Hello Google AI Support,
I am requesting a review of what appears to be a false-positive safety block in the Gemini API.
I am using the gemini-omni-flash-preview model to animate an uploaded image of a fictional, AI-generated presenter. The character is not a real person, public figure, or private individual, and the project does not involve impersonation, deception, sexual content, violence, political persuasion, or any other harmful use.
The purpose is benign animation research. I am building a reusable viseme library for an AI presenter and testing simple mouth movements for the sounds M, B, and P while keeping the camera, background, clothing, and pose stable.
The request was blocked twice with this error:
Error code: 400
Request blocked due to prohibited content guidelines.
Unable to show the generated video. The video was filtered out because it violated Google’s Generative AI Prohibited Use policy.
I also tried a revised prompt that explicitly stated:
- the character is fictional and AI-generated
- she is an adult
- she does not represent or imitate any real person
- the project is for benign facial-animation and viseme research
- the requested movement is limited to neutral lip and jaw motion
The request was still blocked.
Could you please review whether the uploaded image or the generated output is being incorrectly classified? I would also appreciate clarification about whether photorealistic fictional adult characters are supported for image-to-video generation through this API.
Technical details:
Model: gemini-omni-flash-preview
Task: image_to_video
Input type: uploaded PNG image plus text prompt
Output requested: 16:9 video
Use case: fictional presenter viseme and facial-animation research
I am not asking for any safety restriction to be bypassed. I am asking for confirmation about whether this is a false positive, an unsupported use case, or a limitation specific to the current preview model.
Thank you for reviewing this issue.