v2.0

soxoj/socid-extractorv2.0Mar 31, 2026by lipku

AI Summary

Major architectural refactoring to a plugin-based system with RTMP streaming support and integration of Alibaba Cloud QwenTTS.

Key Highlights

  • Refactored to a plugin-based architecture for models, TTS, and transmission
  • Added RTMP output support in streamout directory
  • Integrated common logic into BaseAvatar and BaseAsr
  • Added Alibaba Cloud QwenTTS

New Features

  • Refactor code to move digital human models and audio features to avatars directory
  • Change digital human models, TTS, and transmission methods to plugin approach
  • Organize transmission methods into a separate class in streamout directory, add RTMP output
  • Integrate common inference and reply code into BaseAvatar
  • Integrate audio feature splitting code into BaseAsr
  • Add Alibaba Cloud QwenTTS

Full Release Notes

1,重构代码,数字人模型和音频特征代码移到avatars目录下
2,数字人模型、tts、传输方式改成plugin方式接入
3,传输方式整理成单独的类,放到streamout目录下,添加rtmp输出
4,数字人推理和回贴公共部分代码整合到BaseAvatar中,各模型只需要实现自己特有部分代码
5,音频特征的切分代码整合到BaseAsr中
6,tts添加阿里云qwentts