That idea sits at the heart of my teaching philosophy: creativity grows when curiosity matters more than correctness, play comes before polish, and experimentation is allowed to lead somewhere unexpected.

Each day of SREFtember offers a short, open-ended subject. Pair it with a Midjourney SREF of your choice, interpret it however you like, and see where the combination takes you.

1. Highway in the rain

2. Solitary pool editorial

3. A funeral for a vending machine

4. A cathedral grown from salt

5. A deep-sea courtroom

6. An archaeological dig for a giant toy

7. A museum of captured weather

8. An orchestra among the stars

9. A village carried by migrating animals

10. A ritual for an extinct color

11. A zero-gravity harvest

12. A civilization on a violin string

13. A headless haberdasher

14. A portrait of the North Wind

15. Ocean animals who live in a desert

16. A mechanical forest dreaming

17. The excavation of tomorrow

18. A banquet for aliens

19. A zoo of impossible shadows

20. An idea folding itself

21. Language becoming alive

22. The moon beginning to hatch

23. Gravity leaking from a city

24. A factory manufacturing memories

25. Time discovered on a playground

26. Lightning in a bottle

27. Birth of an artificial mythology

28. Museum of lives never lived

29. Reality molting its skin

30. Robotics Fashion Show

Follow every day or jump in whenever inspiration strikes. Most importantly, have fun.

Share what you create, the prompt you used, and the images you liked. Tag your work with #SREFtember26 ( and me Brian Sykes )so together we can explore everyone’s wildly different interpretations.

As of this post - I have already completed the first 2 challenges... (but YOU interpret them the way YOU want to!!)

Subscribe now

Day 1

Day 2

Let's have a little fun - shall we?! Share a link so others can see what you post!!

When judgment is sharp, AI gets a creative team to distinct work faster. When judgment is weak, AI simply produces more mediocre work at scale.

Right now, organizations are adopting generative tools much faster than their people are developing the criteria to direct them.

Subscribe now

The symptom isn’t a lack of technical skill. Everyone knows how to type a prompt.

The symptom is creative dilution:

• Teams becoming prompt operators instead of art directors

• Briefs getting flattened into generic variations

• Tool experimentation happening without shared principles

• The loss of authorial intent in the final output

Human → AI → Human as the centerpiece

AI can generate, vary, and accelerate. It cannot discern, edit, or take responsibility.

That is why our operating model cannot be “AI replaced the process.” It has to be Human → AI → Human:

1. Human: Intention, observation, context, and creative direction.

2. AI: Exploration, variation, velocity, and execution.

3. Human: Taste, discernment, refinement, authorship, and accountability.

If you skip the first human step, you get generic starting points. If you skip the second human step, you ship generic work.

Tools change every Tuesday. Creative judgment remains the moat.

To the creative directors and marketing leaders in my network:

Where is your team feeling the friction most right now - in the direction, or in the discernment?

Leave a comment

Your Challenge: Build the Next Scene

Choose one character and one clearly defined location. Then create a short video in which the character actively interacts with that environment.

Don’t simply place them in the scene. Give them something to do that depends on the location: clear an obstacle, open a door, cross uneven terrain, react to the lighting, or interact with something already visible in the reference.

For an extra challenge, use the same two references in two different video generators. Compare the results:

• Which one preserves the character more accurately?

• Which one understands the larger environment?

• Which produces the most convincing interaction?

• What changed—even though the references and prompt stayed the same?

The goal isn’t merely to make one successful clip. It’s to discover whether your character and location can become reusable building blocks for the next shot, the next sequence, and eventually an entire story.

Subscribe now

If you saw this post earlier today - you may have seen where I created a character sheet for the redhead in the white filigree dress:

AI Lab

How to Add New Footage to an Existing Clip + Modified Lighting in Stills > With AI

Before & After…

Read more

2 days ago · 2 likes · Brian W. Sykes

But we don’t have to stop with just dropping her into the video like I did in the lesson above. We can take this process to a whole new level!

Grab/Create the Resources

If you want to follow along:

• download the assets of the female character sheet that I created (or go thru the article above and create your own).

• grab the attached reference image of the skatepark (below) that I compiled in Photoshop.

Now, you have 2 files - a model, and a location. The detail in this location shot are ideal for this because it has a clearly defined place, lighting, etc for the AI to interact with.

Get Creative

For the following, I used Adobe Firefly. Under Generate Video, I made the following selections:

The Gemini Omni Flash model let’s me select References of either frames or images. First, I provided the skatepark reference image — which as you note above is far wider than a 9:16 shape - but because it is Gemini/Nano Banana 2 anchored, it is able to remember the full visual reference for the location. Second, I provided the character sheet for the redheaded female. When ‘Frames’ is chosen, you are saying you are going to provide a first/last frame to guide the video creation.

The prompt to direct the video creation (what are you wanting the AI to do with the setting and the character provided):

The Redhead woman in a white filigree dress with fringed bottom, skateboarding thru the skatepark. Doing jumps and kickflips.

This used 300 credits and only took a couple of minutes to render.

Google Gemini

We can also do this in the Gemini platform directly… note above:

• I added the same prompt from before

• I uploaded the 2 references (redhead model and setting of the skatepark)

• I set Landscape 16:9 as the format

• I selected “Create video”

I will share my results for this as a note - but love for you to dig in and see what YOU get!

—

Before & After

I stopped by a local skatepark yesterday that is only a couple miles from where we are staying this month in Knoxville (Fountain City area). I mean to be in touch with the local governing body to see about having lights added to the park for night enjoyment by the tons of kids and young adults that frequent there.

My goal was two-fold:

1. Grab some pictures to document the scene as is, then use AI and show the park with and without lights at night.

2. Capture a little footage to try some AI addition of a skateboarder.

The image above is a photo-merged image (Photoshop ::: File > Automate > Photomerge). We will address that in a moment.

So the human side of this equation at the front side of ‘Human → AI → Human’ is the idea, and the resource collection to tackle 2 projects.

