As you read this we’ve just had a public holiday here so this arrives a day later than usual. But as I write this it’s Friday evening and I’m rushing to get it finished so that I can enjoy the holiday weekend ahead, therefore I’m going to get right on with it.

Hope you’re well!

Synthetic Content

Google released Gemini Omni 1.1 Flash, an update to its video generation and editing model that adds scene extensions, video references, first-and-last frame controls, and optional 360p output for faster results (that can be upscaled to 720p). blog.google

It’s available now in Flow, AI Studio, and an API. I made a very quick test ⬇️ using an image reference and a video reference; I’d also planned to extend it but realised that feature isn’t available in Flow yet, which is disappointing. But you can take a look at my references and prompt (if you use Flow).

A couple of weeks ago [TL 442] I talked about open weights models but I didn’t emphasise why they’re so important. As well as being able to host and run them on your own hardware, which is good for privacy and cost-saving, open weights models can be custom-trained to give them new capabilities — something which would be impossible with closed and proprietary models.

This week fal launched H3 Max, a variant of MiniMax H3 [TL 440] that’s been tweaked with improved visual quality and prompt adherence and accelerated so that it can generate five-second videos in under three seconds. blog.fal.ai

It’s available now on fal and starting to appear in some other creative tools. I ran a couple of quick tests ⬇️; it struggles with scale in the first one, and loses the painted look in the second (some of that could be down to me not knowing its prompting best practices yet). But it’s decent, cheap, and very fast; I generated each 15-second clip in the same time that it takes to watch it.

• Adobe stealth-released Firefly Video 2 (Preview), a video model that can edit clips up to five seconds long.
  There’s been no announcement of it that I can find, I just stumbled across it in the ‘Video > Prompt to edit clip’ interface. I put some tests up on my Insta account. It seems decent, but five seconds is quite short.

Runway introduced Solaris, a real-time Interface World Model that generates dynamic app and web interfaces frame-by-frame without underlying code. runway.com/news

This is really intriguing; I’d love to look into it more deeply but it was announced too late for me to do anything about it in this edition (and I’m running out of space). I may cover it in more detail next week.

• Tavus launched Sparrow-2, a real-time conversational model that analyses whole acoustic scenes to improve voice chats by better handling turn-taking, pauses, interruptions, and background noise. tavus.io

Midjourney started a public test of its v8.2 editing model, with support for natural language edits, multiple image references, inpainting and outpainting. midjourney.com

Midjourney has long been surpassed by other models in photorealism, text rendering, and more; but it’s still peerless for its range of artistic styles and how easy it is to make something lovely. In this test ⬇️ I created a starting image (left) using two character references, and one style, then edited it to show the scene at night (middle), and then the characters turned around (right). It’s pretty impressive, although does seem to sharpen the image each time.

• Bria released Fibo Generate 1.5, an update to its image model that improves visual quality and speed. x.com/bria_ai
  Fibo isn’t the best model by any measure except for commercial safety, but it’s good enough for many use cases. Here ⬇️ is a quick test.

Creative Tools

Adobe Photoshop added new AI capabilities including the AI-Assisted Editor which brings a full prompt-based interface into the app, and Instruct Edit with Masks for natural language editing. blog.adobe.com

Generative AI is getting more deeply integrated into the most popular apps, tools, and operating systems. Some people still think it’s a fad.

• Bria launched Create, a production service that turns assets into editable layers ready for transcreation. bria.ai, instagram.com/bria.ai

• Recraft launched V4 Styles, which strongly applies a visual style from a preset or reference for consistency across generated images. recraft.ai, instagram.com/recraftai

  I’ve found that it too strongly applies a visual style, sometimes overwhelming the subject you prompted for. But I confess that my testing amounted to about five minutes.

• Figma Weave added 3D ‘set direction’ to pose scenes from any angle for consistent spatial environments in generated content. threads.com/@figmaweave
  It’s powered by World Labs’ Marble model [TL 425].

