaps you totally rethink the track and do some seriously amazing remixing violence to it.

The only wrong answer is an answer that sucks… or is 61 seconds long.

Once you’re happy with your cutdown – and only then – click on this link to watch the expertly cutdown version that HBO used.

Coda

The skills you’re unlocking here will carry you well beyond a simple title sequence. Once you learn how to make four minutes fit into three, you’ll see how much power you have, and what compelling results that power can create. Where do you need the key change to land? Where does the pace of your scene desperately need a boost from the soundtrack? How can you turn a crappy AI performance into something with soul?

Once you experience the power of bending a track to the constraints you place upon it, a world of cinematic control will open up for you.

And the emotional force you can impose on your timeline – and your viewer – will explode.

Share

Thanks for reading. I know a lot of you are extremely talented producers, editors, and filmmakers… I’d love to hear your thoughts on how you use music in the comments.

Subscribe now

Subscribe now

Let’s dig in.

I feel like there’s been a dramatic build-up to this moment – the moment we finally start creating video. But here’s the secret: if you’ve taken great care with the first steps I’ve described, this should actually be the easy part.

By now, you have an exacting vision and all the tools you need (your characters, locations, and storyboard frames) to put the pieces together. At this stage, the work shifts from creation to discovery. Your job now is:

• Unearthing problems you didn’t see coming (and turning them into strengths).

• Rejecting shots that turn out A+++ but don’t serve your vision.

• Discovering new ideas because the AI gave you something you weren’t expecting, but love because it perfectly serves your vision.

• Finding new connections that hadn’t occurred to you until you started stitching the footage together.

To create a philosophical framework for these steps – and these issues – I’d like to offer an anecdote.

The legendary film editor Walter Murch (Apocalypse Now, The English Patient) talks about having his assistants print out single frames from every scene in a movie. He looks for those painterly, emotional images that really define a character or a moment. He wallpapers his editing room with them so he can see the “DNA” of the film in a single glance. Walking around during breaks from his keyboard his eyes will flit from frame to frame. He trains his eyes to avoid the linear progression – left to right, top to bottom. Instead his eyes roam the walls at random, always searching to make surprising new connections between disparate characters and unrelated storylines.

What he’s after is some of the creative chaos he remembers from the days before modern digital editing. In the seminal book-length interview between Michael Ondaatje and Murch, “The Conversations,” Murch laments that digital editing prevents the “happy accidents” of the old analog days: the random images he’d stumble across while winding through a reel of film. And the surprising connections he’d make as a result.

As Murch says:

“The computerized system gives you exactly what you ask for. And that’s the danger. You don’t see the things you’re not looking for.”

Well, doesn’t that just perfectly reframe the “hallucinations” of AI? We often see AI’s randomness as a bug, but perhaps Murch would see it as a feature. The accidents break our rigid efficiency. They create connections we never would have seen. They brings the “happy accidents” back into the workflow.

I try to remember that optimistic reframe when I’m screaming at my monitor because the AI doesn’t understand simple English. Speaking of which:

LET’S DO OUR FIRST VIDEO PROMPT

First off, a word about the tools.

Throughout this guide, I have kept my tech stack as simple as possible. Specifically, I have been using Google’s Flow interface to generate both stills and video through their Nano Banana and Veo tools, respectively.

Is that because Flow is the “best” UI, and Google has the best models? Not necessarily. To me, the appeal lies is the simplicity – especially for a “basic” course. The interface is easy, and I like easy. If Flow isn’t your thing, you may prefer to work in Adobe’s Firefly, Higgsfield, Runway, or Artlist. All have more robust workflows. But the principles in this course will apply to whatever tool floats your particular boat.

Whatever model you prefer, when it’s time to generate video, there will be three main approaches on offer:

1. Ingredients to Video: The “ingredients” are the reference images you upload. A frame from your Storyboard Grid, a Character Sheet to lock in your character’s consistency, and a Location Sheet (if needed). As a part of your text prompt (more on this below), you’ll explain “Image 1 is the primary composition. Image 2 is a character reference. Image 3 is a location reference.”

2. Frames to Video: This mode allows you to specify your exact first and last frame (or just one frame), making “Frames to Video” the ideal mode for super specific actions. These are the “small things” we talked about in Part Two of this course – like our clown removing his nose.

   

3. Text to Video: This mode creates video from whole cloth. This is the “slot machine” method. It has its place, but it’s not what we are doing here.

Now Let’s Get Prompting. Here are the 6 basic ingredients for a solid video prompt:

• The Subject & Action (What is happening now?)

• Camera Work (Dolly left, smash zoom, static)

• Visual Style (Cinematic, Noir, Claymation)

• Lighting (Hard, soft, motivated)

• Environment (Windy, dusty, rainy)

• Technical (FPS, lens choice, lens flare)

I know that sounds like a lot, but because of our extensive pre-viz work, most of this information is already baked in. You don’t need to describe the clown’s jacket anymore; the character sheet does that for you. You don’t need to light the scene; your location reference does that for you. You don’t need to place the camera; your storyboard does that for you.

It does help to restate the style of your film in the prompt – especially if it’s something unusual like “tactile claymation miniature animation.” But that will be a text block that you write once and cut-and-paste into every prompt. The part that changes on a shot-by-shot level is (usually) just going to be the Subject, the Action and the Camera Work.

For this example, I’m using the “Frames to Video” approach, and I’ve uploaded the first and last frames seen above. My text prompt is simply: “The clown takes off his nose, slowly shaking his head. Shaky handheld camera pushes in on his face.”

Easy enough. Here’s the very first generation I got back:

Perfect? No. Problems? Sure! But I bet generating it another 10 times would have gotten me something that I like even better.

YOUR SECRET WEAPON: “Annotations”

One of the weirdest ways to interact with AI is also one of the most effective: Annotating your images with ugly colored boxes and writing the direction on the image itself. Oddly enough, this is the best way to get complex, specific action happening on different planes simultaneously. I’ll explain…

The Problem: If I have a lineup of three clowns and I prompt: “Left clown takes off nose, middle clown sits, right clown cries,” the model will likely panic and make them all cry. Or all sit. Or all cry. Or all those things at once.

The Solution:

1. Take your image into a photo editor.

2. Draw a box around each clown. Use high-contrast colors.

   

3. Write the instruction for each specific clown inside the clown’s box, in a text color that matches that box. Number each instruction so the AI knows which order to tackle them in.

   

   

   

4. Crucially, you must add a Master Instruction to your frame: “1. Delete the text and the three colored boxes, then pause 1 second before executing commands 2, 3 and 4 sequentially.” It doesn’t always work (of course!) but the idea is to make the AI delete your visual commands with a beat of separation before the action begins.

5. Finally, your text prompt to accompany the image is something along the lines of: “Using the exact reference image, follow the instructions written in the frame.” (You may also want to specify “no music, no dialogue.”)

All of this sounds insane, and there’s no question that this is a ridiculous way to make a motion picture. But check it out:

That was a whole lot of very specific choreography, and the AI performed pretty well. Not bad!

Here’s another test example. This is a frame from my Welcome To Wonderland trailer (If you haven’t seen it, hey, go on and give this link a click, won’t you?).

In it, I want the mushrooms to glow, the caterpillar to exhale a puff of hookah smoke, and her tail to light up like a lightning bug. Check it out:

Again – not bad! Exerting this type of granular control over the bot is exactly what we need so we can stop generating and start filmmaking.

IN CLOSING

As promised, this course has covered the (Very) Basics of AI Filmmaking. This should be enough to get you started executing high-value, cinematic AI filmmaking right away, using relatively straightforward tools and processes.

When the tools disappoint or frustrate you – and they will! – remember that this is the best the technology has ever been… but it’s also the worst it will ever be. The next generations of the tools are going to blow your mind, and change how movies are made.

And who gets to make them.

It’s a foregone conclusion that the tools & techniques I’ve described here will be totally obsolete in no time. Maybe even by the time you’re reading this. But what’s timeless here is the theory. The philosophy of WHY, the awareness of INTENTION, and the adaptability of HOW.

The tools are just tools. The understanding of how to tell a story with them is everything.

I hope this course has been helpful. Now, go make some great stuff.

Thanks for reading The AI Showrunner. If you enjoyed this course, share your own creations in the comments below.

And now you’re itching to start turning them into fully realized shots, scenes, and films.

First off: The Bad News.

We aren’t there quite yet. There’s still some crucial prep work to do. The maxim, “slow is smooth, and smooth is fast” is our guiding principle here.

Subscribe now

Sure, you could jump right into animating your characters from the assets we’ve built so far. That’ll work. And if the scope of your project is limited, you’ll probably do just fine.

But if you want to deliver high-value, cinematic results (without tearing your hair out) you’ve still got some prep work to do. This is the work that separates “AI Filmmaking” from “Filmmaking.” If you’re still reading, I know which side of the equation you want to be on.