Here is the video clip panning over the skatepark (before) on the left, and on the right is the same skatepark with an AI generated single skateboarder (after):

The Still Images

For this, I went to ChatGPT. I could just as easily have done this with Gemini and its impressive Nano Banana 2, but I felt GPT-Image-2 was up to the task.

First, give the LLM a reference image and clarity verbally on what it is you are trying to accomplish. You have already seen the original photo-merged image at the top of the page. So next are the instructions:

Hold your taters there artist looking for the shortcut to copy-paste. I will give you the prompt—but first I’m going to teach you how to construct one yourself. I am all about teaching people to fish - not just catching all the big ones and handing them over. So, here is how to think through that process…

Tell the AI what you are providing. For this project, it is the skatepark. So I start the prompt by telling it clearly what it is, and what I really want it to begin paying attention to:

This is a photo of a skate park.

This is a heavily used and appreciated facility, but they do not currently have lighting.

Tell the AI what you want it to do. Keep in mind, clarity is of utmost importance. Set parameters of expectation, just as you might instruct or creative direct someone on your team. Keep in mind the notion of explaining something to someone that does not have the ability to read your mind. A great way to do that, is imagine explaining to a 5 year old - how to tie shoe laces, over the phone without video.

I would love to create some images of the skatepark.

First, is the park at night without lights (with blurred images of skateboarders of varying ages - skating)

Second, is the park at night WITH lights, with blurred images of skateboarders of varying ages.

Third is the park in the day time with blurred skateboarders.
Blur the skaters so they are almost not visible.

We want them as single images - so start with Image 1 (park at night). Once completed, I will say NEXT, and you can move on to the next image.

Before sending the AI to task, make sure it is clear. You can do that with a simple ask:

Does this make sense?

ChatGPT responded:

It added a few additional details regarding how it would approach the creation, then began to spit out the images exactly as requested:

You may wonder why I bothered creating a daytime image, since I already had one. In this case, I wanted to show the park alive and active and filled with people (like it normally is - but it was a school day, so empty).

This to me, is simply excellent for the intended purpose. I am happy.

Moving on…

Video Modification with AI

In this next bit, I am going to show you how to go from the original iPhone 12 Max Pro footage (yeah, my phone is ‘that old’, but still works) to the rendered bit with a female skateboarding in the actual footage!

Subscribe now

Read more

Sketchnotes work because they drastically reduce cognitive load. The low-fidelity, hand-drawn aesthetic feels human and approachable, turning complex technical pipelines into instant visual maps anyone can retain.

I used Gemini to act as my Art Director and Content Strategist. I fed it the inspiration images and the source content as follows:

Gemini:

To recreate or expand upon this style, a designer (or an AI model like Nano Banana 2) needs to understand both the visual “sketchnote” vocabulary and the thematic structure.

These images are “Graphic Recordings” by Sunni Brown, capturing a Robert McKee seminar. The style is deliberately “low-fidelity/high-insight,” mimicking a highly creative person’s personal journal.

Gemini went on to define the Visual Style Guide (which it called the ‘Reverse Prompt’), a Content Breakdown (Essence of the Story, Character vs Characterization) and Designer Briefing for NEW Iterations.

So, here is a brief, clear blueprint showing the repeatable modular pipeline for how to instruct Gemini or ChatGPT to grab the necessary content before creating the image. Note the image creation, is the FINAL step.

The 4-Stage Conversion Framework:

1. Source Video → Extract timestamps and core technical steps.

2. Structured Content Breakdown → Distill into bite-sized visual logic.

3. Visual Style Guide (Reverse Prompt) → Lock down linework, palette, and layout rules.

4. Image Generation → Execute the final hand-drawn visual asset.

So, it is good to go. Then I fed the following instructions to Gemini, along with the most recent video I produced:

The 6-Step Pipeline That Makes AI Characters Look Hyper Real.

Gemini:

Here is the structured written overview of the video's content and complete workflow, followed by the hand-drawn sketchnote created in the original Sunni Brown visual style.

Video Overview & Complete 6-Step Workflow Breakdown

Video: The 6-Step Pipeline That Makes AI Characters Look Hyper-Realistic and Remain Consistent!

Creator: Brian Sykes (AI LAB), adapting a commercial production pipeline by David Litwin (Pure Fusion Media / Detail AI).

Core Philosophy: Beating the “Uncanny Valley”

• Most AI generations fall into the “plastic” uncanny valley because users try to prompt everything (face, setting, pose, wardrobe) in a single step.

• The Solution: Treat AI generation like commercial art direction: separate casting, set design, wardrobe, and composition into modular stages, then blend them together.

• Prompt Framework: Powered by the 16 Elements of Generative Control (expanded from Sykes’s 2023 12 Essential Prompt Elements book) to precisely control subject, wardrobe, lighting, optics, and composition.

The 6-Step Pipeline

Step 1: The Editorial Casting Board (Face & Persona)

• Tool: Magnific AI (Image Generator).

• Engine/Model: SeeDream 4 (4K) using the “Fashion Photo” style.

• Prompt Strategy: Generate a “single horizontal 4-panel editorial casting board” on a 16:9 canvas.

• Output: 4 variations × 4 panels = 16 distinct model faces per generation run.

• Action: Iterate prompts for unique hairstyles and personality expressions, select the best face, download full-res PNG, and crop down to an isolated headshot in Photoshop.

Step 2: Creative Direction & Environment Research

• Goal: Define character narrative and visual world (e.g., modern grunge ‘90s rocker at an urban skatepark).

• Tool: Pinterest.

• Action: Search and find reference imagery with strong lighting, camera perspective, and composition (not to copy-paste, but to extract environmental essence).

Step 3: Reverse-Prompting via the 16 Elements Framework

• Tool: ChatGPT.