• Magnific released a desktop app for MacOS and Windows, with the full range of tools plus syncing and storage management. magnific.com, instagram.com/magnific_ai

• Veed redesigned its avatar studio, simplifying the interface for faster creation. instagram.com/veedstudio

• ElevenLabs’ ElevenMusic added Composer, a section-by-section song editing tool that lets users rewrite lyrics, audition multiple takes, and regenerate individual parts of a track without altering the rest. elevenlabs.io

Assistants

Google updated Gemini Live with new voice-driven productivity features, including integration with Spark [previously] for multi-step tasks across Google Workspace apps, Gmail management, and a Daily Brief. blog.google

Some of these features (notably Spark) aren’t available here in the UK yet so I can’t properly test them. Voice interaction in Google apps has become really very, very good; some of it is due to the new Gemini 3.5 Transcribe model, which is now available in AI Studio, Antigravity, and an API.

• Gemini in Chrome can use sections of a page for specific context, and remix images with Personal Intelligence. x.com/googlechrome

• Chrome’s AI Mode in Search came to the UK, letting users select recent tabs and file uploads for context, with a choice of Gemini models. instagram.com/googleuk
  Google Search is now, to all intents and purposes, a more limited version of the Gemini app.

• Google introduced Expert Intelligence, letting users query, analyse, and generate content from trusted publications, starting with ebooks purchased from Google Play. blog.google
  It’s available in Gemini Notebook today, coming to the Gemini app and AI Mode in Search in the future.

• ChatGPT Ads is expanding globally, with self-service access. The company says Ads are predicted to raise $1 billion in revenue this year. openai.com

Work

• Anthropic expanded Claude’s memory across Chat and Cowork, so personal and professional context is available to whichever tool you’re using. claude.com

• Claude Cowork added a built-in browser, letting the desktop app navigate websites, read pages, and fill out forms without requiring a browser extension. claude.com

• ChatGPT Work can use its computer and browser to sign in to websites using a password manager so ChatGPT never sees your username or password. x.com/ChatGPT

• Perplexity launched Portable Computer, AI agent software (built with NVIDIA) that enables users to run multi-step workflows on their own hardware for improved data privacy. perplexity.ai

Model Dumping

It’s clear that Chinese companies are flooding the market with inexpensive and/or open weights models to drive down prices and weaken the domination of US companies — this is a playbook they’ve used before, notably with steel and EVs. Just this week saw the release of three powerful foundation models:

• Alibaba’s Qwen3.8-Flash-Next is a multimodal open-weight AI model designed for agentic workflows and coding, and serves as an architectural preview for its upcoming Qwen4 model. qwen.ai

• Z.ai’s GLM-5.3-Flash is its first natively multimodal AI model, designed for high-efficiency coding and visual reasoning at one-tenth the cost of previous versions. z.ai

• Tencent’s Hy4 preview is a flagship LLM offering a huge (one-million-token) context window and improved capabilities across coding, office work, and scientific research. hy.tencent.ai

I don’t know how good these are and don’t have the time or a useful framework to evaluate them properly, but they all currently sit in the top 10 on the arena.ai leaderboard for agentic coding.

Social & Creators

Instagram added First Draft, a tool to automatically edit and trim a selection of clips for quickly getting to a (fully editable) rough cut. instagram.com

Something to note: this is 100% an AI tool, and most other platforms that offer similar tools tend to promote them as AI. But not this one. I wonder if this is a conscious effort to avoid the noisy online backlash? FWIW I think other platforms and services should focus more on the functionality, and less on the AI.

• Instagram is tightening its rules on AI characters, requiring accounts to label themselves as AI-generated or face restricted reach. creators.instagram.com/blog

• YouTube added Amazon to its Shopping affiliate program, allowing eligible US creators to tag Amazon products in Shorts, long-form videos, and livestreams. blog.youtube

Twitter is back. But not. A new app called twitter.now wants to ‘reclaim’ the social network, including its name which seems to be classed as an abandoned trademark since Musk pushed X. The platform wants to differentiate itself with a trust ranking at its heart, and has opened an early access programme for paying founders — a place can be yours from $20.

