Fast high quality video generation.Update (17/07): now with the new v0.9.8 for improved prompt understanding and detail generation
Model GitHub Diffusers
Number of initial frames to use for conditioning/transformation. Must be N*8+1.
Target video duration (0.3s to 8.5s)
Uses a two-pass generation for better quality, but is slower. Recommended for final output.
Controls how much the prompt influences the output. Higher values = stronger influence.
Must be divisible by 32.