
Switch genres mid-track with Eleven Labs Music V2
Top Stories - Overview
It's been a quiet week for AI production and post news, but the shape of where this is all going got a little clearer: the story seems to no longer be about any one, single model, but how fast they're all collapsing into one place you can actually work, without jumping between tools.
Microsoft showed up at Build with an image model that can finally spell, fixing the thing that other models mangle: type. ElevenLabs shipped music you can legally hand a client, and Runway quietly turned itself into a switchboard for everyone else's generators.
And then Martin Scorsese, of all people, put his name on AI storyboards, which tells us the conversation has finally moved from whether to use AI in production to how to use AI in production. Or, in Scorsese's case, pre-production.
Featured Stories
1. Microsoft unveils MAI-Image-2.5 and Flash variant at Build 2026
Microsoft Blog
What Happened: Microsoft used Build 2026 to broaden access to its MAI-Image-2.5, an updated in-house image model with image-to-image editing, control-with-preservation tools, and notably strong text and layout handling. It debuted at No. 3 among image families on the public arena, with a faster Flash variant alongside, reachable through Microsoft Foundry.
Why Is This Important? Forget for a second where Microsoft put it (PowerPoint and OneDrive). What matters for motion and title artists is that reliable typography and "preserve this, change that" editing are turning into the baseline, not the exception. Between this and Nano Banana 2 topping the arena down in the links below, text rendering is quietly getting solved across, and that's the pressure that drags these same controls into the Adobe tools you actually work in. Expect them sooner rather than later.
Tags: #AITools #GenerativeAI #MotionGraphics #Microsoft #MAIImage
2. ElevenLabs Music v2 switches genres mid-track, cleared for commercial use
TechCrunch
What Happened: ElevenLabs launched Music v2, which shifts genres inside a single track, builds songs section by section, and regenerates any part of a song by prompt without touching the rest. It adds non-musical sound effects, performs more reliably across languages, and ships trained on licensed data with commercial use cleared.
Why Is This Important? For anyone scoring spots or cutting trailers, the licensing line is the whole story. Music trained on licensed data and cleared for commercial use means you can drop a generated cue into a client deliverable without the rights cloud that hangs over Suno and Udio. The section-level regenerate is the practical part day to day: fix a bridge or restyle a chorus without rebuilding the track from scratch.
Tags: #AIMusic #AIAudio #SoundDesign #AITools #ElevenLabs
3. Martin Scorsese signs on with Black Forest Labs, used FLUX to storyboard his next feature
IndieWire
What Happened: Martin Scorsese went public as a partner and adviser to Black Forest Labs, the German FLUX maker, releasing a video of himself using FLUX.2 to storyboard his next feature, "What Happens at Night," with Leonardo DiCaprio and Lawrence. He very carefully framed it strictly as pre-vis, not final imagery.
Why Is This Important? When the most craft-protective director alive uses AI to show a production designer and DP what's in his head, the pre-vis argument is effectively settled. The practical read for working pros: storyboarding and previs are where generative image tools land first and cleanest, because nobody's shipping the output. But that's cold comfort to the storyboard and concept artists now pushing back hard, because pre-vis is their craft, and this is the first place it gets felt.
Tags: #FilmProduction #GenerativeAI #PreVis #BlackForestLabs #AITools
4. Runway opens its API to outside models
Runway Changelog
What Happened: Runway added its Gen-4.5 model to the API and folded in a roster of outside generators that now run inside Runway along with its Premiere and Resolve integrations. Among the arrivals is ByteDance's Seedance 2.0, which brings native synchronized audio plus reference control across text, image, video, and audio inputs.
Why Is This Important? For editors and post teams, this is a workflow story, not a model story. One account and one set of credits, and you can audition models against each other without bouncing between six logins and six export paths. The friction in generative work was never really quality, it was the time lost shuffling between tools.
NOTE: Seedance's reach for consistent looks is real, but before you feed it client material, ByteDance's terms and privacy handling are..thin.
Tags: #VideoGeneration #PostWorkflow #AIVideo #AITools #Runway
5. getimg.ai positioned as a multi-model image and video aggregator
getimage.ai
What Happened: In case you missed it, and you'd be forgiven since there's no launch or funding round behind it: Warsaw-based getimg.ai has quietly become one of the more capable multi-model hubs going. It boasts a deep bench, FLUX.2, Nano Banana 2, Seedream, GPT Image, plus a stack of video generators, into one subscription with automatic model selection and built-in editing.