IDK if this is going to work; I think there’s more or less a market size limit for text-based microblogging platforms, and between X and Threads (and to a lesser degree, Bluesky, LinkedIn, and others) there are already plenty of options available. Social networks rely on a critical mass of users, and it’s quite the ask to get everyone to move to yet another new one — even if it’s an old one.

XR

• Meta rounded up a batch of recent updates to Meta Quest, including full availability for Meta AI, a redesigned navigation interface, improved hand and voice controls, and enhanced privacy settings. meta.com

This was a very quiet week and I don’t have much to say so I’m just going to get straight on with it.

Hope you’re well!

Synthetic Content

You wait ages for an SDR to HDR conversion model, then two turn up at once… Runway’s Ruby and Topaz Labs’ Hyperion 2.5 both let creators convert standard 8-bit video into high dynamic range for post-production workflows.

Also this week came Black Forest Labs’ FLUX Upscale, which upscales video up to native 4K resolution while fixing common visual artefacts, available as a standalone tool and through an API. And Higgsfield’s Relight, which lets users add professional lighting to shots after capture by adjusting light angles, colour, and brightness.

AI video is creeping up to professional (post-)production quality, at which point it will become more or less common and invisible. Another indicator: ByteDance and the Motion Picture Association reached an agreement to “maintain strong guardrails” around the use of intellectual property in generated content — a move that makes Seedance more palatable to the industry.

• Adobe Firefly launched a new unified image and video creation interface in public Beta. threads.com/@adobefirefly
  It’s a welcome step; having separate interfaces for images and videos was becoming increasingly unwieldy. Only seems to be available for personal accounts right now.

• Higgsfield launched Faceless Studio for rapid generation of videos with voiceovers for explainers and stories. threads.com/@higgsfield.ai
  This is going to be incredibly useful for creating YouTube slop.

• FLORA released a Shopify plugin, enabling the production of ads and product images/videos at scale. x.com/floraai
  A very useful addition to FLORA’s Fashion Studio announced last week [TL 442].

• VEED launched Multi-Shot Characters, for avatar videos with multiple camera angles. x.com/veedstudio

• ElevenLabs’ announced the general availability of v3 Conversational, a speech model optimised for real-time use cases. x.com/ElevenLabs

Music

Alibaba launched Happy Shrimp music generator, which can generate songs (with vocals) up to five minutes long. happyshrimp.ai

It’s a very good model; it absolutely nailed the British New Wave Synth-pop style I requested in my test ⬇️. Will Legal teams approve its use in commercial work? Probably not.

Adobe Firefly’s audio generation tools are generally available, letting users create commercially safe music, speech, and sound effects. blog.adobe.com

Lke all of Adobe’s own models, these are… fine. They’re not up there with the best, but they’ll be OK for some uses and they’re commercially safe, which is important. Here ⬇️ is a test I ran for “a brooding, mysterious song, with synth-pop style, for a movie soundtrack”.

• Stability AI updated Stable Audio 3.0’s capabilities with new tools in the Web app to “give iterative direction, try variations, mix and edit, work with multiple tracks, and extend the length” and a DAW plugin for musicians to generate, edit, and mix AI audio directly in their preferred digital audio workstations. stability.ai

Bonus Link

5 Things to Know About Suno’s BMG Licensing Deal is a useful summary of the changes coming to Suno in the near future. The main point to note is that all of Suno’s existing music models are going to be scrapped, replaced by new ones trained on licensed music. Will they be as good? Who knows.

Assistants & Search

• Google rolled out Gemini in Chrome for Android users in the US, adding webpage summarisation, Google app integrations, and an agentic ‘auto browse’ task-completion feature (for paid subscribers). blog.google

• Google introduced new AI-powered learning tools across Search, Lens, and AI Mode, including interactive visual simulations, practice quizzes, Gemini Notebook integration, and custom study file generation. blog.google

• OpenAI introduced ChatGPT for Teens, a dedicated experience for users aged 13 to 17 that features built-in safety protections, parental controls, and guided study tools designed to support learning. openai.com

• ChatGPT Ads expanded to 31 European markets, showing ads to Free and Go tier users. openai.com

• Mozilla updated Smart Window, its (opt-in) AI-powered browsing feature, adding search results powered by Exa, automated tab grouping, and visual history previews. blog.mozilla.org

