Bella Chai

Your laptop is now the voice of your whole house.

Boubou turns open-source TTS models into local runtimes with one command. Your phone, robot or smart home on the same Wi-Fi just talks to it.

100% local. Built on @LiquidAI_ + @huggingface 🤗

github.com/runboubou/boub…

What if voice AI was as easy as Ollama? 🎙️

1. Pick a language
2. Pick a voice
3. Boubou gives you one command

Paste it. Talk to AI on your own machine.
No cloud. No API key.
Your phone and your robot can join too.

npx boubou@beta

Open source ：github.com/runboubou/boub…

I turned my computer into Reachy Mini’s AI server.

Now it’s powered by local Qwen through Boubou, having a very serious conversation with Maya about… cooking ?

Boubou is open source: github.com/runboubou/boub…

And Reachy is just one example.

Any robot or app that supports OpenAI Realtime over WebSocket and lets you change the server URL can connect to Boubou, too.

Built on Hugging Face’s speech-to-speech pipeline, our local runtimes package supported models, voices, and dependencies together—ready to launch with one command.

Pick a runtime, fire it up on your computer, and point your device at it. You’re ready to talk.

Good point, thanks for adding this. Phones can slow down as they heat up, so a fast first run doesn’t tell us whether TTS can keep up with playback over a longer session.

LLMs have llama.cpp. Can speech models run locally the same way?

Yes. The weights and config can go into a GGUF file. The runtime builds a computation graph for the model, and GGML runs it on your CPU or GPU.

Much of the engineering carries over, including model support, quantization and validation. Streaming speech also needs to keep up with playback and stop when the user interrupts.