• Action: Upload the Pinterest reference image and prompt ChatGPT to analyze the composition, lighting, camera lens, and setting to produce a ~2,000-character prompt for Google Nano Banana Pro, organized across the 16 Elements of Generative Control.

Step 4: Scene & Background Generation

• Tool: Magnific AI (Google Nano Banana Pro, 4K).

• Action: Generate 4 environmental scene variations (e.g., skatepark bowl, bridge architecture, dramatic backlighting). Select the frame with the best depth, lighting, and camera angle.

Step 5: Component Integration (Multi-Image Blending)

• Wardrobe Source: Isolate a targeted piece of clothing (e.g., Wornstar White Skull sleeveless hoodie).

• Tool: Magnific AI using multi-image reference handles:

  • @image1: Cropped Character Headshot (Step 1)

  • @image2: Generated Scene/Environment (Step 4)

  • @image3: Wardrobe/Clothing Asset

• Prompt Architecture: “Recreate the exact scene from @image2 (lighting, angle, setting), replace the person with the man in @image1 (hair, face, build), and dress him in the garment from @image3...” with a Natural style setting.

• Result: A hyper-realistic, seamless composite blending face, wardrobe, and scene with consistent lighting.

Step 6: The Campaign Lookbook (Scaling Consistency)

• Goal: Generate a multi-pose commercial campaign around the exact character.

• Tool: Magnific AI (Google Nano Banana Pro, 16:9 widescreen).

• Prompt Strategy: Reference the master render (@image1) and prompt for a 3-panel side-by-side vertical editorial lookbook showing varied poses, quirky expressions, and camera angles.

• Finishing: Slice the resulting 5504×3072 px render in Photoshop into individual assets for social media or ad campaigns.

There you go. A quick way to take your own posted videos and turn them into sketch notes in the style of a favorite creative (including yourself!)

But I am not yet satisfied. It is too perfect for me, I want to push it a bit more. Now this actually took MULTIPLE iterations to get a result I was genuinely pleased with. Here is what it finally created, and below the paywall is how I pushed the AI to give it to me.

Subscribe now

Read more

Consider you just walked thru the sliding glass doors of your local grocery store. Right there in the doorway, you make a split-second commitment: cart, basket, or hands-free.

Not earth shattering, right? For me, going hands-free is an optimistic gesture where I tell myself I only need a few things. And if you are like me, it quickly becomes an unspoken challenge, testing the limits of how much we can balance in our arms before gravity or physics wins. But physical limits are absolute. As your hands begin to fill, your priorities narrow fast. You will pass up items that would have easily landed shopping cart space, because your current capacity simply will not allow them. The container you chose at the door didn’t just hold your items; it governed your attention, and your willingness to explore.

Another time might find you on a long road trip. You have watched the gas gauge slowly drop, passengers are restless for another bathroom break, and everyone is tossing hints of just how hungry they are. You watch exit signs roll past, taking polls on the perfect one that hits everyone’s wishes, holding out hope that just a few miles further satisfaction will be found. Eventually, the urgency for fuel and bathroom overrides the perfect meal selection, and an exit is taken. That single choice immediately limits your food options to whatever sits at the end of that specific off-ramp. We all know the mental friction. Getting back on the highway to search elsewhere after the fuel stop feels like too much friction, so you settle for what’s immediately accessible.

Choice generates more choices, but every initial choice further narrows the field. Weird paradox right?

This is Rule 1 from my [Hu]man Element framework:

Rule 1: The Container Principle

The initial boundary chosen at the threshold does not merely hold capacity; it establishes the operational constraints that dictate all downstream velocity, friction, cognitive load, and possibility.

Subscribe now

You may read up to here and think the point is obvious: of course hands hold less than a shopping cart. Of course Exit 21 only has a Subway and a Shell station. That is just logistics.

But it isn’t. It is architecture.

See, the moment you grab a cart or leave it docked and go hands-free, you did not just choose a carrying capacity. You defined the physics of your entire shopping experience.

Velocity:

Hands-free means the initial goal was out the door in under 10 minutes. The cart—even for the experienced shopper—might have you still in aisle 4 selecting which new high-fiber protein-enriched granola bars to pick over you did not know existed 10 minutes ago. The container set a speed limit on the front side, but it also ramped the speed of shopping progressively if you were hands-free. Because as your hands began to fill, selection was narrowing to a premium out of necessity of space, weight management, balance, (and need we not forget) muscle-fatigue!

Friction:

The few things you decided before entering the store (carton of eggs, gallon of milk, jar of honey, and fresh bagels) became a balancing act in your bare arms. Every step until the register became a mini-crisis. But rarely do I go to the store and come out with only what was on the list. I just came home from a “quick grocery run” of high fiber wraps, and salsa—that added 3 boxes of cereal, a gallon of tea, a half-gallon of milk, and Recees’ Peanut Butter Cups. I more than doubled the shopping results - which had me precariously perching the wares in my arms. That was only 2 extra aisles on a quick run. The awkward carrying friction, imposed a limit on getting more (or even being interested in what more might be.

Cognitive Load:

For the normal shopper, the cart pusher has to actively decide not to buy the seasonal cookies on the endcap. (Note, I said the normal shopper. But if they were chocolate-chunk, small-batch… I might be persuaded to give them a try just this time. For Trish of course.) The average person going hands-free has a different decision that bears different consequences. Can I safely add this to the assortment I am already carrying, or do I risk my load being too cumbersome to arrive at the register without catastriphe. See, the decision tree just got a few more leaves of challenge on the branches. The permiter of hands-free (no longer empty) is impacting the choices being made in real time. The choice of no container did the heavy lifting, so the brain did not have to,

Possibility:

The exit ramp with only a Subway, means no wood-fired pizza, no meaty grease-fest burger, and no grilled chicken on a buttered bun. It means the possibilities are narrowed to what is available with the decision that got you here. Obviously, getting back on the highway in hopes of something being found later is a real possibility, but often the friction of delaying the drive with another exit ramp has gotten us to our next destination late - and still no dinner. That is a real possibility also.