• Amazon’s Alexa+ is now free on all compatible Fire TV devices (in the US), bringing conversational voice search, contextual recommendations, and smart home controls. aboutamazon.com

Work

• ChatGPT launched an Apple Messages plugin, enabling Work and Codex users (on Mac) to search messages, catch up on conversations, and draft replies. instagram.com/chatgpt

• Google Antigravity added Remote Control, a service that lets (paying) users run agentic sessions across different machines (laptops, desktops, and servers). antigravity.google

Everything Else

• Meta AI introduced new features for small businesses, integrating with Facebook and Instagram analytics, Meta Ads, and Google Workspace to analyse organic and paid performance, benchmark competitors, and generate reports. facebook.com/business

• Apple updated its EU app developer terms, replacing the per-install Core Technology Fee with a 5% transaction commission in apps distributed outside the App Store, and letting developers offer alternative payment options. apple.com

Bonus Link

This presentation detailing the recent hack of Hugging Face by an unreleased OpenAI model is kind of jaw-dropping. Yes, it gets technical in a few places, but I urge you to stick with it.

This week saw a notable rise in the release of open-weights AI models, and as this has become a bit of a political talking-point recently I thought it might be useful to define what “open weights” actually means, for anyone who’s wondering.

There’s a common misconception that models contain copies of all of their training data. In reality, when a model is trained very complex algorithms are used to transform the data into a set of numbers that represent it, in order to calculate answers; these are the weights. When a model has open weights these numbers are given to the public. That means anyone can download the model to their computer (or host it on a server), giving privacy and reduced costs, and enabling custom fine-tuning for dedicated purposes.

Open weights is not the same as open source, where the full source code of a software program is made available for free; the creator of the open-weights model can keep the training data and the source code private. And they can limit commercial use of the model through a licensing agreement, letting you use it for your own purposes, but not resell it or any software that’s based on it (this applies to open source too).

Right, definition over, so let’s get on with it. Hope you’re well!

Made By Google 2026

Google’s annual hardware event saw the launch of four new Pixel phones (11, 11 Pro, 11 Pro XL, and 11 Pro Fold). The phones will feature more Gemini Intelligence, some nice camera software upgrades including Magic Capture which chooses and upgrades stills from a video so you don’t need to switch modes, and new Creator Studio with teleprompter and Storyboard editor.

Also announced: Pixel Watch 5, with on-device Gemini and new health features, Pixel Tag tracker, upgraded Pixel Buds capabilities including dynamic noise cancelling and Gemini audio control, and more connected apps for Gemini (which just passed the 1 billion user milestone).

It wasn’t the most exciting set of announcements, but there were some decent quality-of-life improvements. I wasn’t planning to buy a new phone but with the incentives of a Pixel Superfans discount, a trade-in boost, previous Google Store credit, and six months of Gemini Pro, it turned out that I wouldn’t need to spend a lot to get an upgrade; so I did.

Synthetic Content

LTX-2.5 is an updated video model with native audio that brings improved multishot generation, prompt understanding, and visual quality, including HDR workflow support for filmmaking. ltx.io

It’s not a top-tier model but it’s decent, very fast, and cheap. Here ⬇️ are a few quick tests:

Sand AI’s Magi-2 Preview is a huge (114 billion parameter, for the nerds) unified audio and video generation model with an emphasis on cinematography and character performance. sand.ai

This only appears to be available through a (paid) API so I haven’t had the time to set it up and test it yet. I remember not being hugely impressed by the previous version [TL 377] but it’s been over a year since that was released so hopefully this will be better.

Perhaps the most notable feature of both of LTX-2.5 and Magi-2 Preview is that they’re open-weights releases; with these and the recent releases of MiniMax H3 [TL 440] and FLUX 3 Video [TL 441] this is a golden age for open weights video models.

• Seedance 2.5 can now generate native 1080p videos with native 10-bit color, richer textures, and more natural lighting and skin tones. x.com/BytePlusGlobal
  It’s rolling out to all the usual creative tools now.