THE PROBLEM WE’RE SOLVING

In my short Nun of the Above, I had to explain to the machine that I wanted a nun hovering in a floating chair, near the roof of a cathedral, inside the frescoed dome of a cupola, in shadows. Easy enough to imagine, but boy that’s a lot of geography to ask a machine to understand. Add in the fact that I wanted a nun and a cathedral and frescoes but I didn’t want any Christian iconography? This starts to feel like a doomed endeavor.

I hit this wall at full speed. I finally had some momentum and was really building the film and then, WHAM. Flow state, meet brick wall. It took hours to break the stalemate between man and machine. And that’s fine. That’s the work. But getting ripped out of the fun of the actual filmmaking in mid-flow? Awful.

So today we’re going to get ahead of that problem using a simple tool that I’ve written about before – and have since improved upon – with a new set of customizable, relatively intuitive prompts that you can start using today.

The step we’re focusing on here could be called “pre-production,” or “pre-visualization,” or “storyboarding.” It’s all about creating coverage grids that break every single action down into simple components. That’s because AI is good at two things: the big things and the small things. The middle ground is where the danger lies, so this step is all about breaking the medium-sized things into small things. Specific things like single images and single movements… images and movements that you can minutely control.

Because that’s what we’re after: control.

Let’s get to it.

THE STORYBOARD

I’ve written about the 3x3 Grid before. So have a lot of people. It’s a popular technique because it solves two massive problems: Scale and Consistency. There are plenty of videos about this on YouTube to check out. Just search “AI 3x3 grid” and say good-bye to your afternoon. But stick with me for a bit first…

If you read Part 1 of this course, you’ll remember how we created an emotional palette for our character.

Now, we are going to rinse and repeat that same basic concept to create proper cinematic storyboards. If you think of everything we did in Part One as the screenwriting part of the endeavor, this is where we really start directing.

A quick word on what that means before we dig into the nuts and bolts…

The AI haters (or the uninformed) think that AI is just luck, or repetition. They think it’s robotic, or that it removes human vision from the equation. This is where you prove them wrong.

And you do that by declaring yourself the author of these images before you even start prompting.

A true author knows the POV, the attitude, and the reason behind every shot. And how the shots relate to each other. Why they go together. A true author considers all the choices, with questions like:

• Why are we in a close-up?

• What’s in the background?

• Is the character looming over us, or is she dwarfed by the environment?

• What does the blocking tell us about the relationship between the characters, and how does it change during the scene?

There’s a word for this, and that word is filmmaking.

Sure, you can approach prompting like a slot machine and hope that 1 out of 9 images comes out right. That’s valid for brainstorming. But isn’t it more fun to see the movie in your head and then extract it from the machine to your exact specifications? To use the tool that can do anything to instead do exactly what you want?

Yes, that. Let’s do THAT.

Creating The Storyboard Grid

Just like a “real” film begins with a storyboard, an AI film starts with a Storyboard Grid. Depending on the complexity of your action you may need just one of these for a given scene, or you may need a dozen.

Let’s start simple, with a static, dialogue-driven scene. In this case you probably want a wide master, some mediums and close ups of your characters, some insert shots. Your story will of course dictate the specific needs, but for this illustration we’ll keep it pretty simple and ask for classical coverage. Nothing fancy.

Here’s a static scene coverage prompt I use, and which you will want to customize for your needs. For this example I am taking a shortcut and text prompting for characters and locations. But when you’re actually making a film you’d want to upload your Character Sheet and Location Sheet as image references.

Here’s the prompt:

A professional 3x3 split-screen grid, cinematic contact sheet. 9 distinct panels of the identical scene captured simultaneously from different angles.

Subject: [Insert your own description here, such as:] A 1950s astronaut sitting at the counter in a futuristic diner, being served a cup of coffee by a waitress with big 1980s hair.

Constraint: STARK CONSISTENCY. The exact same characters, same costumes, same lighting, same location, and same time of day in all 9 panels.

Grid Layout:

Row 1 (Context): Extreme Long Shot (Environment), Full Body Shot, American Shot (Knees up).

Row 2 (Coverage): Medium Shot (Waist up), Over-the-Shoulder, Frontal Close-Up.

Row 3 (Details): Extreme Close-Up (Eyes/Texture), Low Angle (Heroic), High Angle (Overhead).

Style: [Insert your own description here, such as:] Cinematic period drama, bright colors, high key lighting. Photorealistic, 8k resolution, cinematic lighting.

Technical: Distinct panel borders, no text overlay. The panels should look like a director’s monitor wall. Prioritize images, minimize borders between frames. Add no text.

Take a look at the result and see if you can spot one enormous problem with using a text prompt vs. a carefully curated Character Sheet…

Yeah… why the hell is my astronaut Ryan Gosling? I didn’t ask for that, and don’t want it. Rude. This is yet another example of why – if you want to commercialize your work – you need to start with an image generator trained on fully licensed images. As I mentioned in Part One there are “safe” options out there like Adobe Firefly.

Other problems are:

• First frame is letterboxed for some reason. Not ideal, but doesn’t really matter for our purposes. AI is great at fixing things like this. So: fine.

• No idea what’s going on in bottom-middle frame – total break of the geography.

Despite these issues, only one fatally flawed composition out of nine? Not bad. You can see how if I started with more specific camera direction and image references we’d be in good shape here.

Now let’s make things more complicated and more simple all at the same time.

The above example was a static dialogue scene. Let’s look at an action scene. When I say “action” I don’t mean an epic space battle (even if Mr. Gosling does look ready to deploy). I’m just talking about a simple, action-based moment. Even when you direct your epic space battle, you will be breaking it down into the smallest possible beats and creating a grid for each one.

This is the specificity we need if we want to truly control the AI models and get the results we want. (With the current models, at least.)

For demonstration purposes, let’s just go with a simple beat and build an action grid for it. The action will be: our sad clown coming home to the shabby motel room we designed yesterday, taking off his clown nose, putting it in his pocket, and sitting down on the bed.

Sounds simple, right?

But really that’s quite a lot of action. We have “coming home to the motel room,” “sitting down on the bed,” and “Taking off his clown nose and putting it in his pocket.” Depending on your plans for the editorial flow of the scene, you might actually want to do a 3x3 grid for each one of these individual actions. For this demo, though, we’ll go with the full action description and see what we can learn from the results.

Here’s the prompt I used:

A 3x3 storyboard grid showing a sequential action beat. 9 frames progressing chronologically left-to-right, top-to-bottom.

Action: [Insert your own description here, such as:] The sad clown enters the room and sits down on the bed. He takes off his red clown nose, and puts it in his pocket.

Flow:

• Panels 1-3: The preparation of the action.

• Panels 4-6: The execution of the action (climax of movement).

• Panels 7-9: The aftermath/reaction.

Consistency: Maintain identical character features, clothing, and background environment across all panels.

Style: [Describe your visual style. Lenses, cameras, lighting, vibe, such as:] Cinematic period drama, bright colors, high key lighting. Photorealistic, 8k resolution, cinematic lighting.

Technical: Cinematic storyboard, motion blur where appropriate, coherent lighting continuity, distinct borders between frames. Prioritize images, minimize borders between frames. Prioritize facial acting. No text overlay.

And here’s one of several generations off this prompt:

Initial thoughts:

• The top row: Geography and camera direction are all over the place. I hate it. Great argument for making “Clown comes home” its own action grid. I consider these frames mostly unusable.

• Middle row: Very frustrating continuity error here (he’s using his right hand in one frame and left in the next).

• But then the second two images in the middle row? Love. Those are gold. That’s exactly the specificity I’m looking for. With those two images I can tell the AI an exact start and end frame for the act of removing the nose. If I throw everything out but keep those 2 frames, I’m happy.

We’ll talk more about this in Part Three, but the next pro move is to take those images and craft them even further. Maybe you’d prefer him holding the nose up higher, or smiling sadly at it. When you are exerting this kind of granular force on your output by drilling down to this level of shot-specificity, you can bend the AI to your will.

You can also exert control over the output by getting a lot more detailed in the initial prompt. Telling the model exactly what you want in each of the nine frames will get you that much closer to achieving your vision.

And I haven’t even mentioned style or tone yet. Thats a whole other layer of control you’ll want to impose on your prompts. Below are some Cheat Sheets you can use to get started.

THE CHEAT SHEETS

The real art here is knowing the visual tone you’re going for. Ideally, you create your own visual language, but here are some handy “Style Packs” you can add to the prompts above to instantly dial in a specific vibe.

Please read these prompts as an invitation to expression, not a drag-and-drop substitute for creativity. Make them your own… that’s the fun part.

1. Film Noir

Lighting: Chiaroscuro lighting, harsh shadows, silhouettes, and dramatic shafts of light.

