Veo 3.1 vs Gemini Omni Flash: what’s different and who should use which

Where each model is available now

Much of the “Veo 3.1 vs Omni” question is really about where you can still use each one. Google has moved its consumer products and the Gemini API to Omni, and kept Veo 3.1 on Google Cloud.

Google also said Omni 1.1 Flash requests will route to 1.2 Flash, billed at the lower 1.2 rate, from December 3, 2026. If you used Veo on the Gemini API, our Gemini API shutdown guide walks through the migration paths, and the Flow retirement guide covers Flow users.

Specs side by side

Veo figures come from Google Cloud’s Veo 3.1 documentation. Omni figures come from Google’s Gemini API Omni documentation and price page. Neither company page publishes quality benchmarks against the other model, and we haven’t run Omni ourselves, so this page compares documented capabilities and prices only.

Price per second, and per clip in Flow

Omni Flash is billed per output token on the Gemini API; the ≈ per-second figures are Google’s own approximations. Veo figures are Google Cloud list prices with audio, billed per “count”, which we read as one second. Our column is what the same Veo clip costs through imaginary.app, per second of an 8-second clip, and it is higher than Google’s own price.

For an 8-second 720p clip with audio, that works out to roughly $0.56 on Omni Flash and about $0.40 on Veo 3.1 Lite through Google’s APIs, against $0.64–$0.75 for Lite on imaginary.app. If price per second on Google’s own APIs is your only criterion and you can set up a Google account for API billing, go direct.

Inside Flow the order changes: Omni at 720p costs 3 credits per second, so an 8-second clip uses more credits than Veo 3.1 Lite or Fast, though far fewer than Quality.

When to choose Gemini Omni Flash

• You want one continuous clip longer than 8 seconds without stitching.
• You want to refine a video by describing changes instead of regenerating it from scratch.
• You need more references than Veo allows, including short reference videos.
• You already build on the Gemini API: Google points Gemini API users to Omni Flash, while Veo 3.1 remains on Google Cloud.
• You work in the Gemini app or Flow and are happy with 720p plus upsampling.

When to choose Veo 3.1

• You need 1080p or 4K generated at that resolution rather than upsampled.
• You want a model Google lists as generally available on Google Cloud with a published support date (Veo 3.1 and Fast; Lite is still in Preview).
• Your prompts, shot lists and pipelines are tuned for Veo, and you want them to keep behaving the same way.
• You want Veo 3.1 Fast or Quality after December 5, 2026, when Flow drops them.
• You want Veo without a Google Cloud project: that’s the case imaginary.app covers. Compare the three Veo variants on our Lite vs Fast vs Quality page.

If you are moving off the Gemini API

Google points Gemini API users to gemini-omni-1.2-flash; Veo 3.1 remains on Google Cloud, where Veo 3.1 and Fast are generally available and Lite is in Preview. Using Cloud directly needs a Cloud project, billing and a service account. Another option is a provider that already runs the Cloud models. On imaginary.app that means a veo_ API key, POST /api/generate with veo3-lite, veo3-fast or veo3-quality, and GET /api/status/:taskId to collect the result; the API docs have the full request format.