Microsoft released MAI-Image-2.6, with improvements to text rendering, portraits, and commercial design. microsoft.ai

I made this ⬇️ comparison of 2.6 (left) with 2.5 (right); the newer model is more detailed and photorealistic, but perhaps has less… character?

Music

MiniMax Music 3.0 is a music generation model that can produce tracks up to five minutes long, with improved audio detail and song structure. minimax.io

Like the video models above, this has been released with open weights. In my very limited testing (for example this ⬇️) it seems it’s exceptionally good at structure and vocals, but doesn’t really follow my instructions for instrumentation and genre; that might be because I haven’t learned its prompting syntax.

• Mureka’s V9.5 music generation model promises more natural and expressive tracks across a range of styles. mureka.ai
  There’s no free tier available so I’ve never tried Mureka.

• Suno introduced Studio 2.0, a major update to its browser-based generative music studio, adding MIDI editing, custom audio plugins, an AI assistant, and more. suno.com

Spotify introduced AI Persona badges, a label which shows that “an artist’s identity may be AI-generated and does not represent a real person“. newsroom.spotify.com

Profile-owners can self-declare, and Spotify will also use automated and human judgement on suspected undeclared profiles. Any profile with a label will be excluded from editorial and algorithmic recommendations by default. I found this interesting:

Listeners have been clear in telling us that they don’t like seeing an artist profile that seems human, only to find out that the persona is AI-generated.

So AI output isn’t judged by whether people like it, but whether they feel ‘cheated’ by it. I wonder how soon, if ever, that will change.

Creative Tools

Multiple updates from Higgsfield: Cinema Studio 4.0, brings video generations up to 30 seconds (powered by Seedance 2.5) plus granular controls for camera motion, lens styles, and character emotions; Marketing Studio adds more presets for UGC ads, posters, product shots, etc; and the ChatGPT Plugin gives access to all of its models plus skills for common tasks including narration and subtitles.

• Krea is now available in Slack, giving users access to its creative tools in the messaging app. instagram.com/krea_ai

Getty Images launched an MCP server, making its licensed visual content available to AI agents and workflows. newsroom.gettyimages.com

Even with the general availability of AI images, the requirement for physical photography – whether for events, legal, ethics, or preference — hasn’t gone away. Getty launched its own AI model but it was never any good; licensing and integration seems like a surer bet.

• FLORA introduced Fashion Studio, a design platform targeted at apparel brands, letting designers turn sketches into photorealistic garments and virtual photoshoot assets. flora.ai

• FLORA added Custom Actions, letting users build bespoke tools such as graphical filters and shaders into their workflow. x.com/floraai

• Wan added Agent Skills to its video platform, for building reusable creative workflows. x.com/Alibaba_Wan

• Tavus made it easier to deploy its PAL avatars using a widget, an embed, or on its own landing page. x.com/tavus

• Google Gemini and Flow now let you toggle off the visible watermark from generated images and video. x.com/GeminiApp
  A small but very welcome change.

Assistants

Google released Gemini 3.7 Flash. The new ‘workhorse’ model brings improved performance for more complex (agentic) tasks such as coding and Web design, and in “knowledge-dense fields” e.g. finance and law. blog.google

It’s available now in the Gemini app and Spark, AI Mode in Search, and in Antigravity and AI Studio at a lower introductory price. It’s only three weeks since the last version of Gemini Flash was released; looks like Google is back on it.

xAI released Grok 4.6 “with a particular focus on long-running agents and more ambitious interactive and visual work”. x.ai

It’s available via API, in Grok Build, and Cursor (now officially part of SpaceX).

• Grok Bot is an always-on AI assistant for team workflows, with dedicated cloud computers to automate browser and workplace tasks. x.ai

• Grok added a Voice connector to generate synthetic voices for memos, podcasts, and more. x.com/grok

Please think twice before using Grok.