Containers

The shopping cart, the off-ramp, the browser tab, the calendar block, the project scope, the deadline, the hobbies list. We think that we are making a thousand choices along the way, but we are actually just living within the containers we accepted at the threshold.

• Capacity (Volume) - The scale of the initial container sets the mental default before execution starts. It defines how much we attempt to carry.

• Scope (Perimeter) - The chosen pathway defines the field or arena of immediate execution. Once we chose an exit ramp, attempting to reach outside that local perimeter drains momentum, and forces costly detours.

• Duration (Time) - The time allotted to a project. A task will expand to fill the duration you provide for it. A tight deadline creates urgency and intense focus, while unbounded time invites distraction, and diluted effort.

When you change the container at the outset, the downstream physics change automatically. Less friction, higher velocity, a clearer mind. I start with The Container Principle because every human endeavor has a starting point. But learning to see the container is only the first step in the [Hu]man Element framework. There are fifteen more RULES that define how we think, create, decide, and build once you step inside.

We will take a look at Rule 2, next.

Leave a comment

Why this - now?

THIS has been in the works since I first dove into AI. Shoot, I was working on this concept before I even explored AI and was still running my branding & marketing agency.

You see, everyone is treating AI like a tooling problem—obsessing over prompt syntax, model updates, and faster generation—while ignoring the fact that our creative output is flatlining. The failure isn’t in the technology; it’s that we are stepping up to an infinite canvas with zero boundaries, collapsing our own cognitive architecture under the weight of endless, frictionless choice. If we don’t master the human container at the threshold, AI won’t expand our creative agency—it will simply automate our mental drift.

So this… this is important.

Not because the prompt was weak. Because nobody decided what story was being told.

We’ve spent three years obsessed with the last six inches of the process — the prompt box, the syntax, the magic words. Meanwhile the decisions that actually determine whether work lands happen upstream, before anyone opens a tool. And those decisions haven’t changed in a hundred years of commercial storytelling. They’re just easier to skip now, because the machine will happily produce something regardless.

That’s the trap. AI will never tell you that you started in the wrong place. It will only make you faster at going there.

Four decisions stand between a client brief and a finished frame:

1. What kind of story is this?

2. What moments does it need, and in what order?

3. How do I frame each one?

4. How do I get the model to produce what I actually see in my head?

Miss the first and the other three won’t save you. You’ll generate twelve gorgeous images that add up to nothing, and you won’t be able to name why.

I’ve built a system for each of those four decisions — not theory, but the ones I actually run on client work when a deadline is real and a budget is attached. They’re now live as the Lab Manuals, included with every paid subscription..

Here’s what each one is, and why it exists.

1. Seven Commercial Narratives

The decision:

What kind of story is this →

Seven narrative structures — Problem–Solution, Before–After–Bridge, Hero’s Journey, Origin Story, Testimonial, Brand Manifesto, Epic Quest. Each with its skeleton, the business goal it serves, and a classic example that earned its reputation: Dollar Shave Club, Nike, Apple, Ben & Jerry’s, Patagonia.

Why it’s here: Because “something emotional” is not a brief, and neither is “make it feel premium.” When a client hands you a vague objective, the professional move is to translate it into a structure you can defend in the room. The quick-reference table does exactly that — objective in, narrative shape out. It’s the difference between having a point of view and having a mood board.

2. Eight Essential Story Beats

The decision:

What moments, in what order →

Hook, Setup, Inciting Incident, Rising Action, Midpoint, Climax, Resolution, Call-to-Action. For each: when it lands, the job it does, and lines you can steal the shape of. One continuous example runs the whole arc, so you watch the beats connect rather than meeting eight disconnected fragments.

Why it’s here: A narrative structure without beats is a diagram. This is where it becomes a sequence. It also includes the compression rule most people get wrong — a thirty-second spot only hits four or five beats, but the arc underneath never changes. Problem, struggle, breakthrough, result, action. Cut a beat, not the arc.

3. Fifteen Shots to Tell Any Story

The decision:

How do I frame each moment →

15 shots across 4 phases — Context & Setup, Character & Connection, Action & Progression, Detail & Emphasis. Each with a visual example and the specific job it does in a scene. Then 8 pairings that work, 5 that break, and 4 principles governing the rhythm underneath.

Why it’s here: This one matters more in AI work than it ever did on set. When you shoot, you get coverage — options, safety, room to fix it in the edit. When you generate, every cut has to be decided before it exists. There is no coverage. Which means the sequence has to be right in your head first. That’s why the section on pairings that break is the most useful part of the document: establishing straight to extreme close-up, reaction into establishing. Sequences that fail for structural reasons, not aesthetic ones.

4. Sixteen Elements of Generative Control

The decision:

How do I get the model to produce what I see →

16 variables organized into 4 phases — Narrative, Aesthetic, Cinematic, Technical — plus the Clarity Rule and two worked examples. Built as a desk reference. Print it, hang it, refer to it before you generate.

Why it’s here: Because piling on adjectives and hoping is not a method. When an output comes back wrong, most people rewrite the entire prompt and start over. That’s not iteration — it’s rolling the dice again. Isolate the variable, change one thing, and you’re directing instead of gambling. Same discipline you’d apply on a set where the clock is running.

Already a paid member?
All four are live. Open → the Lab Manuals

Reading free?
You’re not missing the thinking — you get all of it, permanently. The Lab Manuals are the working version: the systems, the sequences, the reference you print and hang. That’s what a paid subscription buys. Upgrade →

Subscribe now

Why these four, together

Narrative → Beats → Shots → Generation. That’s the whole pipeline, and each layer constrains the one below it.

Skip the top and you’re making beautiful frames in service of nothing. Skip the bottom and you have a great idea you can’t execute. The reason this publication keeps returning to human judgment isn’t sentiment — it’s that every one of those four decisions is a decision. The model doesn’t make them. It just carries them out, faithfully, whether they were good or not.

