This is a simple demo created to showcase browser only voicebot capabilities using client side inference.
It only requires CPU and works on any browser. This project has no server and everything is done client side.
The models used here are extremely tiny so please adjust your expectations and try to speak as clearly as possible.
Example questions:
- What is a CPU?
- What is a transformer?
- Who created Facebook?
Following libraries and models where used to make this demo possible:
- transformers.js
- whisper tiny
- hark (for speaker detection)
- LaMini-Flan-T5-783M
- Xenova/speecht5_tts
- antd
- tailwindcss
- react
- vite
- markdown-to-jsx