The last 24 hours in AI video were dominated by Adobe MAX 2026, an open-weight 4K image model from MiniMax, and a reminder that code is becoming part of how videos are made. Here is everything that matters for AI video creators on November 11, 2026, in one page, with what is shipping, what is beta and what is only a claim.

The short version

  • Adobe launched Firefly Video and agents across Firefly and Premiere at MAX 2026, with Kling 4.0 and Seedance 2.5 inside Firefly.
  • Adobe is previewing Project Live Set, real-time generative video editing, at its MAX Sneaks session on November 11.
  • MiniMax released MiniMax H3-Image, an open-weight 4K image model with up to 16 references.
  • Video-Coder, a new paper, makes MiniMax H3 more physically accurate by building a 3D draft in code first.
  • Gaxos launched ArtGen Muse, a lower-cost video model, and AI Era Corp bought rights to a script-to-video pipeline.
  • Google confirmed that Gemini Omni 1.1 Flash calls move to 1.2 on December 3.

1. Adobe Firefly Video and Creative Agent (Adobe MAX 2026)

On November 10, Adobe repositioned Firefly as an all-in-one AI creative studio. Firefly Video takes a project from story to scenes, shots and a timeline, with captions, transitions and b-roll generated in place. Adobe Creative Agent now runs across Firefly, with a Plan Mode and a Model Selector that can choose a model for you. Partner models named by Adobe include Dreamina Seedance 2.5, Kling 4.0, GPT Image 2.6 and Nano Banana 2.1. Premiere mobile adds licensed music from Universal and Warner and a Creative Agent assistant in private beta that makes first cuts from your footage.

Status: Edit Text, Retouch, Relight and Quick Actions are generally available; most other tools are beta or preview. No new pricing was announced. Full details: Adobe Firefly Video at MAX 2026.

2. Adobe Project Live Set: real-time AI video editing (preview)

Adobe scheduled a preview of Project Live Set for its MAX 2026 Sneaks session on November 11 (5:30 p.m. ET), according to Variety: a technology for generating and editing video in real time while it plays. Sneaks are research previews: Adobe has not said whether or when it will ship, so treat it as a direction rather than a product.

3. MiniMax H3-Image: open-weight 4K images

MiniMax released H3-Image on November 11: a 33.1B-parameter model for text-to-image, multi-reference generation and editing, with about 16-megapixel output and up to 16 reference images. MiniMax claims 4K in under 10 seconds. The weights are on Hugging Face under a community license that requires authorization for some uses. It shares its text encoder with the MiniMax H3 video model, which makes it a strong first-frame generator for image-to-video. Full details: MiniMax H3-Image explained.

4. Video-Coder: code-first drafts for more realistic physics

A paper from the University of Central Florida, Microsoft Research, Tsinghua and others trains a 9B agent to write Blender code, render a rough 3D draft and pass it to MiniMax H3. On VBench-2.0, the physical score rises from 54.58% to 78.19%, according to the authors. It is a preprint, but it fits a clear trend: code handles motion and physics, video models handle the look. Full details: Video-Coder explained.

5. Gaxos ArtGen Muse and AI Era's script-to-video deal

Gaxos.ai (Nasdaq: GXAI) launched ArtGen Muse on its ArtGen platform on November 10, pitched as high-quality video at a lower generation cost. The announcement gives no specs, resolution, length or prices. The same day, AI Era Corp. announced it had acquired exclusive rights to an AI script-to-video pipeline and said it can produce an episode for about $78. Both are company press releases with no independent data.

6. Gemini Omni 1.1 Flash moves to 1.2 on December 3

Gemini Omni 1.2 Flash went generally available on November 5 with 15-second clips, six reference images and video output at $12 per million tokens. From December 3, requests to the 1.1 model ID will be routed to 1.2 and billed at 1.2 rates, so developers should re-test their prompts now. Full details: Gemini Omni 1.2 Flash pricing and specs.

Also this week

  • Claude Fable 5.2 (November 9): Anthropic's new model, which Anthropic says can create animations and educational videos with sound in code. It is now available in Nano Banana Video Studio. See what's new in Claude Fable 5.2.
  • ElevenLabs Scribe v3: new transcription models, useful for captions and subtitles, with 27% fewer word errors than v2 in ElevenLabs' own multilingual tests.

What it means for creators

  • Choose models per shot. Adobe now bundles the same partner models other studios use. The skill that matters is knowing which model fits which shot.
  • Start from a strong frame. Image models like MiniMax H3-Image and Nano Banana 2.1 make image-to-video more consistent than text alone.
  • Use code where precision matters. For logos, data, diagrams and product turntables, coded animation beats pixel generation.

You can try that whole workflow, image, video models and coded 3D animation on one timeline, in Nano Banana Video Studio.

FAQ

What was the biggest AI video news on November 11, 2026?

Adobe's Firefly Video launch at MAX 2026, with Kling 4.0 and Seedance 2.5 inside Firefly, and MiniMax's open-weight H3-Image model.

Did any major new video model launch in the last 24 hours?

No flagship video model launched. The main model release was MiniMax H3-Image, an image model. Gaxos launched ArtGen Muse, with no published specs.

Is Adobe Project Live Set available?

No. It is a Sneak, Adobe's preview of technologies that may or may not ship.

Sources: Adobe newsroom; No Film School; Variety; MiniMax; Video-Coder paper; Gaxos press release; AI Era Corp. press release; eesel; The Neuron. Status as of November 11, 2026.