// free tool · runs in your browser · no signup
Veo 3.1 Prompt Checker
check before you spend credits
Paste a video prompt, see what would waste the generation, copy the fixed version. Also for Gemini Omni Flash.
Your prompt has:
6 checks, instantly
Paste a prompt or
Veo 3.1 vs Gemini Omni Flash: where the guides differ
In Flow you can pick either model, and a prompt written for one is not always right for the other. The most important difference is negation. From the Gemini API and Google Cloud guides, checked on 2026-10-25:
| Veo 3.1 | Gemini Omni Flash | |
|---|---|---|
| Things to avoid | List the terms, no "no …" (API: negativePrompt field) | Write it in the prompt: "No dialogue" |
| Clip length | 4, 6 or 8 s (setting) | 3 to 10 s, timecodes like [0-3s] allowed |
| Aspect / resolution | 16:9 or 9:16 · 720p, 1080p, 4K | 16:9 or 9:16 · 720p default |
| Audio | Always generated: describe it | Generated by default: direct it in the prompt |
Note: the Veo 3.1 preview models left the Gemini API on 2026-10-22 (Google points API users to Gemini Omni Flash or GA Veo on its enterprise platform). Veo 3.1 is still offered in Flow.
Frequently asked questions
What does this Veo prompt checker check?
Six things that commonly waste a generation: camera instructions that contradict each other (a static shot that also dollies in), "no …" phrasing inside the prompt, dialogue without a speaker or too long for the clip, length/resolution/aspect ratio written in the text instead of the settings, too many actions or camera moves for one clip, and prompts too long for a short clip. Each finding cites the Google guide or the community source it comes from, with the date we last checked it.
Veo 3.1 or Gemini Omni Flash: which rules change?
The guides disagree on a few points. Veo's guide recommends against "no walls"-style negation and asks you to list unwanted terms (in the API, in the negativePrompt field). Gemini Omni has no negative prompt parameter, and its guide tells you to write negatives in the prompt itself ("No dialogue"). Veo clips are 4, 6 or 8 seconds; Omni clips run 3 to 10 seconds and the Omni guide allows timecodes like [0-3s] to time each beat. Pick the model at the top and the checker applies the right set.
Does it work for prompts I use in Flow or the Gemini app?
Yes. Choose "Flow / Gemini app" as the surface. Those apps have no negative prompt field, so instead of moving terms to a field, the checker rewrites common negations into what should appear (for example "no camera movement" becomes "static locked-off tripod shot").
Will this guarantee a good video?
No. It checks the prompt against what the guides document and against failures creators report often. It does not render anything or predict quality. What it does is remove the known reasons a generation fails before you spend credits on it.
Is it free? Do you store my prompt?
Free, no signup, no limit. The check runs entirely in your browser: the prompt is never sent to a server.
What about Sora?
OpenAI shut down the Sora app in April 2026 and removed Sora from its API in September 2026, so the checker covers Google's current video models: Veo 3.1 and Gemini Omni Flash.