MS delays webrtc audio and video real-time AI calls. End-to-end local model
-
Langchain technology supports Openai,ollama and other common large model interfaces
-
Webrtc underlying millisecond latency,Native webrtc technology, easy to expand and integrate
-
SST,TTS audio recognition and synthesis using local models,Streaming synthesis
-
Vad double check noise reduction and audio monitoring
-
Supportive interruption
-
Supports video frame understanding dialog
-
Support h5, mini program, apps with uniapp
git clone https://github.com/heawon-yoon/webrtc-ai.git
tested with python3.10
cd backend
conda create -n webrtcai python=3.10 -y
conda activate webrtcai
pip install -r requirements.txt
python server.py
need to set the openai key and url from .evn file
如果没有官方的可以使用这个代理openai 申请获取api_key
没有访问限制,超级优惠,2人民币=1美元
OPENAI_API_BASE=https://api.xingyuntujiao.top/v1
OPENAI_API_KEY=XXXX
using uniapp,vue3
cd frontend
npm install -g @vue/cli
npm install
打开Hbuilder工具导入frontend项目. 选择运行到浏览器