fix: correct facial alignment issues and add API server

fix: convert all audio to WAV 16kHz PCM before processing (#379)

fix: ensure upper bond does not go below zero in landmark extraction (#329)

feat: windows infer & gradio (#312)

Update audio_processor.py

feat: data preprocessing and training (#294)

fix: floor (#293)

feat: real-time infer (#286)

fix: infer bug

v1.5

Add import copy to blending.py

improve blending speed

<enhance>: support using float16 in inference to speed up

Add codes for real time inference

Add gradio

fix some cuda related issue when run on M-Series Mac

<enhance>: modified inference codes

initial_commit