GitHub radar
Build your own AI livestream host that runs and chats live
Open-yoki is an open-source framework for an AI livestreaming host: it clones a voice, shows a digital face, remembers viewers, and goes live on a schedule. Configure a personality and you get your own AI streamer.
It's a set of services that give an AI a voice, a moving digital face, viewer memory, and a streaming schedule.
When it helps
Useful if you want to build your own AI host for livestreams or a voice chatbot. Voice and video need your own server (4 cores, 8GB RAM); a regular laptop only runs the text-only test mode.
Pros
- Free test mode runs with zero paid API keys
- Diary and viewer notes are stored in plain text files
- Streams to Binance Square or any RTMP-compatible platform
- Swap the chat model for Claude Code or DeepSeek
Cons
- Voice and a moving face need a separate paid server and keys
- Docs are in Chinese; English README covers only part of it
- It's a demo with no money features — no wallet, trading, or betting
How to set it up — step by step
- 1Open your AI agent (Claude Code or Codex) inside your project folder
- 2Send the agent the text from the block below — it installs Node.js, ffmpeg, and runs test mode
- 3Check the result: open http://127.0.0.1:4370 and watch the streamer run with zero paid keys
- 4Want voice and a moving face? Give the agent OpenRouter, Volcano Engine, and Vidu keys
- 5To build your own streamer, fill in profile.mjs and config.mjs inside the example folder app/idols/
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up open-yoki's local test mode from here: https://github.com/jacke-pp/open-yoki — install Node.js and ffmpeg, run "cd app && npm install && npm test", then "IDOL=example AVATAR=mock MODE=test node server/index.mjs" and open http://127.0.0.1:4370 to see it load with no paid keys at all; if you want to hear the voice and see it move, ask me for OpenRouter, Volcano Engine, and Vidu keys.
You need your own server (the README recommends 4 cores and 8 GB RAM, Linux x86_64) — a regular laptop only runs the no-voice, no-video test mode.
Open on GitHub▌ More finds