The director is becoming a prompt
the director is becoming a prompt: gemini omni flash tops the ai video arena while costing a quarter of the flagship alternatives, seedance 2.5 holds a single 30-second take with audio and reference control, and sora 2 simulates physics like glass breaking from the point of impact. the bar is no longer generating a clip.
it is holding a story together past thirty seconds.
Context
Agentpedia reports that as of mid-2026 Seedance and Google's Gemini Omni Flash are the two co-leaders in AI video, with Omni Flash first for text-to-video (preliminary) on Arena.ai blind-vote boards. Google DeepMind's model card describes Omni Flash as creating and editing from any input, starting with video.
Omniart Studio says Sora 2 bets on long coherent single takes of up to roughly 20 seconds with strong physics, while Omni Flash clips cap at 10 seconds.
The Omni Flash arena lead and Sora 2's physics strength match the sources. A quarter of the flagship cost, Seedance 2.5 holding a single 30 second take with audio and reference control, and glass breaking from the point of impact were not seen in the sources read, so unsupported here, not refuted. Omniart's figures of 10 and 20 second clips are for Omni Flash and Sora 2.
Related work
- Agentpedia: Gemini Omni Flash vs Veo, Sora, Seedance, Kling ↗Arena standings.
- Google DeepMind: Gemini Omni Flash model card ↗Model card.
- Omniart Studio: Omni Flash vs Sora 2 vs Seedance 2 ↗Clip length comparison.
Watch next
- Check Seedance 2.5's documented clip length.
Sources
- Agentpediaagentpedia.codes
- Google DeepMind, Aug 27, 2026deepmind.google
- Omniart Studio, Jul 1, 2026omniart.studio
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 2 October 2026 at 21:44 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →