Mentatcurated
Artificial Intelligence high · first-party

Gemini Omni Flash

Six weeks after its consumer debut, Google's video model opened to developers at ten cents a second — and shipped beside a new image model that makes a picture for three cents.

When Gemini Omni landed in YouTube and the Gemini app in May, Google deferred the developer API and the benchmarks to "later." Later is now: on June 30 the model opened to anyone on the Gemini API and AI Studio, priced at $0.10 per second of 720p video with synced audio. A ten-second clip costs about a dollar — the same price point as Google's own Veo Fast.

Character consistency when changing scenes or panning movements has some limitations but we are working to make this better. — Google

The quieter half of the launch is where the unit economics land. Alongside it, Google shipped Nano Banana 2 Lite, a cheaper, faster image model that renders a picture for 3.4 cents in about four seconds. Because both sit in the same API, a developer can generate a still for pennies and animate it for roughly a dollar in one programmatic session — media generation at the unit cost of a display ad, which is exactly the automated ad-and-marketing-creative pipeline the pricing is aimed at.

Worth keeping the frame honest, though: this is an access-and-price story, not a capability jump. The API launches without the audio-reference upload the consumer version implied, and Google's own copy still concedes that character consistency "has some limitations" when scenes change or the camera pans — the same drift-on-edit flaw the May launch was flagged for, unfixed at the developer opening. The much-repeated "#1 video, Elo 1,527" stat is leaderboard-shopped, too: the same model scores 1,292 on one public arena and 1,404 on another. What's genuinely new here isn't a better clip — it's that runtime video now has a metered API price, and a three-cent image sitting next to it.

The lenses

Novelty 2
Impact · breadth 4
Impact · depth 3
Actionable 3
Substance 4
Hype 3

The facts

What it isText/image-to-video and cheap image generation, both callable from the Gemini API and AI Studio
What it costs$0.10 per second of 720p video with sound (~$1 for 10 seconds); 3.4 cents per image on the new Lite model
AccessPublic preview on the Gemini API, AI Studio, and Vertex — a paid tier, not a free click
Known limitNo audio-reference upload in the API yet; character consistency still breaks on scene changes and panning
Open blog.google →

How this connects

Tap a node to open it