# Gemini Omni 1.2 Flash is GA: 15-second video, lower pricing and API limits

Google launches Gemini Omni 1.2 Flash with 15-second video, cheaper output and higher paid limits. What developers need to know before the December 3 migration.

By Paras Madan | 2026-11-05


Google has released Gemini Omni 1.2 Flash, making its updated video generation and editing model generally available through the Gemini API and Gemini for business. The practical changes are longer clips, a lower output price and more room to iterate on a video without starting again.

The [November 5 launch announcement](https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemini-omni-1-2-flash/) also describes a rollout across Flow, Google Vids, Google Ads and the Gemini app. The Gemini app rollout is for Google AI Ultra subscribers. Product access and API billing are separate, so a consumer subscription should not be treated as an API allowance.

## Fifteen seconds is not the limit for every operation

The new model identifier is gemini-omni-1.2-flash. According to the [developer guide](https://aistudio.google.com/docs/omni), generation accepts whole-second durations from 3 to 15 seconds, with 10 seconds as the default. It produces video with native audio and supports conversational editing through the Interactions API.

Other operations have different constraints:

- First-and-last-frame interpolation supports 3 to 10 seconds, not 15.
- Extension adds 3 to 10 seconds per turn, up to 30 seconds total.
- Up to six reference images are supported. For an 11-to-15-second generation, the reference-video allowance drops to one clip of up to three seconds.
- Extension uses the final 10 seconds as context and appends to the end. It cannot insert a new segment into the middle of a clip.

For a product demo, that makes a useful distinction between producing a longer first shot and building a sequence across several turns. Neither workflow guarantees that packaging text, character identity or physical motion will remain correct. Those still need review in the final output.

## What the lower price means

Google's [pricing page](https://aistudio.google.com/docs/pricing#gemini-omni-1.2-flash) lists video output at $12 per million tokens. At 720p, the documented rate is 5,792 tokens per second, approximately $0.07 per second. That puts a 10-second clip at roughly $0.70 and a 15-second clip at $1.04 for video output.

The page also gives approximate output rates of $0.023 per second at 360p, $0.104 at upsampled 1080p and $0.209 at upsampled 4K. The higher-resolution options should be described as upsampling, not native 4K generation. The Gemini API model is on the paid tier.

Google describes the output-price reduction from Omni 1.1 as 30%. That does not mean every finished asset costs 30% less. A production budget still depends on how many candidates are generated, how often they are edited and which resolution is delivered. Track cost per accepted clip, rather than treating one successful generation as the entire workflow.

## Higher limits still require a queue

The [published standard limits](https://aistudio.google.com/docs/rate-limits#gemini-omni-flash-rate-limits) are 8 requests per minute and 40 per day for Tier 1, 16 per minute and 200 per day for Tier 2, and 40 per minute and 2,000 per day for Tier 3. Google says the paid-tier limits have doubled.

Quotas apply per project, not per API key, and actual capacity can vary. Developers should check the active allowance in AI Studio before planning a batch of hundreds of assets. Extra API keys are not extra project quota.

There are regional and platform differences too. The Developer API guide does not support editing or extending uploaded videos in the EEA, Switzerland or the UK, although model-generated videos are supported. It also lists audio-reference uploads and Provisioned Throughput as unsupported. Google's broader launch mentions enterprise Provisioned Throughput, so teams should verify the capability on their chosen platform rather than transfer a Cloud promise to the Developer API.

## Existing Omni 1.1 integrations have a deadline

The [release notes](https://aistudio.google.com/docs/changelog) say requests to gemini-omni-1.1-flash will automatically route to 1.2 on December 3, 2026, at the newer model's rates. No code change is required for that routing change.

That removes a migration step, not the need for testing. Run representative prompts and edits before the switch, compare output quality and moderation behavior, and check that downstream systems accept the resulting media. A lower price is useful; predictable behavior across an existing content pipeline is what makes the upgrade deployable.

I'm Paras. I work at the intersection of AI, engineering and distribution.

I help technical founders turn AI product changes into clear stories, useful demos and practical adoption decisions.

[Book a 30-minute call](https://cal.com/paras-madan/30min).