• ChatGPT added Computer History, turning activity across apps and websites into ‘memories’ and a timeline that ChatGPT and Codex can reference. x.com/OpenAI

• The Claude plugin for Chrome can run full Cowork sessions, syncing browser tasks and chat context seamlessly across devices. claude.com

Claude will use invisible watermarking in its text output in future releases. anthropic.com

The link above explains how it works; essentially, certain word choices will form a pattern that machines can detect. It mostly works on large blocks of unmodified text; if you heavily edit the output, or even use a custom writing style, it will be less effective. People are freaking out about this, but Gemini has been doing it for two years and the world hasn’t ended yet.

Meta introduced Muse Glimmer, a smaller AI model that’s designed for agent workflows to run locally on consumer hardware. research.meta.ai

Probably the most interesting thing about this is that, continuing the theme of this week’s newsletter, it’s an open-weights release — and Meta says it will open the weights of Muse Spark 1.2 in the near future. It’s worth reading Mark Zuckerberg’s reasoning behind why this is important: The Future is for Everyone.

For everyone to be part of the future, everyone must have the ability to use superintelligence to improve their lives and shape the world. We will offer free versions that will be accessible to billions of people. For those who want to pay to use more compute, there will be a dynamic auction mechanism that will guarantee that everyone gets the lowest price possible for the intelligence and compute they’re using while also ensuring the capacity is used for whatever people collectively find most valuable.

Alibaba released Qwen3.8-27B, an open weights variant of its latest top-class model (which scores particularly highly on the Arena leaderboard for coding) that’s optimised to run on consumer hardware. x.com/Alibaba_Qwen

I don’t cover the release of every LLM (there are too many!) but this deserves a mention because a) it’s another open weights release, and b) Qwen3.8 is a powerful model being given away for free, more or less; an example of AI being used as a soft power play.

• Manus announced its timeline to return to independence following its separation from Meta. Affected users will need to back up and restore recent data, to comply with regulatory requirements. manus.im

• Microsoft is removing some of its custom Copilot features including Group Chat, Labs, and the Mico persona. support.microsoft.com
  These were all part of Mustafa Suleyman’s efforts at “humanist AI”, which tells you something about the company’s plans for the future.

Social

Meta launched Threads as a native VR app for Quest headsets, letting users browse feeds, compose posts using voice dictation, and navigate with hand tracking or controllers. meta.com

Honestly, I didn’t see this coming. But I suppose that as Meta has adapted Threads for the Meta Ray-Ban Display glasses it wasn’t too much extra development work to do this.

• YouTube is changing its Partner Program, raising monetisation eligibility thresholds for new creators and Shorts while expanding Premium Lite revenue sharing globally. blog.youtube

This was an actual rarity: a quiet week! I haven’t had one of those for a long time. So rather than pad this out, I’m just going to get on with it.

Hope you’re well!

Synthetic Content

Black Forest Labs released FLUX 3 Video

Announced last week [TL 440], the multimodal generation model can create up to 20-second multi-scene HD clips with native audio, from text and image references. It promises world knowledge and grounding for more realistic outputs, and has Draft Mode for quickly previewing a clip before a full render.

bfl.ai/blog/flux-3-video

From my limited tests (see some examples here ⬇️) this seems a good model, perhaps not up there with the very best (Seedance) but near the top tier. That said, I haven’t had a lot of time to get used to it, so perhaps I’m being unfair. It’s available in most creative tools now, and BFL has stated that an open weights model will be released in the future.

Alibaba released Wan3.0, a new video model that can generate 30s clips with “reality-grade” quality, from references including text, images, audio, video, and documents. x.com/Alibaba_Wan

It’s currently in Public Beta, only available through Alibaba’s cloud platforms, so I haven’t tried it yet; I’ll cover it in more detail when it’s broadly released.

Alibaba released Qwen-Image-3.0-Pro, a version of its latest image model with higher resolution and detail, notably for dense information layouts. x.com/Alibaba_Qwen

It’s a good model, decently fast and priced, and good quality. I made this ⬇️ quick test; zoom in and read the text on the tattoo.

xAI released Imagine Image 2.0, aimed at “real work” with improved factuality, precision editing, and crisp text rendering. x.ai

It’s available on the Grok mobile and Web apps, using the Quality mode, with templates for common image workflows. It seems to be a completely new model, with structured layout generation similar to Reve and Ideogram [TL 432] (the interface looks very similar to Reve’s). I’m actually pretty impressed by it; take a look at my test ⬇️. But you’ll need to be on a paid plan to get the best out of it; I was able to generate only four images on my free account.