Style: Black and white (or muted Neo-Noir neon), film grain, moody, deep blacks.

Grid Layout:

Row 1 (Atmosphere): Silhouette Shot (framed against window), Dutch Angle (canted 45 degrees), Reflection (seen through mirror/puddle).

Row 2 (Tension): “The Squeeze” (framed through bars/blinds), Deep Shadow Medium Shot, Over-the-Shoulder into darkness.

Row 3 (Power): Extreme Close-Up (eyes/sweat), Extreme Low Angle (looking up from floor), Extreme High Angle (top down view with stark shadows)

2. The “Wes Anderson” Look

Framing: Center-punched framing. The subject is dead center in almost every shot.

Style: Flat lighting (low contrast), pastel color palette, sharp focus, sterile composition.

Grid Layout:

Row 1 (The Space): Extreme Wide (tiny subject in massive room), Flat Profile (mugshot style), Birds’ Eye View of the subjects.

Row 2 (The Portrait): Dead Center Medium (breaking fourth wall), Dead Center Medium reverse/POV shot, Symmetrical Two-Shot.

Row 3 (The Style): Crash Zoom start frame (wide of subject), Crash Zoom end frame (tight on face), Low Angle Center, Extreme Close Up of prop/character detail.

3. Cinema Verité / Documentary

Vibe: Raw, handheld, imperfect, reactive.

Style: High ISO grain, desaturated, natural lighting, motion blur, “found footage” aesthetic.

Grid Layout:

Row 1 (The Environment): Obscured Wide (dirty frame/blocked by foreground), Long Lens Snipe (telephoto/compressed), Tracking Shot (following back of head).

Row 2 (The Human Element): Snap Zoom Medium (slight blur), Profile (unaware/looking off-screen), “GoPro” Angle (fisheye lens as though camera is mounted on a prop).

Row 3 (Texture): Macro Grime (dirt on lens), Lens Flare (blown out lighting), Candid Close-Up (mid-speech).

The Bottom Line

It goes without saying that ChatGPT is a good collaborator here if you aren’t fluent in the language of cinema. But to me, creating these kinds of stylistic markers is where half the fun lives. Being an architect of the coverage, translating your vision into words then into images, dialing the results in until they’re just right.

That’s the place where I get my thrills.

And it’s also the place where we can stop using the term “AI Filmmaking” and just start calling it what it is: Filmmaking.

NEXT TIME ON…

Bet that when you started reading this series you didn’t think it was going to take THREE articles before we created a single second of video. Sorry about that, dear reader. We’re almost there, I swear it.

And I promise you, everything we’ve talked about so far will:

a) Help you fully realize your vision…

b) Make things both easier and much more fun when you get to the next step…

c) Bend the machines to your mighty will…

d) Help you create something that isn’t an “AI film,” but an actual “film-film.”

And I promise, in Part 3, we’ll FINALLY begin putting all this hard work to use and start crafting some video.

Subscribe below to get the update as soon as Part 3 goes live!

Subscribe now

Welcome to my 3-part course in AI filmmaking.

This series will detail the basic steps and prompts I use to make films with AI from the ground up. But it isn’t just about prompting – it’s also about storytelling. Let’s call it 40% film school and 60% interacting with AI.

None of this is complicated. None of this is magic. All of it is accessible to you right this second. The steps can be a bit unwieldy because the current generation of tools is unwieldy. But I promise this: try these steps once, and you’ll be able to rinse, repeat, and start making your first AI film immediately.

All you need is a good idea and a solid POV. The AI can’t make that for you.

Subscribe now

We are going to break this course down into three chapters:

1. The Foundation: Character & Location Design

2. Building the House: Storyboarding & Consistency

3. Time to Shoot: Video Prompt Crafting

Today, we start with The Foundation.

1. CHARACTER DESIGN

Your Cornerstone

If you want your film to have a soul, the first step is great character design.

Before you even begin prompting, you should be able to see the character in your head. Think of yourself not as a prompter but as a screenwriter… because that’s exactly what you are. Don’t relinquish this role to the robots. This is the fun part.

Ask yourself: Who is your character? How do they dress? What baggage do they carry with them, and how is it revealed in their posture, their clothes, their features? What do they think they’re revealing about themselves? And what do they incorrectly think they’re hiding from view? Is your character’s dress expensive but fraying because she’s clinging desperately to a once-glorious, now-faded success?

That’s a lot to cram successfully into an AI prompt. But high-value storytelling is not just about the details, but about the intention behind them. Getting to that intention is the point. That’s where story lives.

Build your world by seeing your character before you create them.

The AI may surprise you and take the character in a different direction. That’s fine – that’s how writers’ rooms work too. But if you have your “north star” already in mind, you’ll be able to look at the results and know immediately if your AI “collaborator” is on the right track. AI filmmaking is not about surprises; it’s about curation.

The “Morgan Freeman” Trap

As I wrote in a previous post, during my project Nun of the Above, I accepted a character sheet that looked more than a bit like Morgan Freeman. It wasn’t exactly what I wanted, but the model’s insistence made me laugh, so I said, “Okay.”

Big mistake.

What I didn’t realize was that once I got to the animation phase, the video model (Veo) would pick up on that resemblance and accentuate it at every turn.

Lesson learned: Make your character sheet perfect before you move on. And if you plan to monetize your work, protect yourself by using a model that is trained on fully licensed images, like Adobe Firefly or BRIA.

The Character Sheet Prompt

A good character sheet shows your character from front and back, with inset features of key details. I prioritize a neutral expression here – we will customize emotions later. This gives you a clean building block.

Crucially, this is also where you set your Tactile Visual Tone. Is the style photorealistic or claymation? Muted colors or garish? Film grain or crisp CGI? Lock those stylistic choices in now.

The only element we don’t want to style heavily here is lighting. We want a clean studio setup for maximum clarity. Lighting will come later.

[Note: throughout this guide I am creating my images and videos in the Google Flow interface, using their NanoBanana and Veo 3.1 tools. Is Google the best solution? No… but it’s where I happen to have cheap credits to burn. The techniques I describe in this course should work just fine on whatever model you prefer.]

Here is the prompt structure I use. The fields that you will primarily want to customize are in [brackets].

Full professional character sheet, studio presentation.

Subject: [Character is an elderly circus clown. Tired eyes, makeup fading and cracked in places. Wardrobe is a rumpled, garish, harlequin clown suit. Slight stains down the front.]

Views: One primary full-body front view and full-body back view of the same character, neutral stance, neutral expression, symmetrical lighting. Include small inset detail panels showing hands, footwear, and one wardrobe detail.

Setting: Clean, light neutral background (off-white or gray). No text, no labels, no logos, no UI elements. No background elements, no props unless integral to the character.

Technical: Precise anatomy, realistic proportions, consistent character identity across all views. Wardrobe clearly visible and unobstructed. Leave negative space suitable for later annotation. No background elements, no props unless integral to the character. Neutral color grading, accurate skin tones, fabric textures readable.

Lighting: soft studio key light, minimal shadow, high clarity.

Style: professional concept art / production character sheet, [cinematic realism. No film grain.]

Camera: very low-distortion, straight-on views. High resolution, sharp focus.

I batch generated 4 images with this prompt, and here’s the one that came closest to what I was envisioning:

If this wasn’t just for demonstration purposes I’d get pretty granular tweaking the image before accepting it. I’d also get way more detailed in the “Subject” field of the initial prompt. But for our purposes let’s consider this guy “almost there.” I do want to change the hair and the shoes, though. If you’re using Google’s Flow interface like I am, all you need to do is click on the generated image, which will bring up the “Edit Image” screen. Enter a simple prompt like:

Change his hair to a droopy, weathered clown wig; and make his shoes bigger and sillier.

And there you have it:

All of the image generators have gotten really good about (usually) only editing the areas that you tell them to. So if you’ve tried this in the past and been flummoxed, it’s time to try again to see how far the tools have come.

For our purposes, this clown is ready for his close-up.

Don’t Skip This Step: Upscale your character sheet to 4k and download it. You want to carry the most fidelity possible into the next stages. If you want to go nuts, and have like 30 bucks to spare, you can buy a subscription to Topaz to upscale past 4k, or play with their “realism” filter. That’s a whole rabbit hole to fall down, but the point is: Resolution is your friend. Perfection at this stage is your friend.

2. EMOTION DESIGN

So, you’ve got your character. But he’s a zombie. He has no emotion. That’s by design… until now.

Next, we want to create a palette of basic expressions. I prefer to create a 3x3 Grid—nine individual expressions all on one sheet. This locks in character consistency and saves you AI credits.

Tell your AI tool to reference your Character Sheet. In Flow, use it as an “Image Prompt” by clicking the little plus sign in the Image Generator field.

