Video Generation in ChatGPT: Limits, Settings, and Rules

ChatGPT's video generation capabilities have clear technical limitations. A resolution of 1024x1024 is unavailable, and 883x883 supports only two frames.

Optimal quality is achieved with a format of 16 frames at 308x308 resolution. Increasing the duration to 64, 128, or 256 frames leads to image degradation and artifact appearance.

Technical nuances of generation

• The API documentation limits the count to 64 frames, although the interface allows selecting more.
• Using reference images significantly reduces noise levels during generation.
• Editing is available only for previously generated files with saved metadata.

Face usage policy

A total ban on generating faces of real people has been lifted. OpenAI permits creating images of public figures by applying targeted protective restrictions instead of a complete block.

Data usage terms

According to Service Terms, visual capabilities accept media as input data. It is prohibited to use them for identity verification, extracting private information, or reproducing appearances without consent.