Google Ships Gemini Omni 1.2 Flash With 15-Second Video and a 30 Percent Price Cut
Google has released Gemini Omni 1.2 Flash, a new version of its video generation model. The model natively creates clips of up to 15 seconds, up from 10 seconds in Omni 1.1, and the duration of each clip can be selected in 1-second increments anywhere from 3 to 15 seconds, so the same model covers both very short loops and longer scenes. The release is generally available now.
📅 November 5, 2026 • blog.google
Pricing has been reduced by 30 percent versus Omni 1.1 and now stands at $0.07 per second of 720p video. Every output carries a SynthID watermark that is described as imperceptible to viewers, which keeps generated clips identifiable as synthetic by detection tools while leaving their appearance unchanged for the audience.
Access is open now on the Gemini API and on Google Cloud, and rollout is under way in Google Vids, Google Ads and the Gemini app, where it covers AI Ultra users. Omni 1.2 Flash supports multi-turn conversational editing, so a clip is refined through follow-up instructions instead of being regenerated from the start, and finished videos can be upscaled to 1080p and 4K.
In a typical production flow a draft clip is generated at 720p, adjusted in conversation until it matches the brief, and then upscaled for delivery, with the cost of each run computed as clip length multiplied by the per-second rate. The always-on watermark applies the same provenance layer to every asset, whether it comes from an API script, a Vids project or an ad campaign, and because one model serves the API, Vids, Ads and the Gemini app, a single generation stack sits behind different surfaces.
IMPACT: Short-form video generation now behaves like a metered cloud resource: duration, price and provenance are explicit numbers, and editing continues in the same conversation instead of moving to a separate tool. For teams that publish video at volume, the practical change is that generated clips can pass through the same review pipeline as recorded footage, with watermark detection available as a standard check on every item that enters the library.
Sources: blog.google
Continue the conversation: https://t.me/AIAutomationArchitecture · Bluesky
Top comments (0)