And then run this Emotion Grid Prompt:

Format: A 3x3 character study grid. 9 distinct panels of the same character in the same lighting conditions, formatted as an actor’s contact sheet. Clean, light neutral background (off-white or gray). No text, no labels, no logos, no UI elements.

Subject: Use Image 1 as an exact character reference. All shots are Medium Close-Ups (Chest up).

Expressions:

• Row 1: Neutral, Slight Smile, Joyful laughter.

• Row 2: Concerned, Suspicious, Angry.

• Row 3: Surprised, Fearful, Exhausted/Sad.

Style: Portrait photography, high fidelity, 85mm lens look, perfect facial consistency, strictly identical clothing in every frame. [Photorealistic], high resolution, cinematic lighting, distinct panel borders. Cinematic storyboard, coherent lighting continuity, distinct borders between frames. The panels should look like a director’s monitor wall. Prioritize images, minimize borders between frames. Prioritize facial acting. No text overlay.

Again, upscale this grid to 4k. Since each face is only 1/9th of the image, you need every pixel you can get. Set this file aside – it will become your bible later.

3. LOCATION, LOCATION, LOCATION

The irritating Hollywood cliché is that “the location is a character.” It’s an annoying trope (and an even more annoying network note) but there’s some truth to it.

But the location isn’t a character… the location reflects your character. Think of the chaos of the mansion in the opening scene of Kubrick’s Lolita, full of covered up secrets, hedonism, and shame:

There’s so much going on in that scene. It looks like chaos. But every detail reflects something about what’s going on in the characters’ souls. What’s happing inside them is visible everywhere.

Before typing a single prompt, you should envision your location as specifically as you envisioned your clown. What are the colors? What are the textures? What is the weather? And why? What story is the location built to tell?

And Finally: The Lighting.

Where we avoided subjective lighting in our character sheet, here we want to indulge it. This is where we (finally!) start setting the cinematic mood, choosing colors, and exploring the tone. Here we are creating the environment that our characters will inhabit, and set the tone that will inform our storytelling.

We’ll start just like we started the Emotion Grid. Though here I will use only 4 frames to allow for greater detail. If you don’t need this many angles, by all means go full-frame on these generations.

Here’s the Location Concept Prompt I’ve been using:

Subject: Cinematic location concept art for a film environment.

Single cohesive location: [a shabby motel room, with tacky art on the walls, peeling wallpaper, threadbare carpet, low incandescent lighting. Nighttime. Some street light coming through the window.]

Format: 2x2 grid showing four angles of the same location: primary angle, reverse angle, top-down view, side angle.

Details: Architecture and spatial layout clearly readable. Environment feels lived-in and story-bearing. No characters in frame. No text.

Style: Professional production design reference, film-ready. Intentional cinematic framing. Composition prioritizes mood, scale, and narrative implication. High resolution, sharp environmental detail. Foreground / midground / background separation present. Surfaces and materials believable and tactile. Clean frame edges, no UI, no logos, no watermarks.

Obviously you will want to expand upon the style you’re going for, the same way you did in the character sheet. Is it a photorealistic look, CGI, etc? As always the more detail the better. Give the AI a good brain dump and it will be happy to receive.

So here’s what this prompt generated:

Not bad for a first iteration. There’s plenty I’d want to customize, and as always, AI struggles with perfect geographical consistency. For instance, the bed is on a different wall in almost every angle – yikes. But this was the first attempt. Try the same prompt 4 more times and you’ll get something you can call a keeper.

You might think of all this iterating as proof that the AI is random and weak… and you wouldn’t be wrong. But I prefer to see it differently: this is the same kind of ideation and iteration I’ve seen on every production I’ve ever been a part of. Your production designer throws a bunch of ideas at you: some are great but wrong… some are terrible and wrong… and one or two will be not only right, but on the trail of perfection. Your job is to chart the course.

Each iteration shouldn’t be viewed as the end of creativity, but as an invitation to expression.

UP NEXT: Building the House

Now we have our Cast (The Character Sheet & Emotion Grid) and our Set (The Location Concept).

In Part 2, we are going to put them together. I’ll show you how to use Storyboarding to combine these assets into coherent scenes before we ever generate a single frame of video.

Stay tuned.

So today, I want to stop and look at the elephant in the room: JOBS.

Subscribe now

As I discussed in a recent note, we have two choices right now:

1. Worry about the consequences.

2. Prepare for the possibilities.

Let’s all choose option two.

This is going to be a landmark year for the entertainment industry. While the job market is going to change, I believe that change will bring an explosion of new opportunities rather than just the deletion of old ones.

In no particular order, here is my forecast for a few of the Hollywood trades.

The Actors

If you haven’t seen the latest “performance capture” tools – specifically the new Motion Control features in models like Kling 2.6 – get thee to YouTube, or just go straight to the source at the Kling website.

Until now, Motion Capture was a specialized, expensive ordeal involving spandex suits and sensors balls glued to Andy Serkis. But for our uses? The game has changed. The tech is now available to everyone.

But this isn’t just about cool toys. It represents a massive shift for actors. For the first time, actors can be cast for personalities, not just faces.

Every actor has heard a casting note like, “great energy, but not a leading man.” In AI filmmaking that’s going to be a thing of the past – and good riddance. We all know that “not a leading man” has often been barely coded racism or narrow-mindedness. That’s not a hot take. That’s just true.

• The Tech: We can now capture a nuanced human performance and map it seamlessly onto any character model.

• The Opportunity: A tremendous new pool of talent becomes accessible to filmmakers. And a tremendous set of opportunities becomes accessible to actors. A performer who doesn’t “look the part” can now be the part. “Looking the part” is irrelevant. Performance is everything. Heart is everything.

My Prediction:

I’d bet a dollar that someone is already developing the “Match.com for Generative AI.” Imagine an app that pairs performers from around the world with AI filmmakers working at all budget levels, tracks the use of the performers’ work, and processes profit sharing in real-time via smart contracts. Anyone. Anywhere.

(If no one is building that: Get going. And dibs – I get 10%.)

The drawbacks will of course be the same as the drawbacks of shooting movies on blue screens. And, yeah, those are significant. No actor likes acting with drapery for a scene partner. And who could blame them? Balancing this is the tremendous, democratizing opportunities to be created. I cast that as a net win for performers.

The Visual Departments

Broadly including Cinematography, Production Design, Costume, Props, Hair, Makeup, VFX, and others.

We are moving toward a world of cheap, accessible, full 3D environment renders generated from a single still frame.

More importantly: these will be worlds where every element is discrete and manipulatable in real-time.

(If you want a fun term to drop at parties, look up Gaussian Splatting. I learned it recently and I can’t stop trying to work it into conversation.)

• The Tech: We will be creating fully navigable, photorealistic 3D scenes where you can say, “Hey set dressing, change the drapes,” or “Hey production design, move that window north for better lighting,” and see it happen instantly. And the changes will happen on the final product, not on a sketch or mock-up. The tools already basically exist… but they’re about to get really, really useful.

• The Opportunity: Just like in traditional filmmaking, all the details matter, and it takes a team of creatives to populate the space. But now instead of shopping for shirts at the Glendale Galleria, wardrobe designers will be building their own digital couture catalogs that live in the cloud. Prop masters will license out their unique digital toolboxes. Individual artists will no longer just be “work-for-hire” crew; they will be asset creators with portable portfolios that they own and license to productions. You won’t just bring your skills to a job; you’ll bring your IP library.

Post-Production

“Fix it in post” is about to be a thing of the past. As a former editor, I am beyond cool with that. Instead, we are entering the era of “Make it in Post.” (As if editors’ egos aren’t big enough already...)

• The Tech: The editor says, “I need a shot of the dog looking left,” and the AI generates it. Perfectly matched lighting. Perfectly matched dog. Why is this so important? For non-editors: Inserts are the lifeblood of scene construction. They’re used to save performances and pace (and add meaning and drama, too). Yet, traditionally, when you run out of time on set, inserts are the first thing to get deleted from the schedule. And that’s a potentially critical error.

• The Opportunity: The AI edit bay becomes a creative cockpit where those missing pieces can be generated on demand. Screenwriting, directing, and editing will begin to fuse into one.

This might be tough news for 2nd Unit Directors, but it is incredible news for Editors, Writers, and Directors.

The Lifestyle Shift

I know a lot of us are preemptively mourning the loss of our way of life. This is a 100% valid emotion. But let’s look at that “way of life” critically.

Filmmaking has always been about trying to do too much, with too little, in too short a time. Hence: the 12-16 hour workdays.

I’ve spent a fair share of my life in those trenches. But I have to ask: Is this really the status quo we want to fight to preserve? Just because it’s the way it’s always been... does that make it healthy? Does that make it right?

