/video2video

Automatic1111 Stable Diffusion WebUI Video Extension

Primary LanguagePython

Changes made in this fork:

Support for interrogator

Add audio back

Improved strength arrangement

Disable latent memory on scene switch, based on MAE

Support for ControlNet:

How to use: When generating, enable the ControlNet tab (Make sure you have ControlNet extension installed), and select your settings Then check the "Using ControlNet" tab in the script settings Generate and it will use ControlNet

Automatic1111 Stable Diffusion WebUI Video2Video Extension

Pluging for img2img video processing

  • No more image files on hard disk.
  • Video fps can be set as original, or changed.
  • Now with latent space temporal blending.

Result saved to output folder img2img-video as MP4 file in H264 encoding.

Added optional temporal blending for latent space. Applied per each step between previous and current frame.

Need a FFmpeg. For OS Windows implemented automatic installation of FFmpeg.

Under development, bugs applied.

Dependencies

ffmpeg

skvideo (pip install sk-video)

TODO

Bug: latent blending will work right only for batch_size = 1