Google: Gemini Omni 1.2 Flash API
Gemini Omni 1.2 Flash is Google's model for video generation and conversational editing: it turns text, images and video into video with native audio up to 15 seconds, refines and extends it over multiple turns in natural language, and supports first and last frames, up to 6 reference images and up to 4K output.
- Context window: 131,072 tokens
- Max output: 65,536 tokens
- Input: text, image, video
- Output: video
- Reasoning: Supported
- File input: Supported
- Released: 2026-11-05
Frequently Asked Questions
Why does submission say my balance is insufficient when it covers one video?
On submission the gateway reserves an estimate from resolution × duration: 10 seconds without `duration`, and the 15-second maximum for edits. On completion you pay the actual upstream usage and the rest is returned. Setting `duration` makes the reservation match the real cost.
How do I keep editing the previous video?
With the unified task API, put the earlier task ID in `input.previous_interaction_id` and describe the change; to extend instead, add `"task": "extend"` and `duration`. With the native API, use the previous response's `id` as `previous_interaction_id`. No need to upload the video again; the request goes back to the upstream account that made it.
What is the context window of Gemini Omni 1.2 Flash?
Gemini Omni 1.2 Flash supports a context window of up to 131,072 tokens.
Does Gemini Omni 1.2 Flash support reasoning?
Yes. Gemini Omni 1.2 Flash is a reasoning-capable model.