Google made its conversational video model generally available on August 27, 2026. The model ID is gemini-omni-1.1-flash, replacing the gemini-omni-flash-preview endpoint released on June 30. If you use the preview endpoint, migrate before Google disables it on September 30, 2026.
This is more than a version change. The GA release adds four production-focused capabilities:
- Scene extension up to 40 seconds
- First-and-last-frame control
- Lower-cost 360p drafts
- 4K upscaling
This guide covers the new capabilities, pricing, supported platforms, limitations, and migration steps.
For background on the Omni family and Google’s reasoning-first approach to video generation, read What Is Gemini Omni. This article focuses on the 1.1 release.
What Gemini Omni 1.1 Flash does
Omni 1.1 Flash accepts text, images, and video, then returns video. It runs through the Interactions API instead of generateContent, the same surface Google uses for stateful, multi-turn generation.
That state lets you generate a clip and then edit it in a follow-up turn without re-uploading it.
The model supports five task types:
- Text to video: Generate a clip from a prompt.
- Image to video: Use one image as the opening frame or as a motion and style guide.
- Reference to video: Provide up to three reference clips, each up to three seconds, to guide movement, appearance, and character consistency.
- Edit: Modify content in a video generated earlier in the conversation.
- Extend: Append another 10 seconds to the end of a clip.
Audio in reference clips is ignored. The model uses them only for motion and appearance.
Extend scenes with 10 seconds of context
Scene extension is the biggest change. The preview model examined only the final second of a clip before continuing it. That could preserve the color palette, but characters often drifted and camera movements reset.
Omni 1.1 analyzes up to 10 seconds of previous context. Google describes the result as “improved visual consistency and narrative adherence”.
You can extend a clip in 10-second increments, up to 40 seconds cumulatively.
Plan around these constraints:
- Extension only appends to the end. You cannot prepend footage or insert content in the middle.
- Uploaded videos can contain up to 10 seconds of input.
- Multi-turn interactions can retain earlier state, allowing the model to work with the existing interaction context.
The 40-second scene extension walkthrough covers the request format and common seams.
Control the first and last frames
You can provide a first frame, a last frame, and a prompt describing the movement between them. The model generates the footage connecting the two images.
Useful applications include:
- Camera orbits
- Zoom transitions
- Looping clips
- More predictable video-tool output
Text-to-video leaves every frame to the model. Keyframe interpolation fixes two points and limits what the model must invent between them.
Use 360p for cheaper drafts
Omni 1.1 generates 360p previews up to 60% faster than 720p at one-third the cost.
Google’s recommended workflow is:
- Draft at 360p until the prompt works.
- Re-render the selected clip at 720p, 1080p, or 4K.
Because most prompt-iteration clips are discarded, cheaper drafts can significantly reduce production costs. See the full pricing breakdown for the per-second calculations.
Set the resolution in the response format. The 1080p and 4k options upscale generated frames rather than producing native high-resolution renders.
Where Gemini Omni 1.1 Flash runs
The model is available through:
- The Gemini API in Google AI Studio
- The Gemini Enterprise Agent Platform API
- Google Flow for AI Plus, Pro, and Ultra subscribers
- The Gemini app, where subscribers can use scene extension
Google also announced launch partners running the model in their products:
- Adobe Firefly
- Figma Weave
- Runway
- GMI Cloud
The API has no free tier. Unlike text Flash models, where AI Studio provides a rate-limited free tier, Omni output is billed from the first request.
The model is free on YouTube Shorts and YouTube Create, but those are separate products with different limits. See how free access works for a comparison of each option.
Limitations to account for
Review these restrictions before designing around the model:
-
Limited prompt controls: System instructions, temperature,
top_p, stop sequences, and negative prompts are unsupported. To exclude something, state it in the regular prompt. - Regional restrictions: Uploading and editing videos are unavailable in the European Economic Area, Switzerland, and the UK. Editing images containing minors is also unavailable there. Model-generated videos remain editable in those regions.
- Recognizable people: Editing certain identifiable individuals is blocked.
- Dialogue on uploaded videos: You can extend an uploaded clip silently or use a multi-turn interaction, but you cannot add dialogue while extending someone else’s upload.
- One video for reasoning: The model cannot reason across multiple input videos.
- English support: English is fully supported; other languages are untested.
Every output includes an invisible SynthID watermark that viewers cannot see but software can detect. Account for this if your product makes provenance claims.
Gemini Omni 1.1 Flash or Veo 3.1?
Google now offers two video-generation families through the same API key:
- Veo 3.1: A cinematic renderer with native audio
- Omni 1.1 Flash: A conversational model designed for multi-turn generation and editing
An Omni 1.1 720p second costs approximately one-quarter as much as a standard Veo 3.1 second. Veo does not provide an equivalent multi-turn edit loop.
Read the side-by-side comparison to determine which model fits your use case. If you already integrate Veo, the Veo 3.1 API guide remains current. This release does not deprecate Veo 3.1.
Migrate from the preview endpoint
Requests using gemini-omni-flash-preview stop working after September 30, 2026.
The basic migration is usually a one-line model ID change:
gemini-omni-flash-preview
becomes:
gemini-omni-1.1-flash
Before switching, verify two response differences.
1. Resolution defaults
720p is now the default. The 360p and 4K options are new. If your code assumes a fixed output size, verify the actual resolution returned by the GA model.
2. Large response handling
Videos larger than 4 MB are returned as a URI instead of inline base64. If your handler reads only output_video.data, higher-resolution responses may appear to be empty.
The safest migration process is:
- Save the request in an API client.
- Store the model ID in an environment variable.
- Run the same request against both model IDs.
- Compare the response shape at every resolution you plan to support.
- Update handlers to support both inline data and URI responses.
- Switch production traffic before the deadline.
Apidog is useful for this workflow: create one saved request, put the model ID in an environment variable, switch environments, and compare the responses. The API walkthrough shows the setup step by step.
FAQ
What is the model ID for Gemini Omni 1.1 Flash?
gemini-omni-1.1-flash. The retiring preview ID is gemini-omni-flash-preview.
When did Gemini Omni 1.1 Flash launch?
The GA release launched on August 27, 2026. The preview launched on June 30, 2026.
Is Gemini Omni 1.1 Flash free?
No. Omni has no free API tier, so every generation is billed. Text models such as Gemini 3.6 Flash still provide a free AI Studio tier.
How long can a Gemini Omni video be?
Clips generate at 10 seconds. Scene extension can increase the total duration to 40 seconds in 10-second increments.
Does Gemini Omni generate audio?
The documentation covers video output and ignores audio in reference clips. Veo 3.1 is the model with native audio generation. If audio is required, start with the Veo 3.1 API.
Can I edit a video I filmed myself?
Yes. Uploaded input is limited to 10 seconds, and the feature is unavailable in the EEA, Switzerland, and the UK. Upload the video with the Files API and pass its URI as input.
Ship against the GA model
Omni 1.1 Flash is the first version that looks practical for production: scene extension maintains shot continuity, keyframes make output more predictable, and 360p drafts make iteration affordable.
Create one saved request against gemini-omni-1.1-flash, test every resolution you plan to support, handle URI-based responses, and migrate before the end of September. Download Apidog to keep the migration check saved before the deadline.
Top comments (0)