Gemini Omni 1.2 Flash makes 15-second video clips at $0.07 per second

A team that needs a 12-second product clip could not get one from Gemini Omni 1.1 Flash in one pass. The old model stopped at 10 seconds. The team had to stitch two clips and watch for the join. Google released Gemini Omni 1.2 Flash this week. The new model does up to 15 seconds in one pass. The price is $0.07 per second for 720p video. Google says that is 30 percent cheaper than Omni 1.1 Flash.

The blog post is by Anish Nangia, a product manager at Google DeepMind. The model is available in the Gemini API through Google AI Studio and the Google Cloud console. It also runs inside Google's own video tools: Flow, Google Vids, Google Ads, and the Gemini app for Google AI Ultra subscribers.

What is new

The clip length moved from 10 seconds to 15 seconds. Clip length can be set in one-second steps, from 3 seconds to 15. Google says this gives room to write a full social ad spot in one generation. The old model needed two or more.

Google also claims better reference following. The team names three cases. A character stays the same across shots. A product keeps its shape when the camera moves. A visual style stays consistent across transitions. The post also says dialogue is more fluid, with clearer lip sync and better handling of more than one speaker.

A new editing mode lets a user ask for changes to a clip in several turns, inside the same session. The reader can keep the existing clip and change one part of it.

What Google did not publish

The post has no benchmark scores. The claims about consistency, lip sync and multi-speaker audio are described in prose, not in numbers. There is no named customer and no independent test. The example uses are Google's own: virtual try-on sequences, catalog spins, branded video ads, and multi-character dialogue scenes.

The 30 percent figure is Google's own, comparing its own current and previous version. Google does not publish Omni 1.1 Flash's price in the same post. The figure cannot be checked against two posted prices inside this release. Omni 1.1 Flash's price can be read from Google's model documentation.

What it means for a team buying video generation

The price is the part that changes who can test this. At $0.07 per second, a one-minute test clip costs $4.20 in model calls. At 15 seconds in one pass, a team can produce a full short ad without stitching. A team that wrote off AI video on the old clip length can pick it up again. A team that could afford the old price but not at scale can now run more variations before picking one.

Two steps help a team adopting this. Run the same prompt side by side on Omni 1.1 and 1.2 before migrating any production prompt. Also measure the editing mode on a task that used to need two generations. That is where the saving is biggest.

Source

• Primary source: Google: Introducing Gemini Omni 1.2 Flash, by Anish Nangia, Google DeepMind.