Generating videos

Video generation works like image generation, with a few extra options depending on the model: duration, resolution, start/end frames, and reference inputs (images, video, or audio on models that support omni references).

Tip: models differ significantly in strengths - lip-sync quality, motion, prompt adherence, and supported languages vary by model. If a result disappoints, try the same prompt on a different model before rewriting it.

Updated on: 07/07/2026