Google recently introduced Gemini Omni Flash, which is a video generation model that also supports editing. However, some users are facing prompt rejection issues.
Reports on Reddit and posts on X claim that almost all prompts “violate policy.” Gemini Omni Flash throws an error that says, “I can’t generate that video. Try describing another idea…” and then asks the user to review video policy guidelines. Various replies to the posts linked above also confirm getting pointless rejections.

This doesn’t make sense, since simple requests such as “Cat walks between rabbits,” or “Make this cat wear a white hat,” or “Change the background to night” should not actually violate any policies. This automatic flagging of basic prompts for no good reason is likely a false positive on Google’s end.

Gemini Omni was introduced recently at Google I/O, and it’s supposed to combine images, audio, video, and text as input to generate high-quality videos. It was advertised as a tool to conveniently edit your videos. However, most prompts with videos as a reference seemingly fail; editing videos is completely broken.
Thankfully, a VP at Google (Josh Woodward) has noticed posts about this issue, and through a reply on X, stated that this shouldn’t be happening. Josh also said that Google would investigate and requested the underlying videos from the original poster.

“b/515000564” was mentioned, which is Google’s internal bug tracker reference number. Users have asked for a public issue tracker link instead, which would be more helpful.
I wasn’t able to find a reliable workaround, so users will have to wait until Google resolves the issue.
In other Google news, the company has started rolling out the “Ask Gemini” button within the Google app, but users want to turn it off already. You can read about that here.