Launched this week
Google Omni 1.2 Flash
Gemini Omni 1.2 Flash: 15s AI video at $0.07 per second
109 followers
Gemini Omni 1.2 Flash: 15s AI video at $0.07 per second
109 followers
Gemini Omni 1.2 Flash is Google DeepMind’s video generation and editing model, now generally available. Per Google, clips run up to 15 seconds, up from 10, and 720p costs $0.07 per second, 30% less than Omni 1.1. New: multi-turn editing without regenerating, steadier characters and styles, clearer lip-sync and smoother camera moves. Draft at 360p or 720p, then upscale to 1080p or 4K. In the Gemini API, Google Cloud, Flow and Vids.
Free Options
Launch Team
LoomHit record, hand off to AI
Promoted
Gemini Omni 1.2 Flash is the latest update to Google DeepMind’s video generation and editing model, and Google says it’s generally available now. It makes longer clips for less money, and it lets you keep editing the same clip.
What changed from Omni 1.1:
Clips run up to 15 seconds, up from 10, and you can set the length from 3 to 15 seconds
The price is down 30%, to $0.07 per second for 720p video
Multi-turn editing: keep refining the same clip, like swapping a product or background, changing the lighting or weather, or switching the aspect ratio, without regenerating it
Characters, products and styles hold from start to finish, with less drift or morphing during motion and camera moves
More natural dialogue, clearer lip-sync, multi-speaker voices, and continuous camera moves like FPV flythroughs, zoom-outs and orbits
It’s built around a draft-then-upscale workflow: draft at 360p or 720p, make your edits, then upscale to 1080p or 4K. Inputs include character, product and style images, short motion clips, audio cues, first and last keyframes and voice references. Google lists uses like virtual try-on, catalog spins and branded video ads from product photography. Outputs carry an imperceptible SynthID watermark.
Where to use it: the Gemini API in Google AI Studio and Vertex AI on Google Cloud (model ID gemini-omni-1.2-flash), plus Flow, Google Vids (15-second clips at no extra cost with a Google account or Workspace subscription) and Google Ads’ Asset Studio. The Gemini app is rolling out to Google AI Ultra subscribers.
Good fit if you make ads, product videos or short clips and want to refine one video over several rounds instead of starting over each time.
The multi-turn editing is the part I'm most curious about — I record a lot of short Mac app demos, and the thing that usually breaks them is the UI drifting or on-screen text morphing between edits. Does refining the same 15s clip over a few rounds actually hold text and interface details steady, or does it start to drift? The draft-then-upscale workflow at $0.07/sec makes trying a few takes cheap enough to find out.
Editing a clip across several turns without regenerating it is the real shift here, because the usual loss was a good take ruined by one wrong detail. Drafting at low resolution and upscaling later also fits how people actually iterate. How well does a character or product hold its identity after four or five edit turns, or does drift build up the way it does with repeated image edits?