Update — October 13, 2026: Gemini Omni 1.1 Flash is Google’s current video model and the one to compare against, not Veo 3.1. It ranks above Veo 3.1 on both Artificial Analysis boards and costs a fraction of Veo’s standard tier.
Gemini Omni 1.1 Flash, released by Google DeepMind on August 27, 2026, treats video as a conversation. You generate a clip from text or an image, then keep talking to it — “make it dusk,” “remove the car,” “continue the scene” — and the model applies the change while keeping the parts you liked. Under the hood it is natively multimodal: it processes text, images, audio and video together, and every clip comes back with generated sound.
What it does
- Generate. Text-to-video and image-to-video, 3–10 seconds at 24 fps, with synchronized audio generated alongside the picture.
- Edit by conversation. Through the Interactions API, each follow-up message edits the previous result instead of starting from scratch.
- Extend. Append to the end of a clip, turn after turn, to build a longer scene. Extension works on clips the model generated and on uploaded clips of 10 seconds or less.
- Interpolate. Give a first frame and a last frame and the model fills in the transition.
- Reference. Supply images of a character or product, or up to three short video clips of a likeness, so the subject stays consistent.
- Resolution. Choose 360p for drafts, 720p (the default and native render), or upscaled 1080p and 4K.
Google puts world knowledge at the centre of the pitch: the model combines an understanding of physics with Gemini’s knowledge of history and culture, aiming past photorealism toward scenes that make sense.
How it ranks
| Artificial Analysis (October 2026) | Gemini Omni 1.1 Flash | Wan 3.0 | MiniMax H3 (768p) | Veo 3.1 |
|---|---|---|---|---|
| Text-to-video Elo (with audio) | 1115 (#8) | 1158 (#1) | 1141 (#4) | 965 (#19) |
| Image-to-video Elo (with audio) | 1130 (#3) | 1118 (#5) | 1131 (#1) | 1034 (#13) |
| Price per minute (1080p) | $9.12 | $12.00 | $4.80 (T2V) | $24.00 |
The text-to-video and image-to-video boards use separate scales, each anchored at Kling 3.0 1080p Pro = 1000, so compare within a row. The top three image-to-video entries are separated by a single Elo point, so the order among them is noise. A Gemini Omni Flash entry also leads Artificial Analysis’s silent image-to-video board (Elo 1368).
Pricing
Google bills video by output token, not by the second, and quotes about $0.10 per second at 720p.
| Output resolution | Tokens per second | Approx. price per second |
|---|---|---|
| 360p (draft) | 1,931 | $0.034 |
| 720p (default) | 5,792 | $0.101 |
| 1080p (upscaled) | 8,688 | $0.152 |
| 4K (upscaled) | 17,376 | $0.304 |
Video output is $17.50 per 1M tokens; text, image, video and audio input are $1.50 per 1M. There is no free API tier — the free tier column in Google’s pricing table reads “Not available” — so the first clip you generate bills your account. For comparison, Veo 3.1’s standard tier is $0.40 per second at 1080p.
Limitations
- Short clips. Each generation tops out at 10 seconds. Longer scenes mean chaining extensions, and each extension bills its own output.
- Upscaled, not native, high resolution. Google’s docs label the 1080p and 4K tiers “upscaled.”
- Regional limits. Editing or extending uploaded videos is not available in the EEA, Switzerland or the UK, though editing clips the model generated is.
- Reference limits. Audio references are not supported; video references accept at most three clips of three seconds each, and any audio in them is ignored. Voice editing is not supported.
- Language. English is fully supported; Google has not evaluated other languages.
- Watermarking. All output carries a SynthID watermark.
Who it’s for
Omni 1.1 Flash is the right pick when you want to iterate rather than re-roll — product teams animating still images, creators who refine a shot through conversation, and developers who want editing and extension in one API. If you need a single 30-second take, look at Seedance 2.5; if text-to-video quality at the lowest price matters most, look at MiniMax H3 or Wan 3.0.
Best for
- Animating product shots, illustrations and photos with sound (image-to-video)
- Iterating on a clip through conversation — change the lighting, remove an object, swap a prop
- Building a longer scene by extending a generated clip turn by turn
- Cheap 360p drafts before committing to a 720p or upscaled final
Pros & cons
Strengths
- Conversational editing — describe a change and the model applies it while keeping the rest of the clip, through the Interactions API
- Ties for the top of Artificial Analysis image-to-video (Elo 1130) at $9.12 per minute, versus $24.60 for Seedance 2.5
- Scene extension, first-and-last-frame interpolation and subject references (images of a character or product) built in
- Cheap 360p draft tier at about $0.034 per second for prototyping before you spend on a final render
- Every clip carries an invisible SynthID watermark for provenance
Limitations
- Short clips — each generation is 3–10 seconds, so longer scenes mean chaining extensions
- 1080p and 4K outputs are upscaled from a 720p-class render rather than generated natively
- Only
- Editing or extending uploaded videos is unavailable in the EEA, Switzerland and the UK, and audio references are not supported
- English is the only language Google fully evaluated
Frequently asked questions
What is Gemini Omni 1.1 Flash?
It is Google DeepMind's video generation and editing model, available as gemini-omni-1.1-flash on the Gemini API and Gemini Enterprise Agent Platform. It takes text, images and video as input and returns a clip with generated audio, and you refine the result by chatting with it. Google released the production-ready 1.1 version on August 27, 2026.
How much does Gemini Omni 1.1 Flash cost?
Video output is billed at $17.50 per 1M tokens, which Google converts to about $0.10 per second at 720p. At the same rate a second costs roughly $0.034 at 360p, $0.152 at 1080p and $0.304 at 4K. Input is $1.50 per 1M tokens for text, image, video or audio, and there is no free API tier. Artificial Analysis lists $9.12 per minute, which is the 1080p rate.
How long can a Gemini Omni 1.1 Flash video be?
Each generation is 3 to 10 seconds at 24 frames per second. Longer scenes are built by extending a clip at its end, one turn at a time. Uploaded clips for editing or extension must be 10 seconds or shorter.
Is the 4K output real 4K?
Google's documentation labels 1080p and 4K as upscaled outputs; the default and native render is 720p. Treat the 4K tier as an upscale for delivery, not as native 4K generation.
How does Gemini Omni 1.1 Flash rank?
On Artificial Analysis's blind-vote boards it is