V100 + FasterLivePortrait + FlashHead Deployment Recipe

This V100 deployment recipe is currently maintained in Chinese:

Read the Chinese guide.

It covers a single NVIDIA V100 32GB host running OpenTalking with:

• FasterLivePortrait through OmniRT for real-person video-driven rendering.
• FlashHead through a WebSocket bridge for image-conditioned generation.
• V100-specific notes for FP16, TensorRT 8.6, disabled torch.compile, CUDA libraries, and WebRTC port exposure.