Gemini Omni 1.2 Flash is now generally available in the Gemini API, and Google moved it to GA on November 5, 2026. It makes clips up to 15 seconds long, edits them through conversation and cuts the video output price by 30%. On December 3, every request that still names Omni 1.1 Flash will be served by 1.2 instead.
That last change is the one to plan for. Your code will keep working, but the model behind the name changes, and so do the output and the bill. This post covers what Google shipped, what the price change means per clip, what to test before December 3 and how to pin your settings. If you are new to the model, start with our Gemini Omni Flash explainer from July. Facts are as of November 8, 2026.
The short version: Gemini Omni 1.2 Flash is cheaper per second and can make longer clips. If your site, plugin or workflow calls gemini-omni-1.1-flash, switch the model string to gemini-omni-1.2-flash now and set the clip length and resolution in every request. Then run your usual prompts and compare. Do it before December 3.
What Google released with Gemini Omni 1.2 Flash
According to Google’s Gemini API release notes, the November 5 release includes:
- Video generation and conversational editing up to 15 seconds, up from 10.
- First and last frame interpolation from 3 to 10 seconds.
- Video extension of 3 to 10 seconds per turn, up to 30 seconds in total.
- Better character and subject consistency across up to 6 reference images.
- A 30% lower video output price, now $12.00 per 1M tokens.
- Doubled rate limits on the paid tier.
Google’s Gemini Omni Flash model page lists the model code as gemini-omni-1.2-flash. It accepts text, image and video input. Output is listed as video with native audio, plus text. Clips run at 24 frames per second in 360p, 720p, 1080p or 4K.
Google’s Omni Flash developer guide adds the working details. Editing runs through the Interactions API. Each follow-up prompt builds on the last clip, and the model keeps the parts of the scene you did not mention. The guide says reference images can cover up to two characters, with up to three images each.
We did not find a Google blog post for this release. The release notes, model page, guide and pricing page are the sources here. Those pages describe the Gemini API. They do not say whether the Gemini app or Google Flow now runs 1.2.
What happens to Omni 1.1 Flash on December 3
The release notes say that from December 3, 2026, API requests to gemini-omni-1.1-flash "will automatically route to gemini-omni-1.2-flash at Gemini Omni 1.2 Flash rates." The developer guide calls this a four-week transition window and says no code changes are required.
Google’s Gemini deprecations page lists December 3 as the shutdown date for 1.1, with 1.2 as the replacement. The older gemini-omni-flash-preview was already shut down on October 22. We covered that date in our post on Gemini shutdown dates this fall.
"No code changes required" means your requests will not fail. It does not mean your results stay the same. After December 3, the name gemini-omni-1.1-flash points at a different model. You cannot keep 1.1 by pinning its name, because the name itself moves.
Google has done this before with text models. In October, requests to Gemini 3.7 Flash began routing to 3.8 Flash. Our Gemini 3.8 Flash cost-per-task post showed how a same-price swap can still change what a task costs.
Gemini Omni 1.2 Flash pricing, in one table
All prices below come from Google’s Gemini API pricing page as of November 8, 2026. Both models are paid tier only. Neither has a free tier. Google bills video by output tokens, at 5,792 tokens per second of 720p video.
| Item (Standard, paid tier) | Omni 1.1 Flash | Omni 1.2 Flash |
|---|---|---|
| Input, per 1M tokens (text, image, video) | $1.50 | $1.50 |
| Text output, per 1M tokens | $9.00 | $9.00 |
| Video output, per 1M tokens | $17.50 | $12.00 |
| Effective price per second, 720p | About $0.10 | About $0.07 |
| 10-second 720p clip | About $1.01 (our math) | About $0.70 |
| 15-second 720p clip | Not available (10-second limit) | About $1.04 |
| Per second at 360p / 1080p / 4K | Not listed | $0.023 / $0.104 / $0.209 |
The per-second price drops by about a third. The per-clip price depends on length. A 15-second clip on 1.2 costs about the same as a 10-second clip did on 1.1. Upscaled 1080p and 4K cost more per second than 720p. If you budget per clip, check which length and resolution your workflow actually asks for.
Why the reroute can change output and cost
Three changes matter most when a 1.1 request lands on Gemini Omni 1.2 Flash.
Clip length. The developer guide says you can set duration in response_format to any whole second from 3 to 15. If you leave it out, the model defaults to 10 seconds. If you set it, the model uses your value "instead of any length mentioned in your text prompt." So a prompt that asks for a longer clip in plain words may now get one, and you pay per second.
Audio. The model page lists 1.2 output as video with native audio and text. It lists 1.1 output as video only. If your pipeline expects silent clips, or adds its own music and voiceover, test what comes back.
Look and consistency. Google says 1.2 improves character and subject consistency. That is good news, but a recurring brand character or mascot may still look different. Clients who approved a look on 1.1 should see 1.2 output before it goes live.
What to test on Gemini Omni 1.2 Flash before December 3
Work through this list on a staging site or a test project, with a spending cap set.