The rule this publication runs on

I don’t paywall the belief. I paywall the implementation.

Every argument I make about why judgment, taste, and authorship still matter stays free. Permanently. Those ideas are supposed to travel — to LinkedIn, into your team’s Slack, into a conversation you have with a client next week. Locking them would defeat the point of writing them.

The Lab Manuals are the other half. The systems, the sequences, the reference you print and hang.

If you’re a paid member: they’re live now. Go get them → [The Lab Manuals]

If you’re reading free: you’re not missing the thinking. You’re getting all of it. What’s behind the wall is the working version — and that is what your subscription buys.

Subscribe now

More coming. The next one I will be adding soon, is a worksheet rather than a reference: a one-page canvas for running a project through Human → AI → Human, with the decision rights written down before the tool opens.

— Brian Sykes

Share

For many creatives, the act of production is intoxicating. The daily ritual of writing, designing, or building creates a comforting feedback loop—a feeling of progress rooted entirely in effort, not outcomes. Because this production habit is so deeply seated, it feels like success. When the actual results lag, the instinctive response is rarely to pivot; instead, it is to double down on the same actions with greater intensity. This creates a dangerous productivity trap where movement is confused with progress.

To break this loop, creatives must transition from raw habit to structured growth using the APIE framework: Assess, Plan, Implement, and Evaluate. While most creators excel at planning and implementation, the entire system collapses at the final letter. Evaluation is the most frequently skipped step, yet it is the only one that turns raw effort into actionable insight. Without evaluation, there is no learning, no adjustment, and no genuine gain—just a hamster wheel of high-energy stagnation.

Subscribe now

The Confession: Falling Into My Own Production Trap

I preach the [Hu]man Element constantly, yet I fell into the exact same habit loop. I created these observations of RULES and FORCES that act upon and explain the very motivators and inhibitors that inflict humankind - and still find myself falling prey to them also.

Over the course of over three years, I have written and published hundreds of articles on my AI LAB Substack. I hit the "Top 100 in Design" leaderboard, celebrated the occasional revenue spike, and kept my head down, producing more.

I was stuck in Implementation.

To close the loop, I took the two highest points in my journey of new subscriptions and followers in the span of three years of archive data, grabbed dashboard screenshots, and traffic metrics and fed them into AI with a single directive:

Decipher the outliers. Tell me why my biggest spikes happened, what flopped, and what the data actually says about what creative professionals want.

Here is what that data and an objective AI audit revealed.

What the Data Revealed: The Three Outlier Lessons

1. The Acquisition Outlier: March 2025 (The “Signal” Spike)

In March 2025, one single article generated 419 new subscribers at an 8.84% conversion rate. The other 29 articles published that month combined for just 22 subscribers (0.09% conversion). That single post accounted for 95% of all subscriber growth.

Why did that 1 article in 2025 explode?

• The Moment (24–72 Hour Window): Published 48 hours after OpenAI rolled out native image generation, right when the industry was flooded with confusion.

• Orientation Over Hype: It clarified the exact confusion (“DALL-E 3? Nope, it’s native 4o...”).

• Visible Trial & Error: Complete prompt chains, failed iterations, and corrective prompts—teaching how to think, not just showing pretty pictures.

• Retained Human Identity: The recognizable presence (newsboy cap, aqua glasses, directorial judgment) grounded the fast-moving tech in human continuity. In other words - it looked like me.

2. The Authority Outlier: August 2025 (The “System” Spike)

Across 24 articles in August, the average conversion was 0.9 subscribers per 1,000 views. But one post radically outperformed:

“15 Shots to Tell Any Story: (1) Context & Setup” generated 6.25 subscribers per 1,000 views—nearly 7x the monthly average.

The Insight:

• Generic tool news failed: Articles covering “Nano Banana AI” or “Midjourney HD Video” generated 0 subscribers.

• Systems won: Readers didn’t subscribe for tool updates; they subscribed to acquire a creative operating system.

• The Series Decay Trap: Numbered sequels (“Part 3: Action & Progression”) saw steep drops because new visitors felt they were late. Every installment must stand on its own feet with an acquisition hook. » This is tough for me as a writer and educator. Most SERIES are required for the sheer volume of key information to be shared and taught. How do I continue the core ideas across multiple articles, and keep the momentum for the follow-on posts? Maybe he better way is a VIDEO, and an e-Book?

3. The Revenue Outlier: September 2025 (The “Application” Peak)

September recorded my highest grossing revenue ($1,740), driven heavily by direct and search traffic (over 70% direct).

The articles that sustained that climb weren’t philosophical musings—they were concrete, professional-grade utilities (e.g., Advanced Nano Banana Techniques and Building a Custom Photoshop AI Plugin). Readers convert when I show them how to integrate tools into an existing commercial workflow.

The AI LAB Editorial Engine: Signal ➔ System ➔ Application

From this evaluation, I extracted a repeatable three-part publishing doctrine:

1. SIGNAL (Acquisition)
   Catch consequential tool/model shifts early ➔ Resolve confusion ➔ Rapid Field Test.
   
2. SYSTEM (Authority)
   Package creative wisdom into named, proprietary frameworks (e.g., The 15 Shots System).
   
3. APPLICATION (Conversion)
   Deliver advanced workflow integrations, production templates, and decision models.

The Real Test: Putting the Findings into Public Practice

Having the insight is meaningless without Implementation and ongoing Evaluation.

So here is my public announcement:

1. I am retiring raw daily production volume in favor of compounding editorial intellectual property.

2. Every major technical breakdown will follow the Rapid Response Field Test format: What changed ➔ Where people are confused ➔ Visual stress test ➔ Where it breaks ➔ How human direction fixes it.

3. The Paywall Rule: Don’t paywall the belief—paywall the implementation.

Subscribe now

🔒 Paid Member Resource: The Personal Archive AI Audit Kit

Having the APIE theory is step one. To complete the Evaluation step without spending 40 hours reading spreadsheets, run your archive through this exact multimodal prompt chain.

