Seperation the voice to 2 track/person by voicesep, keep the office effect in Modelscope demo meanwhile accelerate the resample speed. 保证了voicesep的modelscope上的demo的效果的同时，加速运行时间。
https://github.com/GuardSkill/ComfyUI-VoiceSep
git clone https://github.com/GuardSkill/ComfyUI-VoiceSep
comfy node install voicesep
各种节点：支持 Lora Block Weight、区域节点、后端缓存、提示词工具、列表工具、噪声（种子）工具……与 Impact Pack。
为 https://github.com/Eyeline-Research/Go-with-the-Flow 从视频生成噪声的节点。