Talking with Gemma 4 31B!

Hi! I'm Andi from Hugging Face. This is a fully open-source and free to test/pull/modify demo I'm bringing today.

It's a voice demo creating a pipeline of:
 - Nvidia's parakeet
 - Gemma 4 31B (served by cerebras!)
 - My  custom inference for Qwen3TTS

The whole stack is fully open-source , and is a drop-in replacement for OpenAI's realtime API. You can run it locally, I get similar latencies with a macbook pro M3 36GB and Gemma 4 E4B.

Here to the web based demo featured in the video , everything is running in the cloud.

For those who have been following, yes, this is the pipeline that runs on reachy minis :)