Read more

If you have never experienced Mixboard, there are several AI platforms that have created versions of it. Adobe Firefly has a model called BOARDS that I talk about in my latest book - CREATIVE CONTROL. BOARDS is a pretty robus…

Read more

Sam explained:

Read more

From yesterday…

On the 7th of January, 1839, François Arago stood up in front of the French Académie des Sciences and announced that Louis Daguerre had solved photography.

He didn’t elaborate or lend explanation. The process was still a state asset under negotiation. He just said it was done.

Nearly 300 miles away in Wiltshire, William Henry Fox Talbot pan…

Read more

Look closely at the image above - View from the Window at Le Gras and you’ll notice something impossible. The sunlight falls on the buildings from the left. It also falls on them from the right.

Both, at once, in the same frame.

That isn’t damage or a flaw in the plate. Joseph Nicéphore Niépce’s exposure ran so long that the sun had crossed the entire sky while the picture was being made. What you’re looking at isn’t just a moment frozen in time. It’s a summer day, averaged.

It explains why there are no wildflowers in the first fifteen years of photography. A breeze would be enough to erase a subject completely. The medium didn’t begin with what photographers wanted to shoot — it began with the short list of things that would hold still long enough to be shot at all. Rooftops. Plaster casts. Table settings. Dead game birds. Basically the still life studies that were being captured as paintings.

Every technique that follows in this series is, in one way or another, an argument with that limitation.

Why do people keep inventing new ways to do a thing that already works?

I might be the world’s worst for this very thing. My wife jokingly asks (as I set out to make homemade pasta) - “Why not just buy some pasta from the store?” Genuinely a reasonable question - and totally how she would solve the problem that technically isn’t a problem. A solution has been found that works, is affordable, and is readily available in many varieties.

So why do I (being one of those sorts of people) decide to keep inventing new ways to address this non-issue? The best way to answer that, might be this… it is not for lack of a solution I do it. For me, it is the discovery and learning, and the finished result that I accomplished.

Let’s examine this in the world of image making.

The ‘picture’ was never the hard part

Here’s the thing most people get backwards about the invention of photography.

Nobody needed to invent the image. The camera obscura had been described by Mozi in China in the fifth century BCE, worked out optically by Ibn al-Haytham around 1021, and was in common use by Renaissance painters as a drawing aid. For roughly two thousand years, you could walk into a darkened room, watch the world outside appear upside down on the wall, and trace it.

The image they could ‘create’. Keeping it was the problem.

Johann Heinrich Schulze established the raw material in 1727, when he found that silver salts darken on exposure to light. Thomas Wedgwood and Humphry Davy took the obvious next step around 1802 and made silhouettes on silver-nitrate-treated leather. They worked beautifully... for just a little bit, but the image continued to darken in ordinary daylight until the whole surface went black and the picture disappeared into itself.

Wedgwood and Davy could capture. They could not stop. For a generation, that was the wall.

Niépce’s answer was permenance (via asphalt)

Niépce solved it by abandoning silver entirely.

He coated a polished pewter plate with bitumen of Judea — a naturally occurring asphalt used at the time as an etching resist. Bitumen has a useful property: light hardens it. Where the projected image was bright, the coating cured; where it was dark, it stayed soluble. A wash of lavender oil and white petroleum rinsed away everything that hadn’t hardened, leaving the image standing in cured asphalt on bare metal.

Nothing in it was light-sensitive anymore. The picture was chemically finished, and therefore permanent.

He called it heliography (sun-writing). The first successful one, in 1822, was a contact copy of an engraving of Pope Pius VII. It survived until 1909, when an attempt to photograph it for reproduction destroyed it.

View from the Window at Le Gras dates to 1826 or 1827 and still exists, at the Harry Ransom Center in Austin. Its exposure is traditionally given as eight hours, though researchers have argued the double-sided lighting implies several days. Either figure lands in the same place: the first photograph is not an instant. It’s a duration.

Worth noting for later in this series — Niépce also used bitumen to etch printing plates, producing the first photomechanical reproductions. The photogravure and photo intaglio processes we’ll get to later, trace directly back to this same asphalt.

The ‘photo’ type, you probably never heard of

In the summer of 1832, Niépce and Daguerre — by then partners — made something interesting.

They took the residue left from distilling lavender oil, dissolved it in alcohol, coated it onto a silvered plate, and exposed it. Development happened in the fumes of white petroleum. The result was a soft whitish image sitting on mirror-bright silver, appearing positive or negative depending entirely on how you tilted it.

They named it the physautotype. Exposures still ran about eight hours.

It’s the forgotten branch of the family tree, and it deserves a place in this series precisely because it went nowhere. Not every technique becomes the future. Some are simply someone finding out what happens (ie - making homemade noodles).

Niépce died in July 1833, still working in hours.

Daguerre’s real breakthrough

Daguerre kept going alone, and in 1835 arrived at the idea that actually made photography possible.

Everything before him required the exposure to complete the change. The light had to do all the work, start to finish. That’s why it took a day. Think in relation to something being sun-bleached, it is the length of exposure that changes the visual properties.

Daguerre found that a brief exposure — far too short to produce anything visible — left an invisible latent image on an iodized silver plate. Mercury vapor then amplified it into a picture. The light only had to start the process. Chemistry finished it.

Exposure times collapsed from hours to minutes.

That single conceptual move is the most consequential idea in this entire series, and it isn’t about materials at all. It’s about noticing that you don’t have to finish the job in one step. Nearly every process that followed — every developed-out paper, every wet plate, every roll of film — is a descendant of that observation.

Two years later he had a fixing method. In 1839 the process was announced publicly, and John Herschel supplied hyposulphite of soda as a superior fixer, drawing on a chemical property he’d identified back in 1819.

The wall came down that year. Part 2 picks up there, with two rival processes announced within weeks of each other.

