Google Omni Flash generates clips at four fixed lengths: 4, 6, 8, and 10 seconds. Google's API accepts a 3–10 second range, but the UI exposes only the four duration buttons, and some app surfaces default or lock to 10 seconds. There is no native extension for Omni-generated clips, so 10 seconds is a hard ceiling.
Pick from four fixed durations — 4, 6, 8, or 10 seconds — when generating with Google Omni Flash. At the API level Google accepts 3–10 seconds, but in practice the interface gives you the four preset buttons, and some workflows default or lock to the 10-second maximum, so plan your shots in those increments rather than assuming arbitrary lengths.
For dialogue shots, don't automatically pick the 10-second option. Across 30+ test outputs, 6–7 seconds is the consistent lip sync ceiling with a single speaker in frame — past that point, sync quality degrades. Use the 6-second preset for talking shots and reserve 8- and 10-second generations for action, environment, or motion-graphics clips where lip sync isn't a factor.
Clip length also interacts with resolution cost. Omni Flash outputs 720p by default, upscales to 1080p at no cost, and charges the equivalent of a full generation for 4K — so a 10-second 4K clip effectively costs double a 10-second 1080p one. If you're generating long clips for a big-screen deliverable, budget the 4K pass per clip.
The 10-second cap is firm because scene extension currently works only on clips generated in Veo 3.1, not on Omni-generated content. To go longer than 10 seconds you either cut multiple Omni clips together in the edit, or generate on a model with a longer native duration — Veo 3.1 tops out around 8 seconds per generation, while Seedance 2.0 reaches 15 seconds natively. Inside invideo you have all of these models in one place, and the invideo agent routes each shot to whichever model's duration ceiling fits it, so a 12-second continuous shot goes to Seedance 2.0 while a 6-second dialogue clip stays on Omni Flash.
Watch some of these to see what works for you:
In our research we found that 6 or 7 seconds is kind of the ceiling of where the model is going to perform consistently well.
— invideo's creative team, after testing 30+ Omni Flash outputs