Instead, let’s imagine a world where an Art Director collaborates on a production from anywhere. Tweaking a 3D set while living somewhere affordable and full of joy, rather than a 2-bedroom condo in Encino. (No hate for Encino – but it ain’t Tuscany).

One last thought about the old ways. Even if the old ways aren’t necessarily great ways, they aren’t going anywhere anytime soon. There will be reductions, yes. Painful ones, at that. As we all know: this has already happened. But this shift doesn’t represent the end of Hollywood as we know it. After all: physical filmmaking is fun. And as long as Tom Cruise still wants to do his own stunts, “real” shooting isn’t going anywhere. As improbable as it seems, I think Tom still has a few years left in him.

Change is upsetting, but I’d argue that instead of grieving, our energies are better spent anticipating (and building) the new opportunities.

The Bottom Line

I know the view through my rose-tinted glasses is a bit overly simplistic.

“Changes to the job market” is often a euphemism for layoffs. I know that, too. I’m not blind to the pain. No one who truly loves this business is. But in this case, I believe that the change is also a promise of new capacity.

When we lower the friction of production, we all get to make more stuff. And hasn’t making-more-stuff always been the goal?

Imagine never turning down a project because of scheduling. Imagine collaborating with artists you love because location is no longer a barrier. Imagine being able to be a one-man/one-woman band, making that passion project that no one was willing to fund.

If you’re afraid of AI, remember: The tools are dumb. They have no soul. That’s where YOU come in. It needs us.

And now, more than ever before, the possibilities are ours to create.

Thanks for reading The AI Showrunner. If you’re ready to start preparing for the possibilities, hit subscribe below.

Subscribe now

#FutureOfWork #Hollywood #EntertainmentIndustry #CreativeEconomy #GenerativeAI #AIFilmmaking #PostProduction #VFX #ActorsLife #ProductionDesign #GaussianSplatting

If you haven’t seen the film yet, please give me six minutes of your life by checking it out right here.

For the curious, the skeptics, and the fellow creators whose eyes are currently bleeding from endless AI prompting: Here are the specs.

Subscribe now

The Concept: A Mystery from 1984

The film was inspired by a specific page from the classic children’s book The Mysteries of Harris Burdick by Chris Van Allsburg.

For those who didn’t grow up with it, Harris Burdick is a legendary creative writing prompt disguised as a picture book. The premise is that a mysterious artist left a series of drawings with a publisher, along with a title and a single line of text for each, and then disappeared. The stories were never written. It is up to the reader to imagine them.

I spent hours as a kid flipping through those pages, doing just that.

My first couple of public-facing AI projects were both pretty noisy, fast-paced trailers which you can view here. I knew I wanted my next AI project to be different. I wanted a slower pace. I wanted a narrative structure that couldn’t hide behind quick cuts. And most importantly, I wanted to move away from the “glossy sci-fi” look that dominates so much of generative video right now. And I didn’t want to try to compete at the “photorealistic game.” I wanted something that didn’t look “photoreal”… I wanted something that looked really real.

I wanted something tactile. I wanted characters that felt like you could touch them.

Which is how I got to claymation. My earliest filmmaking experiences were making Super 8 animations as a kid. Somehow, combining my nostalgia for Van Allsburg’s book with my nostalgia for the stop-motion form just made sense.

The Writing: The AI Writer’s Room

As far as the written word goes, I always think of AI as a collaborator rather than a writer. I treat my interactions with the machine like a virtual writer’s room: as long as you train it on the elements of drama and teach it to stop kissing your ass, bouncing ideas around with Gemini can be incredibly productive. There are good ideas and bad ideas, flat-out dumb ideas, and also weird ideas that lead you to right ideas.

I love doing the writing myself, but as an experiment, I wanted to see how Gemini would do with the vagueness of Van Allsburg’s prompt.

So I uploaded the image from the book, and intentionally kept my own prompt pretty vague as well. I told Gemini about the “tactile, heavy, slightly ominous but whimsical” tone I was looking for, and asked for a narration that could be read aloud in under 3-minutes.

I hit send and Gemini got to work rocking my world.

That first draft was evocative, sensory, and weird. It nailed the skeleton of the piece immediately. A lot needed to change, but even before I finished reading the draft I already knew this was something I could work with. I was going to make this film.

[Mind you: I’ve had many, many more AI brainstorming sessions with precisely the opposite reaction.]

That said, by the final draft the themes and attitude had changed dramatically as I made them my own. Gemini tells me that I only retained 32% of its original draft (love the specificity). I’m not bragging when I say that the true soul and deeper meaning of the piece came entirely from me (and my co-creator, Robert). But I will very happily admit that the basic structure – and the framing of the inciting incident – came from the AI. It was an excellent launch point. It was the proof of concept I needed to confidently move forward. And fast.

I’ve never written a first draft of anything from a screenplay to a Substack note where more than 32% of the ideas survived from the first to the final version. My analysis: Gemini killed it… at least in the early framing of things.

(If you’d like to see the original raw output versus the final script, let me know in the comments.)

The Specs: The ROI on Time

So how long does it take to make a 6-minute claymation film with AI?

There were times – usually when I was failing to make the video model understand the meaning of the work “up” – where it felt like it took forever. But looking at the calendar, the speed is actually terrifying.

• Concept Discussion: December 26

• Character Design: December 27

• A Day Off: December 28 (Hey! It was the holidays!)

• Production Start: December 29

• Picture Lock & Mix: January 12

Total Time: 17 Days. Including that day off!

Total Labor: Approx. 60 Hours.

Not counting the time spent pacing and thinking (sometimes cursing), that averages out to 10 hours of labor per minute of screen time.

When you think about what it would have taken to produce this the old-fashioned way? Yikes. At 24 frames per second, we’re talking about posing and shooting roughly 8,640 individual frames. I haven’t been that 12-year-old in the bedroom with a Super 8 camera for a long time, but I’m pretty sure doing this practically would have taken months.

And How About the Budget?

Here are my hard costs, not including the aging hardware I already own.

• Google Pro Ultra: $125/month (they’ve been running a 50% off deal for 3-months at 25,000 generative credits/month). All images and animation were produced through Veo, as well as a lot of the sound design.

• Firefly: $9.99/month. I’m told this is a limited time price, and it’s a STEAL. (The “generative extend” function alone saved me hours, and more importantly saved some otherwise great shots that Veo just couldn’t stick the landing on).

• ElevenLabs: $22/month. All music and voices were created here, except for the voice of Pi the cat, who simply cannot be replicated by a mere machine.

• Premiere Pro: $39.50/month.

• Topaz (for upscaling to 4k): $37/month plus another $40 for extra cloud computing credits.

So: that’s a grand total of $274.49. Less than 50 bucks a minute, which in my book is pretty good.

The day after I released the film, I submitted it to a handful of film festivals. Not a ton: only six. One of them even gave me a fee waiver as a returning alumnus (thanks, Seattle International Film Festival!)

These six submissions priced out at $279.50. Which you may have noticed is $5.01 more than the film cost to make.

Hilarious. Maybe we should all get together and start a film festival. Is that where the REAL money is?

The Takeaway

Here’s the real treat in all of this, and the reason I am optimistic about 2026:

Without these tools, there is exactly zero chance I would have made this little film.

The barrier to entry – the time, the cost, the physical space required, learning to sculpt clay figurines! – was too high. Never gonna happen.

So what a gift from the droids! They allowed me to work in a medium I loved as a child, or at least a reasonable simulacrum of it, that I otherwise never would have touched again.

That’s just so freaking cool.

And so well worth my fifty bucks a minute.

Thanks for watching, and thanks for being part of this strange and wonderful journey with me.

Share

#AIFilmmaking #BehindTheScenes #CreativeProcess #GenerativeAI #StopMotion #Claymation #HarrisBurdick #IndieFilm #BehindTheScenes #FilmFestival #FilmmakerLife #ShortFilm #LowBudgetFilm #ProductionBudget #GoogleVeo #Gemini #AdobeFirefly #ElevenLabs

Subscribe now

Subscribe now

This project was an experiment in pushing AI video to replicate the tactile, imperfect feel of stop-motion claymation. It’s a story about a mysterious donation, a parish with secrets, and what happens when the laws of physics – much like faith – become merely a suggestion.

I’d love to hear your interpretation of the ending in the comments. I’ll be diving into the filmmaking philosophy and the logistics in the coming days, so please stay tuned.

In the meantime, watch the full film here:

Thanks for watching. If you choose to like, subscribe or comment, I promise that both the algorithm and I will greatly appreciate it.

#AIFilmmaking #Claymation #ShortFilm #Storytelling #GenerativeAI #IndieFilm

Share

Somewhere, right now, in a messy apartment or a quiet coffee shop, the history of cinema is being rewritten.

The first truly great AI filmmaker is sitting at their laptop. They are working on the first AI feature film that will “come out of nowhere” and change everything. They are tinkering with prompts, fighting with frame rates, and wrestling a machine into submission.

