Pick a preset or upload any edit, add photos of your characters, and get your own version. You connect the Higgsfield MCP and type /katana in Claude. 762 likes and 59K views. Katana first shipped Oct 8 per the changelog.
RuntimeWire reports that Higgsfield's Oct 12 guide pitches batch requests via ChatGPT or Claude and recurring scheduled jobs in Supercomputer. Each run still costs credits and still needs human review.
'Go from a prompt to generating, editing, and iterating on images and videos, all without leaving Claude.' The MCP first launched Jun 18. Its Claude page lists GPT Image 2, Nano Banana Pro/2/2 Lite, Seedream 5.0/4.5 and Kling 3.0.
'do you wanna join OpenArt? 👀' got 164 likes and 12K views, with an Ashby jobs link in the thread.
Two new blog guides aimed squarely at product and e-commerce video buyers. The Ad Awards winners were due Oct 12, but I couldn't find a published winners list yet.
1,272 likes and 130K views. Hedra posted 'Something new from Krea coming soon. On Hedra.' and fal teased it too. Heather Cooper says it lands Wednesday. Krea already has a Krea 3 page, and partner each::labs quotes Krea calling Krea 3 'our new foundation model, with amazing editing capabilities'. It's not confirmed that this tease is Krea 3.
He locks an ultra-realistic first frame in Magnific One, then animates it in Seedance, with full prompts shared. The main thread has 768 likes and 168K views, and a follow-up has 7K views. The API and changelog have nothing new since Magnific One shipped Oct 8.
Midjourney's updates (Oct 7: shared folders, Thinking Mode test) and BFL's release notes (FLUX 3 Image, Oct 1) have nothing new. Neither account posted on X in the window.
Vidu's pitch is better facial emotion and micro-expressions, plus a run of creator spotlights using invite codes.
Draggable virtual lights with direction, color, brightness and softness controls.
The full launch is still listed only as 'October', with no date. Kling 4.0 Flash has been in limited early access since Sep 28.
Theoretically Media's subtitled demo uses King of Comedy footage and got 473 likes and 28K views. Reports say you can draw a camera trajectory to previsualize a shot.
Prompts are shared in the thread.
Led by Dharana Capital, with Founders Future, Neotribe, YC and f7. It's based in California and Bengaluru, and says 1,000+ companies (Atlassian, Zendesk, Gainsight) use it, with an 80% cut in video costs.
Heather Cooper gave Claude six prompts. It picked 12 shots, generated a score and title card, and graded the edit. invideo MCP launched Oct 2.
Turns an idea or script into a short film: story development, shot planning and character looks.
Starts from one image with WASD controls, at 24fps on a single H800, generating video and stereo sound together. Preview weights are on Hugging Face.
Heather Cooper (@HBCoop_) is the most useful creator today. She put two Midjourney characters into one scene via Dreamina and Seedance 2.5 (https://x.com/HBCoop_/status/2109735399820325121), had Claude edit inside invideo over MCP, and says Krea's launch lands Wednesday. TechHalla (@techhalla) is all-in on Magnific One + Seedance first-frame workflows (168K views). Min Choi (@minchoi) showed Manus 2.0 building a playable 3D game from one prompt (https://x.com/minchoi/status/2109774874919002194) and amplified Karpathy's capability-gap post (46K views). Theoretically Media (@TheoMediaAI) had the first look at ByteDance's world model. Dave Clark (@Diesol) did an interview on AI in Hollywood. Kris Kashtanova (@icreatelife) shared photo-to-diptych prompts. Chubby (@kimmonismus) and Justine Moore (@venturetwins) spent the day on Meta's Muse Wonder frontier-model launch, which is general AI rather than image/video. Rory Flynn, Nick St. Pierre, Ammaar Reshi and PJ Accetturo didn't post in the window. I couldn't read Leonardo's account.
X: latest posts from 32 watchlist accounts via FxTwitter, window since Oct 12 1:30am IST. 31 read fine and Leonardo failed. Midjourney, BFL, Magnific, Runway, Pika, Google DeepMind and Stability had no posts in the window. Official changelogs checked: Higgsfield, BFL, Magnific API, OpenArt what's-new and blog. None had new entries after Oct 8 except OpenArt's blog. Web and press searches covered launches, pricing and funding. Instagram and TikTok can't be read without a login, so they're not covered. Added to the watchlist: Dreamina/Seedance (@Dreamina_AI), invideo, Manus, Noiz AI, Vixel AI, Clueso.