intabai / video gen

github
models (cached in browser)
show / hide model cache Models download once into your browser and persist across visits. Nothing leaves your machine.
prompt
input image (optional, for image-to-video) leave empty for text-to-video. otherwise, the image is center-cropped to a square (largest centered square that fits) and scaled to the model's resolution; it becomes the first frame of the generated clip.
options
output is fixed at 480×480, 81 frames @ 16 fps, 4 UniPC steps - the model only supports this geometry.
advanced
- leave blank for a fresh random seed each run.
- show a verbose per-stage log (timings, token counts, per-step progress) under the progress bar. noisy; useful when something looks wrong.