I suspect there isn’t just one of them. There are thousands.

Maybe one of them is you.

Subscribe now

We are staring down the barrel of a massive cultural shift. The question isn’t IF we’re going to see an aha moment that forces us to reconsider whose voices get our attention… it’s WHEN?

Is this major new voice going to come out of left field in the next six months? Possible but still a bit soon… I give it a 10% chance.

In the next year? Call it 25%.

Extend the timeline out to five years? It’s a sure thing. This is going to happen. Some unknown entertainment industry outsider is going to be the next great feature filmmaker. And change everything.

And soon.

The Tipping Point is Here

As you may know, I’ve recently started creating and sharing cinematic storytelling made entirely with AI (see my work here). The films I am making right now were technically impossible six months ago.

And frankly, they’ll be obsolete six months from now.

I speak regularly with people who have insider access to the next generation of tools being developed by the major players. They all say the same thing: Get ready.

We are about to see more controls, better performances, locked-on character consistency, massively upgraded tools for human collaboration between departments, and the power to surgically revise the most complex generated video. At scale, and at speed. The boundaries that are frustrating you today are about to dissolve. And the filmmakers who are currently struggling against those boundaries are going to hit the ground at a full sprint when the walls come down.

How do you get there? How do you not fall behind?

The Key is to Tune Out the Noise

I know how it feels to open your browser and feel immediately behind the curve.

You blink for a week, and your tech stack is outdated. A new model drops at 2:00 AM on a Tuesday. Your favorite app changes its interface overnight. It feels like a full-time job just chasing AI’s tail.

But here is the secret: Everyone is chasing the tail.

Anyone who tells you they have this fully mastered is either lying or just quaintly uninformed. So, take a breath. Tune out the noise. Stop obsessing over every white paper and new app feature and focus on your workflow. Find the tools that speak to your style and go deep on them.

That is what the next Spielberg is doing right now. They aren’t trying to use every tool; they are trying to tell one story.

The Call to Action

The barrier to entry has never been lower. And yet beyond the gates lie the most powerful storytelling tools ever created by human hands. They are infinite. They are miraculous.

But they are also dumb. They have no soul. They have no trauma, no joy, no heartbreak, and no perspective.

They have everything they need to change the world, except for one thing.

You.

So stop worrying about being late. You’re right on time. Now, get back to work.

Over the Christmas holiday, my creative partner and I were brainstorming visuals for our next film.

If you’ve been following along, our first two releases (Welcome to Wonderland & Human Resources) were quick cutting teaser trailers. This time, we wanted to slow down. We wanted to tell a story where the viewer is expected to connect with a character for longer than a fleeting 90 seconds.

Subscribe now

We settled on a stop-motion claymation aesthetic. There is something undeniably fun about bringing the highest-tech storytelling tool in history (generative AI) to the lowest-fi medium imaginable (clay).

As part of our visual research, we watched a lot of award-winning short films. They were stunning, tactile, and obviously labored over with love. But watching them, I had a realization that I suspect many creators are having right now: I don’t know if I can go back.

Having spent a decent percentage of my lifetime on or around sets, I know the blood, sweat, permits, and budget required to get the shot. Once you’ve experienced the “speed of thought” workflow that AI offers, the idea of returning to the sheer logistical heaviness of physical production feels... unattractive.

Traditional stop-motion requires an infinite amount of patience – a virtue I seem to have lost somewhere between working with the Vanderpump “kids” in West Hollywood and my time in a Serbian soundstage for Martin Scorsese Presents: The Saints.

The barrier to entry with AI is so low – and the feedback loop is so fast – that the traditional “Army of People” model of filmmaking is starting to feel like a distant memory. I respect the hustle, but I think I’m done with the heavy lifting. With AI, we can capture the “hand-animated” claymation aesthetic, but with the ability to iterate in minutes rather than months. It’s hard to un-ring that bell.

(Of course, I realize that “heavy lifting” is really just code for “jobs.” There is a massive, complex conversation to be had here about the economic threats – and the new opportunities – that AI is creating for the industry workforce. It’s the elephant in the room, and I promise we aren’t going to ignore it. But that topic deserves more than a footnote, so I’m saving that deep dive for a future post.)

But back to our research for the new project, which sparked a debate about the two biggest buzzwords of 2025: Micro-Dramas and Verticals.

I’m willing to bet that by this time next year, the “Micro-Drama” hype will have faded. But “Verticals”? They aren’t going anywhere.

The Widescreen Dogma

As a filmmaker of a certain age, I have spent my entire career telling stories in widescreen. It is the visual language of cinema. It’s also biology: Our skulls are built with two eyes, side by side. We scan the horizon. We look left and right before crossing the street. We (usually) read left to right.

To a traditionalist, filming vertically feels like a violation of human physiology.

I remember back in 2014, when I was a Co-Executive Producer on Season 4 of Vanderpump Rules. We were just starting to incorporate the cast’s social media posts and self-shot phone footage into the episodes. It was absolute murder trying to get them to remember to turn their phones sideways to match the real cameras’ aspect ratio.

At the time, it felt like sacrilege to even consider airing vertical footage on a cable show. We were formalists. We were unwilling to bend.

Cut to 2025.

Half the industry is still offended by the vertical frame. The other half is looking seriously at 9:16. But while the industry fights about aspect ratios, does the audience even care? Is the 9:16 frame actually an insult to cinema?

“The Wire” and the “God” of Composition

I’ve been revisiting a wonderful essay by David Simon (creator of The Wire) from a few years back. HBO had decided to remaster The Wire into 16:9 widescreen to fit modern TVs, but they forgot to consult Simon. They assumed he didn’t want to be bothered, and wouldn’t care.

He cared.

Simon argued that changing the aspect ratio fundamentally changed the story. He noted that the 4:3 ratio of the original show created a sense of entrapment – visualizing the way the characters were stuck in the systems of Baltimore. Opening up the frame to widescreen added “dead space” that diluted the tension.

As Simon wrote:

“There are many scenes in The Wire where the composition was not merely about what was in the frame, but what was excluded... To open up the shot... is to lose the claustrophobia that was essential to the scene.”

The lesson? The frame isn’t just a container. The frame is the story.

So, what is the story of the Vertical Frame?

Obviously, it works for “tall things”– runway models, skyscrapers, rocket launches. You may have noticed (or maybe you didn’t) that The New York Times has quietly transitioned much of their video journalism to vertical. It works beautifully for the “interview” format: a single face, speaking directly to you. It feels intimate. It works so well that you don’t give it a second thought.

But can it work for narrative?

The Flying Nun and the Woodblock Print

For our new claymation project, we are drawing inspiration from The Mysteries of Harris Burdick by Chris Van Allsburg.

If you don’t know the book, it’s a cult classic of children’s literature. It consists of a series of surreal drawings that imply a larger, often slightly creepy, story. The images come with a title and a subtitle, and nothing else. They are an invitation to creativity. One of the most famous images features a nun sitting in a chair, floating up into a cathedral.

A nun floating up.

That is a vertical story. It begs for a vertical frame. And yet, as we experiment with the concept, old habits die hard. I still find myself generating the shots in widescreen, telling myself we will “crop it later” for the phones. This is an inherent resistance to the format I claim to be interested in, I know.

But then, last week, I went to LACMA (The Los Angeles County Museum of Art), and I saw something that snapped the whole debate into focus.

I was looking at 18th and 19th-century Japanese woodblock prints by masters like Utagawa Kunisada, Kawase Hasui, and Suzuki Harunobu.

We tend to think of vertical storytelling as a symptom of the “TikTok Era” – a degradation of art caused by our addiction to phones.

But looking at these prints, I realized that was total nonsense. Artists have been using the vertical format for centuries. They used it to emphasize isolation, to show hierarchy (the high and the low), and to capture the beauty of rain falling or trees growing.

They didn’t paint vertically because they forgot to turn the canvas sideways. They did it because the composition demanded it.

The idea of a filmmaker shooting vertically isn’t the heresy we traditionalists think it is. It’s actually part of a venerable tradition. Maybe 9:16 filmmaking doesn’t have to mean copying TikTok influencers; maybe it can mean channeling Ukiyo-e.

So, for 2026, I’m making a resolution. I’m going to stop treating the vertical frame like a compromise, and start treating it like a canvas. As with Chris Van Allsburg’s book, I’m going to consider it an invitation to creativity.

Hey, if vertical composition was good enough for the Japanese masters, it’s more than good enough for an AI claymation nun.

Happy New Year and keep creating.

To misquote Home Alone 2: Merry Christmas, you filthy animals.

I wasn’t planning on posting today. I figured I’d be drinking eggnog and ignoring my laptop. But then the internet gave me a gift, and I feel compelled to pass it along.