Making these style images with AI

With any of the image generators, you can write a prompt that asks very simply for it to produce “a daguerreotype image” and get a monochromatic (maybe b/w) image with texture and age marks. You can describe the process, and end up with quite possibly a confused AI as to what the subject is and where you are going with things.

Let me show you.

The less than ideal options first.

A daguerreotype-style image of a 19th-century photographer in 1830s France, operating an early wooden box camera on a tripod in an outdoor courtyard, wearing period-accurate formal attire with a vest and cravat, surrounded by vintage photography tools and early equipment, soft natural lighting, shallow depth of field, historic and scientific atmosphere, documentary realism --no modern outfits, color, smiling, digital elements --raw --ar 7:5 --stylize 250

This is a Midjourney prompt, and here is the results:

It has a period vibe in the clothing, focus and equipment… but color was not available for quite a while yet for this photo style.

Subscribe now

Read more

Earlier in this series I mentioned that the current approach builds on ’s original workflow. I took what he provided and added a bit more human-in-the-loop control, expanding his structure into what you now learning to my “16 Elements of Generative Control.”

This piece is that expansion in full: every element, broken out by phase, with the vocabulary to make each one land. This one took me all day to put together, but I really want you to have the Desktop Resource and how to use this in your prompting.

Subscribe now

How These 16 Actually Work in a Prompt

There’s no interface here — no dial, no toggle, nothing to click. Every one of the 16 Elements lives inside the same plain block of text you type into Midjourney, Nano Banana 2, or GPT-Image-2. What actually varies from element to element is three things: word count, word choice, and word position.

Word count and choice determine influence. A one-word Setting — “rooftop” — barely registers in the output. A loaded, sensory Setting — “cracked rooftop tar, humming AC units, string lights sagging between poles” — visibly shapes the result. The element itself didn’t change; how many words you spent on it, and how specific those words were, did.

Position matters. These models tend to weight words closer to the front of the prompt slightly more heavily than words near the end. So the order you write the 16 Elements in isn’t cosmetic — whatever you list first is quietly getting more attention than whatever you tack on at the end.

You’re working inside a token budget. Every model has a practical limit on how much prompt text it processes effectively. Tokens are the currency, and that currency is finite. Try to write all 16 elements out in exhaustive detail and you’ll blow past that budget — diluting everything, and pushing your lowest-priority details past the point the model reliably reads them at all. This is why earlier, I requested ChatGPT to write my prompt to fit 2000 characters.

That’s the real mechanic behind “priority of focus.” It’s not a metaphor — it’s a budget. Decide which three or four elements are doing the actual work in this generation, put them first, spend most of your words on them, and either compress the rest to a phrase or cut them entirely.

The Clarity Rule

Before you touch a single one of the 16 Elements, start here.

Priority of focus beats volume of detail. Write one core-intent sentence — plain, specific, and about the one result that matters most for this generation — before you add anything else. That sentence is the Clarity Rule check: if you can’t state your intent in one sentence, you don’t have a prompt yet, you have a wish list.

Casting a face? Strip the background out of that sentence entirely. Building a world? Make the subject generic and let the sentence describe the space instead. The core-intent sentence tells you, in advance, which of the 16 Elements below are even relevant to this generation — everything else stays quiet, or gets left out completely.

This is also where you decide your passes. Don’t force one prompt to carry the casting, the world, the lighting, and the color grade all at once — composite them later. The Clarity Rule is the filter that keeps a sixteen-item vocabulary from turning into a sixteen-item demand list.

Every element that follows is optional. The one core-intent sentence is not.

Phase 01 — The Narrative Layer

Elements 1–4 · Who, What & Where

Before any style or lighting is applied, the model has to understand the physical reality of the scene. This is the layer most people rush past — but it’s the one that determines whether everything you build on top of it feels inhabited or just decorated.

Read more

I can play trumpet, saxophone, flute, guitar, and piano — in steadily diminishing order of proficiency. I haven’t touched any of them in years.

That wasn’t a loss. That was a decision.

Every YES in your life arrives attached to a…

Read more

There, I said it.

Read more

To play catchup - P 1-5 Overview can be found at • Mastering the Core Pipeline

P6 • Motion & Temporal Space - Lookbook asset creation

P7 • Spatial Relocation - character placed sporadically in a scene

P8 • Model Sheets & Character Systems - multi-outfit/setting model sheets

So far you have seen me create 2 surfer dudes (one with dreads and one with wavy hair), a skateboarding rocker, and a superhero redhead.

All of this is truly great… but there is more here that you might not have put together - yet. What if each collection of characters is now interchangeable. Let me show you what I mean.

Replace the man in @img2 with the woman in @img1 . Make sure the Photo reveals a 25-year-old white woman with long, wavy red hair and a muscular, athletic physique, wearing a spandex suit of blue and orange, walks out of the ocean carrying a surfboard. The background is the sea and the beach

See that? The picture I created before of our man coming out of the water carrying his surfboard - is now the woman with her super hero costume, and red hair.

Let’s try something a bit more complex… Here is our rocker dude with a guitar, thrashing at the skatepark. I really like this collection of 3 images:

Let’s swap out our dude - for the woman. But we can let the collection keep the same outfit the dude is wearing…

Sweet! Now, let’s really push the envelope a bit. What if we liked this fun, modern wedding dress as an idea:

I took the dress image to ChatGPT and gave it the following prompt:

I want you to write a 2000 character prompt that produces a wedding dress with this essence. The person is non-consequential. The importance is the dress and its styling for a Nano Banana Pro prompt to render.

Running this is Magnific, using Seedance 5 Pro 2K, which gave me the following dress:

Now watch this… Let’s put this dress on our redhead playing the guitar in the skatepark.

Subscribe now

Read more

To play catchup - Parts 1-5 Overview can be found at ‘Mastering the Core Pipeline’.

