VideoGenerationModel

A loaded text-to-video model.

generate

generate(self, prompt: 'str', **options: 'Any') -> 'Any'

Generate a clip; returns an object with frames ([F, H, W, 3]), fps and seed. Options: first_frame / last_frame ([H, W, 3] UInt8 tensors the clip is anchored to, on families that condition on keyframes), negative_prompt, width, height, num_frames, fps, num_inference_steps, guidance_scale, seed, extra.