Gemini Omni Flash

Gemini Omni Flash is the Google any-input-to-video model on the paid Gemini API tier. It generates and edits video from text, image, audio, or video prompts, with keyframe interpolation, clip extension, and native audio built in.

Input costs $1.50 per 1M tokens for any modality, output text is $9.00 per 1M, and video output is $17.50 per 1M tokens at 5,792 tokens per second of 720p, an effective price of about $0.10 per second. There is no free tier.

Veo 3.1 is the cinematic text-to-video flagship focused on filmic quality, while Omni Flash is built for flexible input and fast editing loops such as restyling footage, interpolating keyframes, and extending clips from any starting modality.

Yes. Editing is a core capability alongside generation: you can restyle scenes, interpolate between keyframes, and extend clips, feeding video, images, audio, or text prompts as the starting point.

The API has no free tier for Omni models, but you can test prompts in Google AI Studio with a paid key before committing, and Google AI subscription plans include Gemini video features in the Gemini app.

It accepts text, images, video, or audio and produces video with synchronized native audio, so one model handles generation, restyling, editing, keyframe interpolation, and clip extension. This removes the need to stitch separate tools for each step of a video pipeline.

Veo 3.1 is the cinematic text to video flagship focused on filmic generation quality, while Omni Flash focuses on flexible inputs, fast editing loops, and interactive transformation. Teams often use Veo for hero shots and Omni Flash for iteration and pipeline automation, both on the Gemini 3.8 generation.

Video output currently targets 720p, billed at 5,792 tokens per second, which works out to about $0.10 per second of video. Higher resolutions may arrive on later endpoints, but 720p with synchronized audio already covers social clips, previews, and most editing workflows.

No. The model is generally available only on the paid tier of the Gemini API, so experimentation happens in Google AI Studio against a paid key. Vertex AI adds provisioned throughput for production traffic, with input priced at $1.50 per 1M tokens across any modality.

Yes. Clip extension is one of its core modes: you feed an existing clip and the model continues the motion and audio natively, which is useful for building longer b roll or stretching a 3 second shot into a fuller sequence. Keyframe interpolation works the same way between two stills or two clips.