Subscribe now

First, a disclaimer: If you are a casual reader here for the fun videos and storytelling essays, feel free to skip this one. Go enjoy your family. This post is pure, uncut technical nerdery for the builders in the audience who are currently staring at a blinking cursor, and slowly going mad trying to get an AI character to look the same in two consecutive shots.

The Problem

I spent the last 48 hours banging my head against the wall. I was trying to build a scene for a new project, and the AI tools were simply refusing to cooperate. I’d get a great wide shot, but the close-up would look like a different actor. I’d get the actor right, but the lighting would shift. You know the drill. I thought I’d beaten this particular problem… but this week I’ve been striking out.

I was ready to quit. Then, I found a random YouTube video (shoutout to the algorithm), which referenced a Reddit thread (shoutout to the anonymous Redditor), which contained a prompt that completely unlocked my workflow.

The Solution: The “Cinematic Contact Sheet”

The concept is simple but brilliant. Instead of asking the AI for one image, you ask it for a 3x3 Grid containing 9 specific camera angles of your scene.

Because the model generates all 9 images simultaneously in one context window, the consistency is nearly perfect. Same lighting. Same actor. Same vibe.

The Workflow

I’m using this in NanoBanana Pro but the logic should hold up in other multimodal tools.

1. Upload your Reference Image (your character or location).

2. Paste the Prompt below.

3. Fill in the “SCENE INPUT” at the bottom with your specific action.

4. Hit Run.

The AI spits out a 9-panel grid. From there, you use that image as a reference in a new prompt, asking simply: “Extract image from row 1, column 2” (or whichever shot you like).

Suddenly, you have your wide, your medium, and your close-up, all ready to animate. And all perfectly consistent (well… mostly consistent.)

Here’s the prompt for you…

The Prompt (Copy/Paste this):

<instruction> Analyze the entire movie scene. Identify ALL key subjects present (whether it’s a single person, a group/couple, a vehicle, or a specific object) and their spatial relationship/interaction. Generate a cohesive 3x3 grid “Cinematic Contact Sheet” featuring 9 distinct camera shots of exactly these subjects in the same environment. You must adapt the standard cinematic shot types to fit the content (e.g., if a group, keep the group together; if an object, frame the whole object):

**Row 1 (Establishing Context):**

1. **Extreme Long Shot (ELS): The subject(s) are seen small within the vast environment.

2. **Long Shot (LS): The complete subject(s) or group is visible from top to bottom (head to toe / wheels to roof).

3. **Medium Long Shot (American/3-4): Framed from knees up (for people) or a 3/4 view (for objects).

**Row 2 (The Core Coverage):**

4. **Medium Shot (MS): Framed from the waist up (or the central core of the object). Focus on interaction/action.

5. **Medium Close-Up (MCU): Framed from chest up. Intimate framing of the main subject(s).

6. **Close-Up (CU): Tight framing on the face(s) or the “front” of the object.

**Row 3 (Details & Angles):**

7. **Extreme Close-Up (ECU): Macro detail focusing intensely on a key feature (eyes, hands, logo, texture).

8. **Low Angle Shot (Worm’s Eye): Looking up at the subject(s) from the ground (imposing/heroic).

9. **High Angle Shot (Bird’s Eye): Looking down on the subject(s) from above. Ensure strict consistency: The same people/objects, same clothes, and same lighting across all 9 panels. The depth of field should shift realistically (bokeh in close-ups). </instruction>

A professional 3x3 cinematic storyboard grid containing 9 panels. The grid showcases the specific subjects/scene from the input image in a comprehensive range of focal lengths.

**Top Row:** Wide environmental shot, Full view, 3/4 cut.

**Middle Row:** Waist-up view, Chest-up view, Face/Front close-up.

**Bottom Row:** Macro detail, Low Angle, High Angle. All frames feature photorealistic textures, consistent cinematic color grading, and correct framing for the specific number of subjects or objects analyzed.

Analyze the input image and identify the main subject(s). Maintain perfect consistency in appearance, proportions, materials, colors, and style across all frames. Read the SCENE INPUT and generate a cinematic 9-scene sequence that progresses logically from start to finish. Each frame should represent the next meaningful moment based on the scene description.

The AI chooses all camera angles and framing automatically. Ensure cinematic lighting, consistent color grading, realistic depth of field, and coherent environmental evolution. No repeated shots. No added text.

SCENE INPUT: [describe your scene here]

Frame 1:

Frame 2:

Frame 3:

Frame 4:

Frame 5:

Frame 6:

Frame 7:

Frame 8:

Frame 9:

The Result

Why all those blank “Frame 1, Frame 2, Frame 3” lines at the end of the prompt? No idea. All I know is it works better with them than without. We like it when AI works better…

Here’s a quick sample of the test that sold me on this method. I used two reference images: an image of a waiting room (made in NanoBanana) and an image of Paul Dano (who has been through enough lately so maybe I shouldn’t put him through this as well).

In the “SCENE INPUT” line of the prompt I entered the details about the creative intention of my scene:

A man approaches an ancient old receptionist at a mental institution and asks for something. The woman is hard of hearing and does not understand. Quirky, indie film style in the vein of Wes Anderson.

And here’s what Google Flow gave me:

It’s not perfect, and that one frame has an annoying subtitle – but that’s easily handled. Notice how the character consistency holds up from the wide shot all the way to the top-down angle. And Flow understood the creative intention well enough to generate images that are pre-loaded with the emotional action I’m after.

Of course the prompt is easily edited to get whatever style, tone, and angles you want.

Consider this my gift to you. Now go make something cool before New Year’s.

Happy Holidays.

And where the future will take us.

Subscribe now

If you haven’t watched my “Welcome to Wonderland” trailer yet, give it a watch. (Don’t worry, I’ll wait).

The concept began like every project begins: with an idea. A friend proposed the concept of “mining the public domain” for an AI film. As we brainstormed, I kept thinking about the letters “PD.”

Public Domain.

Police Department.

Private Detective.

I knew there was a world worth exploring there. But an idea is nothing without execution. Below is the breakdown of how I built the world of Wonderland – from the initial character generations to the “broken tools” I used to make them speak.

The Workflow (The “How-To”)

I know many of you are here for the technical specs, so let’s get into the weeds before we talk creative theory.

When I started this little film, I made a conscious choice to use as few tools as possible. I wanted to see if I could build a cohesive cinematic universe using a streamlined tech stack. I wanted to see how low the barrier to entry could really be, while still creating a high-value narrative that didn’t feel like AI.

1. The Image Generator: Whisk

For initial image generation, I leaned almost entirely on Google’s experimental Whisk interface. Unlike Midjourney – which I love but have never once considered “fun to use” – Whisk is highly intuitive. It allows you to upload Subject (character), Scene (location), and Style (tone/lighting) references and blend them together. It’s drag-and-drop. And for the moment, it’s free. We like free.

Let’s take a look at the interface:

• The Sidebar: Once you’ve generated (or imported) your character art and style images, you populate Whisk’s sidebar with Subject, Scene, and Location reference images. Then you toggle them on and off at will to tell Whisk, “this character in this location with this color palette.” Other tools do this too, but Whisk manages to make it feel fun. Creativity is supposed to be fun.

• The “Precise” Toggle: In the text input bar, Whisk has a toggle for “Precise Reference.” When on, it locks your character perfectly. When off, the droids take the ball and run with it. I found that toggling “Precise” off often gave me the best creative accidents, which I could then refine – often by subtraction.

• The Refine Tool: Sometimes buggy, sometimes useless, but more often than not, spot on. Use this to perfect your images. It can handle text, move the camera, or give you an excellent reverse angle of your location. Occasionally it gets stumped, but that’s better than asking an image generator to change the color of a hat and getting back a completely different hat on a completely different guy.

• Volume: I generated hundreds of images to get the handful you see in the trailer. It was a process of curation as much as creation. But with the relatively quick, unlimited generations on Whisk, volume was not a problem. In fact, it often led to surprising sources of inspiration. Creativity breeds creativity. More on this below.

2. The Animation: Veo 3.1 & Runway Gen-4

We’ve got lots of way to access these tools, separately or in bundles. I used the Runway website as my gateway. Why? Well, Runway is where I currently have credits to burn. When those credits are gone I’m looking forward to going deep with Adobe Firefly – I suspect that Adobe is quietly positioning itself to be an innovative leader in this space. But I digress.

For my workflow on this project, Veo 3.1 handled the lion’s share of animations. It renders light and texture beautifully – just look at the rain on the streets.

• Troubleshooting: For no good reason that I could tell, Veo 3.1 rejected a not-insignificant number of my prompts. 95% of the time there was nothing objectionable, so I assume these were fleeting server errors. Often, running the same prompt through “last week’s model” (Veo 3.0) solved the problem—perhaps 3.0 is slightly less discerning about gritty noir content? When that failed, Runway Gen-4 was there to pick up the slack.