Creative Tools

Bytedance’s Seedance 2.5 is now generally available in pretty much all of the major creative tools, and through an API for developers. See last issue [TL 440] for an overview.

Adobe launched a unified plugin for ChatGPT that gives access to over 70 creative and productivity tools in Chat, Work, and Codex, letting users generate and edit visual assets and documents directly in chat. blog.adobe.com

This appears to be basically making the Firefly Assistant available through ChatGPT, which feels like a logical move.

invideo continued adding features to Agent Two: Workflows for repeating common tasks; Notebooks, to allow manual control at any stage of an agent flow; Slate, a timeline editor; Folders and Connect to give external users access to work in progress; Stock library footage; and an Enterprise tier for large organisations.

I haven’t used invideo but I’m impressed by these latest updates; they show that the company is ambitious, and how agents are completely transforming creative tools. And there are still two days of releases yet to come.

• fal launched Agent, a conversational interface to access any of the multiple models available on its platform. instagram.com/madeonfal
  Until this point fal has largely been an API platform with playground features; this seems to be a move to turn it into a full creative platform. There are already plenty of those, but fal does have a massive selection of models.

• Pika launched API Club, a $10 monthly membership that gives large discounts on leading synthetic media models. instagram.com
  This has the feeling of a flailing company trying to juice user numbers; it admits that it makes almost no profit from this service. I’m going to predict that Pika doesn’t last much longer in its current form.

• Suno teased new transparency and disclosure tools that “align with emerging industry standards” (not named), with digital watermarking coming soon. suno.com
  It also outlined its principles for responsible AI music creation, undoubtedly related to the recent copyright case that went against them in Germany; read Technollama for a good take on what happened and what it means.

Bonus Link

I began to suspect that the AI argument contains two different anxieties that are easily mistaken for one another. One concerns livelihood: what happens to me if a skill for which people once paid me becomes abundant? The other concerns identity: what happens to me if an ability that distinguished me becomes abundant?

The first question is economic and can be brutally real. The second is stranger.

The Generative Gazette

If anyone can create a great song, what makes me special?

There is something strange about the current anxiety over AI making music, and I think the easiest way to see it is to actually forget about music for a moment…

Read more

24 days ago · 1 like · Paul Henry Smith

Assistants

• OpenAI updated GPT-5.6 Sol in ChatGPT to give more reliable facts and focused answers to users on paid tiers, while providing free users unlimited text access to GPT-5.6 Luna. openai.com

• Meta launched Muse Code, a terminal coding agent powered by the updated Muse Spark 1.2 model that’s designed for long and complex agentic developer workflows. research.meta.ai

Social & Creators

• YouTube released new features for creators: Live Practice Mode to test streams privately before going live, and data-driven insights on hook, visuals, and pacing before posting. threads.com/@updatesfromyoutube

• YouTube’s shopping affiliate programme is now available in the UK, with select partner merchants (Wayfair, Currys, Debenhams, Boots, M&S, Etsy). socialmediatoday.com

• X launched the Original Content Rewards Program, replacing its previous Revenue Sharing (which rewarded only virality, not originality) with payments for top-performing “content you personally create that reflects your own voice, perspective, expertise, or creativity”. x.com/XCreators

• Disney and TikTok announced a content-sharing deal that will bring Disney’s IP into TikTok for eligible creators to use, with the best selected to show up in the Verts feed on Disney+. newsroom.tiktok.com

• Reddit is expanding its Rules Hub suite of content moderation tools, and announced plans to gradually restrict legacy features (including Old Reddit) to combat unauthenticated scraping. redditinc.com

Instead of an intro this week I direct you to my blog; I don’t post there often but this week the spirit moved me to write AI & Books: notes from a moral panic.