-
Find every call. Search your code, plugins, automations and saved prompts for
gemini-omni-1.1-flash. Agencies should check each client project. -
Switch the model string now. Change it to
gemini-omni-1.2-flashand set the date in your change log. Then you choose when the switch happens. -
Set duration and resolution in every request. Do not rely on defaults or on words in the prompt. This is the closest thing to pinning you have.
-
Run your standard prompts side by side. Use the same text and reference images on both models while 1.1 still exists. Compare length, audio, motion and how faces and logos hold up.
-
Check multi-turn edits. If staff edit clips over several turns, repeat a typical session. Note where the model keeps or drops details.
-
Recheck reference images. The limit is now 6 images. If you send more or fewer, confirm how the model uses them.
-
Update cost estimates. Recalculate per-clip cost at your real length and resolution. Set or update the project spend cap in AI Studio.
-
Review data storage. Omni runs on the Interactions API, which stores requests by default. Our post on Gemini API logging defaults covers what is kept and how to turn it off.
Our SDK versions post covers pinning on the SDK side. The same habit applies here. Pin what you can, and write down what you cannot.
Who should care: agencies and nonprofits making short video
Most of our readers use Omni Flash for short social clips, event teasers, explainer loops and animated still images. For that work, Gemini Omni 1.2 Flash can fit a 15-second social clip in one generation. Before, you needed an extension turn.
For nonprofits, the lower price helps if you make many short variants for testing. For agencies, the risk is a silent change in a client deliverable. A scheduled automation that builds weekly clips will keep running on December 3. Its output may not match what the client approved.
If you also make still images, check our Nano Banana explainer. Google’s deprecations page now lists gemini-omni-1.2-flash as the recommended replacement for the Veo 3.1 preview API models, which shut down on October 22. If you were on those Veo previews, Omni is where Google points you.
One more reroute: Gemini Embedding 2 preview
The same release notes include a smaller change. Since November 6, 2026, requests to gemini-embedding-2-preview route to gemini-embedding-2 at Gemini Embedding 2 rates. Google says the GA model "uses the same API schema, pricing, and embedding space," so no re-indexing of stored vectors is needed. Update the model string anyway, so your code says what it runs.
How other platforms are handling video models
Video and image models are changing at other platforms too. OpenAI’s API deprecations page shows the Videos API and Sora 2 models removed on September 24, 2026, with no replacement listed. We noted that date in our Assistants API shutdown post. The same page lists older GPT Image models ending December 1, 2026, and states six months’ minimum notice for generally available models. Google gave Omni 1.1 four weeks, though requests reroute rather than fail. Outside the big labs, we covered ByteDance’s Seedance 2.5 and Black Forest Labs’ FLUX 3 in July. For choosing among Google’s current models, see which Gemini model to use.
Frequently Asked Questions
What is Gemini Omni 1.2 Flash?
It is Google’s video generation and editing model in the Gemini API, made generally available on November 5, 2026. Google’s release notes say it makes and edits clips up to 15 seconds, extends clips up to 30 seconds in total and keeps characters consistent across up to 6 reference images.
What happens to Omni 1.1 Flash on December 3, 2026?
Google says API requests to gemini-omni-1.1-flash will route to gemini-omni-1.2-flash and be billed at 1.2 rates. Calls will not fail. The model behind the name changes, so output and cost can change.
How much does Gemini Omni 1.2 Flash cost?
As of November 8, 2026, Google lists video output at $12.00 per 1M tokens, down from $17.50 for 1.1. Google puts that at about $0.07 per second of 720p video, about $0.70 for a 10-second clip and about $1.04 for a 15-second clip. There is no free tier.
Is the 30% price cut a 30% cut per clip?
Only if your clips stay the same length and resolution. The per-second price drops by about a third. A 15-second clip on 1.2 costs about the same as a 10-second clip did on 1.1, and 1080p or 4K costs more per second.
How do I pin a Gemini Omni version?
Use the full model string gemini-omni-1.2-flash and set duration and resolution in response_format on every request. After December 3 you cannot stay on 1.1, because its name will serve 1.2.
Does Gemini Omni 1.2 Flash add audio to clips?
Google’s model page lists 1.2 output as video with native audio and text, while 1.1 is listed as video only. Test whether your pipeline handles the audio track.
What is the default clip length?
Google’s developer guide says the default is 10 seconds when you do not set duration. If you set duration, the model uses it instead of any length you mention in the prompt.
Do I need to re-index after the Gemini Embedding 2 change?
Google says no. Since November 6, 2026, gemini-embedding-2-preview routes to gemini-embedding-2, which uses the same schema, pricing and embedding space. Update the model string so your code matches what runs.
Is Gemini Omni 1.2 Flash in the Gemini app or Google Flow?
The Google pages we read for this post describe the Gemini API only. They do not say whether the Gemini app or Flow uses 1.2. Check Google’s product pages before you assume it.
