Gemini Omni Flash

Google's multimodal high-speed video model — generation and editing with native audio. The only model in NGMC that can take an existing video and transform it with a prompt.

Capabilities

Feature	Support
Text-to-Video	Yes
Image-to-Video	Yes (first frame)
Reference-to-Video	Yes (up to 30 images)
Video Editing	Yes (1 input video, up to 10s)
Resolution	720p
Aspect Ratios	16:9, 9:16
Duration	Up to 10s
Frame Rate	24 fps
Native Audio	Yes
Seed / Negative Prompt	No

Video Editing Mode

Connect a video (≤10 seconds) as the input and describe the transformation:

Keep the actor's performance but move the scene to a neon-lit alley
at night, rain falling.

Omni Flash re-renders the clip following your instruction — restyling, relighting, swapping environments — while preserving the motion of the source. No other model here edits existing footage.

30 Reference Images

The largest image-reference budget of any video model (Seedance takes 9, Veo 3). Use it for dense style bibles: character sheet + wardrobe + props + environment + color script in one run.

Fast, Short, Sound-On

Omni Flash trades resolution ceiling and duration for speed and completeness: ~1 minute to a 720p clip with a synchronized audio track. It's the quickest way to a watchable, audible shot — ideal for animatics and previz.

Prompting Tips

In edit mode, say what to keep. "Keep the camera move and timing, change everything else to watercolor" beats listing every change.
Describe the soundscape. Audio is native — "distant thunder, dripping water" shapes the track.
Storyboard at Omni speed, deliver on Veo/Seedance. Same prompt language transfers up-market for the 1080p/4K final.

Limitations

720p only; 16:9 and 9:16 only
10-second ceiling on both output and input video
One input video at a time in edit mode
No seed, no negative prompt

Capabilities​

Video Editing Mode​

30 Reference Images​

Fast, Short, Sound-On​

Prompting Tips​

Limitations​

See Also​