Also, I’m running a new experiment this week, trialling a generated overview of the top stories. This is my first effort; I’m aware of its defects, but I’ll refine it over the next few week…

Read more

This week Substack announced that it has added an AI text detector, estimating how much of any post’s content was human-written. Because, it says:

Read more

Britain’s second heatwave is over and the weather is back to being season-appropriate, so I’ve been outside making the most of it — hence I don’t have time to write more of an intro.

So, let’s get on with it. Hope you’re well!

Synthetic Content

Decart AI released Lucy 2.5, a real-time video …

Read more

No intro again this week, instead I’ve written a few paras about teens’ social media use at the end.

So, let’s get on with it. Hope you’re well!

Synthetic Content

Meta launched Muse Image

The new image generation and editing model features advanced reasoning capabilities, including tool use such as browsi…

Read more

This was a quiet news-week because of the 4th July celebrations in the US, meaning this is a fairly short newsletter. I’ve taken the opportunity to write a piece at the end, which I hope you find useful.

So, let’s get on with it. Hope you’re well!

Synthetic Audio-visual

Google released Nano…

Read more

Instead of an intro I’ve written about the workflow I use to write this email, which you can find at the end if you’re interested.

So, let’s get on with it. Hope you’re well!

Synthetic Content

Lots of small quality-of-life improvements to video models this month: Happy Horse 1.1 adds “enhanced motion, textu…

Read more

I’m trying to get a better grip of agentic workflows and MCP, so I paid for a Manus subscription and set myself a starter task: to create this week’s cover illustration. These were the steps I settled on:

1. Check the Airbase table [via MCP] where I store each week’s stories, and get all of the written summaries for each story.

2. From those, synthesise a…

Read more

Photography is the easiest way to create content today. As long as you have a camera, you can just point and click, and you have an image. But not every photograph is a good photograph. Your smartphone camera will run all sorts of algorithms to make your photograph look better, but there’s no guarantee of a good photograph. You can buy yourself th…

Read more

In lieu of an introduction I’ve written a short aside further down this week’s edition.

So let’s get on with it. Hope you’re well!

Synthetic Audiovisual

Reve 2.0

This new image model uses a novel structured layout approach to generation, providing precise visual control over individual elements in an image.

b…

Read more

This was a very quiet week for news, presumably because there was a public holiday in the USA. I’m currently recuperating from a minor surgery and I don’t have anything useful to say, so I’m just going to get on with it.

Hope you’re well! (I am.)

Synthetic Audio-Video

Microsoft announced MAI-Image-2.5, h…

Read more

It’s been two weeks since the last proper edition and a lot has happened, plus I’ve included some announcements from Google I/O that didn’t make it into the mid-week special. So this edition will be light on comment to avoid going over the length limit.

In which case, let’s get on with it. Hope you‘re well!

Synthetic Content

Runway launched Aleph 2.0…

Read more

I’m back from my break (Copenhagen, I loved it) with an unusual mid-week edition just to cover the news from this year’s Google I/O. There was a ton of announcements as usual, so I’m just going to highlight a few below. I’ve had to write this very quickly so I apologise for the brevity and any mistakes. The regular newsletter will be back on Monda…

Read more

This week’s edition is short and, if I’m being honest, somewhat bitty. You should know that there will be no regular edition next week as I’m off for a short break. However, I might send out a mid-week edition that covers the news from Google I/O 2026, which is happening next Tuesday, 19th May.

I think there are going to be a number of exciting ann…

Read more

NB This edition is a day later than usual because of the public holiday here in the UK.

In the past few weeks some clear trends have been emerging around AI assistants moving into creative design and workplace tools. I work in a marketing agency and creative design is a large part of our work, so we need to try to predict the impact of this — but p…

Read more

Everyone should read more history books. Right now we’re in a period of change that can feel overwhelming, and history is a way to put it all in context and make sense of it.

For example, I recently read Empires of Light, about the competition to turn electric light into a consumer product; the full transformation of leisure and work when homes and…

Read more

This was a busy week, it took me a lot of time to put this newsletter together, and I don’t have much space to write an intro, so let’s get on with it. Hope you’re well!

Synthetic Audio-Visual

Midjourney released V8.1 (alpha) with the more consistent aesthetic of previous versions, the return of image prompts, and improvements to resolution and spee…

Read more