* add training code and corresponding config yaml files
* add some auxiliary funciton in utils
* fix a parameter bug in motion module
* fix mask size issue in stage2 dataset module "talk_video.py"
* WebUI + Audio Fix
1. audio fix: explicitly specify the audio codec in `util.py`, otherwise the video is technically corrupt and doesn't play sound
2. web ui: gradio web ui
3. print the current step while running inference
gradio
* lint
* update
xumingw