• AI’s Smokeshow: I’ve noticed across platforms that when AI sees smoke in a frame it thinks, “Smoke! I’m good at smoke,” and adds it… everywhere. For instance, a shot of Tinkerbell had an ashtray in the frame. With no prompting, the video generators picked up on that visual cue and gave me a forest fire. Removing the ashtray from the original image was a fix for this AI weirdness.

  

3. The Performances: Act-Two & ElevenLabs

• ElevenLabs: Most of the dialogue and VO was created via the text-to-speech tool from ElevenLabs. I find that ElevenLabs has a frustrating interface, and often requires generation after generation to get a line with something like actual emotion. But the generations are fast, so you can afford to roll the dice repeatedly until you get a take with emotional coherence. My complaints about their UI aside, ElevenLabs gets the job done.

• Runway Act-Two: For key lines, the droids lack nuance. This should surprise no one. To insert a little comic timing, I turned to Runway’s Act-Two feature, which maps a webcam performance onto a character. I acted out the lines myself (embarrassing, but effective) and Act-Two made my characters comply (after several failed generations, that is). Right now this feels like the only way to get the timing you want… though I hope someone has other ideas they’d like to share in the comments. This leaves you with a clip of your character, but they’re speaking in your voice, not theirs. Which brings us back to ElevenLabs…

• The Voice Changer: Unless you do a great Tinkerbell impression, the final step is exporting that performance audio to ElevenLabs’ Voice Changer. This swaps your voice for the AI actor’s, while keeping your cadence and emphasis. It’s a multi-step process, but an acceptable exchange to force natural human timing into an AI performance. I found that in the back and forth between apps the lip sync gets a bit loose. But – especially for such a fast-cutting piece – I was willing to live with it.

The Creative Breakdown (The “Why”)

Building the tech stack is easy. Building the world is the hard part. Here’s how I approached the casting and design of the Wonderland universe.

1. Casting Wonderland’s Anti-Heroes

My first instinct was to cast Winnie the Pooh in the Philip Marlowe role. To me, a hard-boiled teddy bear was just fundamentally funny.

But I hit a wall: AI models essentially refuse to make Pooh look tough. They always want him to smile. (Almost as though they’ve been trained on Disney’s copyrighted library. Hmmmm…)

Then, the model gave me a gift: an image of Pooh with a bottle labeled “Hunny.” It clicked. Pooh isn’t a hero; he’s a glutton. That’s greed. That’s a villain. A few prompts later, I had Mobster Pooh – a character screaming for his own spinoff.

This is an excellent demonstration of what true human-AI collaboration looks like. The AI didn’t suggest that Pooh become a villain, but it showed me an image that tapped into the character in a way that made me say, “Wait a second…”

That’s how collaboration works.

Every noir needs a Femme Fatale, and for Wonderland, Alice was the obvious choice. But AI has a bias problem here: it really, really wanted her to look like a modern Instagram influencer.

To fix this, I steered the prompts away from the Disney aesthetic and toward the original Tenniel illustrations (Public Domain), while keeping her recognizable enough for a general audience. Giving Alice a bottle labeled “Drink Me” and a white rabbit bartender sure helped.

2. Designing the “Neon-Noir”

To sell the idea that these disparate characters live in the same city, I needed a unifying visual language. I chose an “Old Hollywood” noir aesthetic – rain-slicked streets, high-contrast lighting, and surreal anachronisms. But our characters span generations, so the overall look needed to be able to contain multitudes. Throughout the piece we have black and white images, color images, and images that mix the two. Keeping them “speaking the same language” is a big ask.

The key was to really hone in on a couple of distinct vibes to guide the visuals throughout.

• Exteriors: Bathed in neon greens, oranges, and purples (The “Wonderland” palette). Wet, rainy, reflective.

• Interiors: Muted ambers, high-key lighting, and floating dust motes (The “Detective” palette). Aged, rustic, shabby.

Nailing these vibes in the development phase – and sticking to them throughout generation – was crucial.

3. The “Micro-Expression”

There is a specific shot in the trailer that I think is the key to the whole piece. It’s the final shot of Peter at his desk.

What makes this captivating? It’s not just the rendering. It’s a micro-expression that happens about a frame or two before the edit point.

A micro-expression is a look we don’t consciously see, but we feel. In that last split second, there is a hardening of his eyes – a flicker of perception. It says to us, “I am thinking.”

This is the holy grail of AI filmmaking. Humans are terrible liars; our faces give us away. As AI filmmakers, we need to become students of these micro-expressions. When the tool gives you a flicker of life – something that feels “unplanned” – you have to seize upon it. Rewrite the whole scene to make it work. That 1/12th of a second of “life” is what sells the other 89 seconds of VFX.

Final Thoughts

Six months ago, this trailer would have been impossible. The text generation on the passport alone would have broken the best models. Today, it’s a reality.

The barrier to entry is gone. The tools are broken, weird, and frustrating, but they are also magic.

I didn’t get into prompt crafting specifics in this essay, because frankly ChatGPT and Gemini can both kill that game for you. But if there’s a specific shot or moment you want to know about, drop a question in the comments – I’d love to discuss!

AI filmmaking is actually nothing new.

I’ve been generating with AI since the early days, but lately I’ve gotten serious about creating high-value content with AI tools. (By the way, so have the tools.) And I’ve got some great news for my colleagues in unscripted TV: you already possess the most important tool you need to master AI storytelling.

And it isn’t Midjourney.

The realization hit me recently. Let me describe an on-screen performance: It’s wooden. The characters don’t seem to understand the words they’re saying. They smile when they should cry. They refuse to make eye contact, and when they do, there is no life behind the eyes.

Am I talking about an AI-generated actor? Or am I talking about a typical reality show cast member?

The Answer is Both.

I have tremendous empathy for reality talent. It’s an impossible job: be yourself, but also be a heightened, emotional version of yourself – on command. You aren’t an actor, yet you must act. You aren’t a writer, yet you need a witty comeback for every occasion. It’s a recipe for awkward, disjointed footage.

But that chaos is exactly why unscripted editors and producers are uniquely qualified for the AI era.

We are used to shooting for three hours to get a two-minute scene. We know that somewhere in that mountain of raw footage is the look, the breath, or the accidental line we need to tell the story. We turn friends into enemies and awkward silences into seething tension. We sculpt performances out of nothing.

The Reality Skillset is the AI Skillset

Current AI models spit out random vocal performances and inconsistent emotions. Which is exactly the same problem we solve in the edit bay every day.

If you feel intimidated by AI, don’t be. Think about the “useless” footage you’ve saved in post-production. Think about the scenes you built from scraps. This is the exact same workflow. Your career in reality TV has made you a master of “fixing it in post.”

It’s all about forcing unsatisfactory raw material into the shape of entertainment. Call it making a silk purse out of a sow’s ear, or building a house with broken tools.

AI still needs you to make it happen.

#AI #Storytelling #Hollywood #Creativity #IndieFilm #GenerativeAI

Subscribe now

I’m getting a distinct sense of déjà vu.

When I was a kid in the early 90s, Hollywood was pumping out sequels. The radio was choked with hair metal. Everything felt polished, expensive, and completely safe. It was a monoculture.

Then, the static hit.

Nirvana released Nevermind. Robert Rodriguez made El Mariachi for $7,000. Kevin Smith maxed out his credit cards for Clerks.

It was messy. It was loud. But it was alive.

I’m starting to feel that exact same energy right now with AI.

For the last decade, entertainment has drifted back into that “hair metal” phase. Massive budgets, massive IP, massive risk-aversion. I love a good blockbuster, but I’ve spent more time napping during superhero movies than I care to admit. The polish is there, but the pulse has faded.

Enter the disruption.

Just like cheap 16mm film and distortion pedals democratized creativity in the 90s, AI is handing the most powerful storytelling tools in history to… well, everyone.

Is there going to be a lot of noise? Yes. Is there going to be slop? Absolutely.

But go back and listen to the garage bands of 1992. There was plenty of slop then, too. That didn’t matter. What mattered was that for the first time, the barrier to entry wasn’t a studio gatekeeper or a million-dollar budget. It was having something to say.

We are standing at the edge of a new Independent Era.

The tools to create world-class imagery, sound, and narrative are no longer locked behind the studio gates. They are sitting on your laptop.

To my fellow creatives: Don’t let the panic about the industry shifts blind you to the opportunity.

• Learn the tools.

• Ignore the gatekeepers.

• Make something weird.

• Make something personal.

The industry is changing, and that’s scary. But if you tune out the fear, you might hear the feedback loop of a new generation getting ready to scream.

I’m here for it.

#AI #Storytelling #Hollywood #Creativity #IndieFilm #GenerativeAI