Why Is This Important? For a solo creator or small shop drowning in single-tool subscriptions, an aggregator that picks the right model for you can replace a drawer full of logins and a pile of separate bills. The tradeoff (because there’s always a tradeoff) is the familiar one: you sit at least a layer back from each model's newest features and tightest controls. It’s worth looking at if you’re drowning in subscriptions, less so if you need the bleeding edge.
Tags: #AITools #GenerativeAI #VideoGeneration #ImageGeneration #getimg
6. Google Veo 3.1 free for personal Google accounts via Vids Chrome Unboxed
What Happened: Google opened Veo 3.1 to any personal Google account through Google Vids, dropping the prior Workspace requirement. Free accounts get a monthly allotment of generations, roughly ten, with clips around eight seconds, created from text prompts or by animating a still image.
Why Is This Important? The significance isn't the free tier itself, it's that one of the better video models now sits a click away inside software you likely already have open. That changes the conversation on set and in the edit bay: the people briefing you (i.e. your clients) can generate a rough version themselves.
Tags: #VideoGeneration #AIVideo #AITools #Google #Veo
Worth a Listen
Intelligent Machines Podacast #873: "AI in Hollywood" with Robert Tercek
TWiT.TV I YouTube
What Happened: Futurist and former MTV creative director Robert Tercek talks about what generative tools are actually doing to the movie business: the new guild and studio agreements covering AI and synthetic actors, why a broken financial model is forcing Hollywood to reinvent itself, and the coming flood of scripted work as the tools get cheaper and YouTubers like Kane Pixels (Backrooms) cross into professional production. He walks through AI as a production assistant for plot structure, continuity, and asset management, and makes the case, by way of his own Neura Studios, that animation pipelines sidestep the uncanny-valley problems that still dog live-action.
Why it's worth your time: Tercek lands where most of us already are, that AI has no discernment and no taste, so it works as a collaborator and falls apart as a replacement. It's also a clear articulation of the role working professionals actually play in an AI workflow, and a useful frame of reference for the Martin Scorsese news this week.
General AI News
Nano Banana 2 leads LM Arena text-to-image leaderboard — https://getimg.ai/blog/gpt-image-2-rumours-leaks-release-date-2026 — Google's Gemini 3.1 Flash Image holds the top spot on the public text-to-image arena, with OpenAI close behind.
EU AI Act accelerates synthetic-media transparency to December 2, 2026 — https://www.aiapps.com/blog/ai-news-breakthroughs-launches-trends-must-read/ — Disclosure and C2PA watermarking rules move up, hitting anyone delivering AI content to European audiences.
Microsoft adds MAI-Voice-2 and MAI-Transcribe-1.5 at Build 2026 — https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/new-mai-models-in-microsoft-foundry-across-text-image-voice-and-speech/4524632 — A multilingual TTS model with voice cloning and an updated transcription model aimed at enterprise audio workflows.
OpenAI launches self-serve Ads Manager inside ChatGPT — https://www.marketingprofs.com/opinions/2026/54655/ai-update-may-8-2026-ai-news-and-views-from-the-past-week — A self-serve ad platform inside ChatGPT, with major agency holding companies on board.
Google I/O 2026 recap: Gemini 3.5, Flow Music, Google Pics — https://blog.google/innovation-and-ai/technology/developers-tools/google-io-2026-collection/ — The developer-conference roundup, including Gemini 3.5, upgrades to Flow Music, and a new design tool called Google Pics.
Anthropic files confidentially for IPO — https://www.theverge.com/ai-artificial-intelligence/941016/anthropic-has-officially-filed-to-go-public — Anthropic filed a confidential draft to go public, opening the year's busiest AI IPO window.
Kuaishou weighs a KLING AI spin-off — https://baike.baidu.com/en/item/KLING/1478239 — Kuaishou's board is weighing a restructuring that would carve its KLING video model into its own entity.
Alibaba Cloud's Bailian to host third-party models including KLING — https://baike.baidu.com/en/item/KLING/1478239 — Alibaba's Bailian platform will keep open access and host outside models including KLING, another point on the aggregation curve.