P6 • Motion & Temporal Space - Lookbook asset creation

P7 • Spatial Relocation - character placed sporadically in a scene

Let’s jump in to today’s article - Model Sheets & Character Systems.

If you have been playing with AI for a while, you recall the challenge only a couple years back gave us, in even getting a consistent character. Whether that was wrangling with a LoRA, or using a Character Reference in Midjourney - our earliest solutions in the space still gave us a challenge to effectively define a model that realistically existed in multiple views. Here is one of a puppet character I used to tell a segment of the [Hu]man Element story just last year… yet note - while it somewhat works (puppets are more visibly forgiving than photo-real humans), I had to choose between the best of multiple renders to get even this level of consistency… and that was under a year ago.

Now, we are able to produce a character with great consistency in a variety of platforms. Let me reiterate that consistent that is not real, is no longer an acceptable standard expected from AI. You may recall ‘Brad’ from my series earlier this year - where we built a Character Reference Sheet. These are fantastic, if you have one character, but the moment you introduce multiple characters into the shot, you then begin producing some conflicting information for the AI to distinguish between. Having multiple faces visible on a single reference sheet can confuse the model, leading to identity flip-flopping or "model bleed" when multiple characters are in a scene.

My friend Tim Simmons has a couple solutions he presently utilizes with great success. One he calls The 360-Video Extraction Method, and the other he calls The Face-Blocking Technique. Let me walk thru both of these.

The 360-Video Extraction Method

Instead of struggling to prompt consistent 3/4 and rear angles in an image generator (though Nano Banana Pro and 2 do it exceptionally well), generate a head-on view first, then run it through a video model with a simple prompt like "Turn the character around." You can screenshot the resulting video, and get perfect 360-degree angles because the video model handles the spatial logic and structure.

Let me show you in steps…

First, I want a simple character to work from. Following the same 4-up method of character creation as I shared in P2, here is a red head female in a black t-shirt.

Give me 3-up images of the model in <<<img1>>> as a super-hero character, separated by a thin black line, and each costume totally different of the 3 images. Full body view. Keep the curly red hair visible. Make the costume form-fitting performance material made of muted primary colors (different in each of the 3 images) that best match her red hair. Keep the background neutral white. Pose the woman from the front, straight on view, hands akimbo (fists on her hips, elbows out). Make her look confident and strong. Do not include any type of emblem. All white background.

I chose this model version… and used Photoshop to isolate her.

I used the Google Veo model to create this video:

All you need to do, is screenshot the various angles that you need, or you can do what I found to be easier (if you have Photoshop). Open the short clip in Adobe Photoshop, and pan the scroll head, then Export > Quick Export PNG.

Subscribe now

Read more

Today, I am going to expand on that with 2 unique directions.

First, the reformatting of the existing prompt request (how you ask), to totally reimagine the possible. Second… well, that’s just too cool for words right here at the starting line, so we will look at that one in just a bit. But a hint is your character + an environment = placed anywhere!

Before we dive into this article, if you haven’t seen the Part 5 article - click it and read it. There are some excellent pointers in there. And I will add, if you haven’t taken the almost-30 minutes to watch the YouTube video I dropped on Tuesday (July 28), be sure and watch to get up to speed with where this lesson kicks off.

In the Video, I produced this great character with some wild hair, and a neat sleeveless hoodie.

Technique 1: Action Alteration (Modifying the prompt to change the activity while preserving the setting/outfit).

Using the following prompt (ran a few times), we got a lot of neat images:

I need to create an entire photo campaign around the man and garment in img1 so use is scene as a reference and create three new identical vertical photos separated by a thin white line that show this same man in different positions and camera angles in some shots. some close ups but make the garment the hero. shallow depth of field. high fashion magazine cover photo but don’t add any words. Add personality and a little quirkiness to the poses. Shot on a DSLR camera, with Kodachrome film.

These are awesome! So much personality expressed, while retaining the exact same character throughout. While I love the vibe - what if we wanted the man to be doing something else - but in the same outfit, same scene… just otherwise occupied?

I need to create an entire photo campaign around the man and garment in @img1 so use is scene as a reference and create three new identical vertical photos separated by a thin white line that show this same man in different positions removing the zip hoodie, and giving it and his skateboard to a young teen boy (African American with great afro). Utilize varied camera angles in the shots. Some close ups but make the garment the hero. Shallow depth of field. High fashion magazine cover photo but don’t add any words. Add personality and a little quirkiness to the poses. Shot on a DSLR camera, with Kodachrome film.

Well, you just need to modify the action you want seen in the scene!

But that is not all… what if strumming a guitar was more the vibe?

I need to create an entire photo campaign around the man and garment in @img1 so use is scene as a reference and create three new identical vertical photos separated by a thin white line that show this same man in different positions sitting on his skateboard, strumming a vintage Les Paul guitar in black. Utilize varied camera angles in the shots. Some close ups but make the garment the hero. Shallow depth of field. High fashion magazine cover photo but don’t add any words. Add personality and a little quirkiness to the poses. Shot on a DSLR camera, with Kodachrome film.

See the vibe? You can nail your character, and vary the activity easily with a slight prompt modification. This is where I find Magnific with the varied partner tools (like Nano Banana Pro/2) do an exceptional job!

But I called this Part 7 • Spatial Relocation - and sure our rocker skate-border is showing up in random ways in each scene generation, that is not technically “Spatial Relocation”.

Technique 2: Spatial Relocation (Injecting the character into a completely foreign, pre-generated environment).

So, what IS Spatial Relocation? Glad you asked. I am going to show you how we can supply our character (decked out like we want them to be) and a scene - and let AI drop our character “in scene” randomly interacting with the space… and here is how we do that!

Subscribe now

Let’s start with a place to put our character. I have genuinely enjoyed the underpass skatepark - but let me take the skate-boarding rocker to a whole new place… How about a modern mall food court? I found this picture of just the place (to serve as AI inspiration.) Here goes the magic. First step - ChatGPT (you know the drill if you read the prior posts).

Read more