Hugging Face Realtime

An open, real-time voice chat built on Hugging Face's speech-to-speech stack.

Powered by Inference Endpoints Cerebras Built by Hugging Face tfrere A-Mahla andito

Tap to start

Conversation

About

Speech-to-speech is Hugging Face's open framework for real-time voice agents. Rather than one end-to-end model, it chains four open models from the Hub (speech detection, transcription, a vision-language model, and synthesis), so any stage can be swapped or run locally. This demo wires that pipeline to hosted inference.

View the project on GitHub

The pipeline

  1. You speak
  2. VAD detects speech silero-vad
  3. STT transcribes it nvidia/parakeet-tdt-1.1b
  4. VLM analyses and composes the reply · via Cerebras google/gemma-4-31B-it
  5. TTS speaks it back Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice
  6. The orb replies

Settings

Noise gate Off
Off −3 dB
Mutes the mic below the handle so room noise isn't sent. Slide fully left to turn it off; the bar shows your live input while in a call.
Connect first, then come back here to apply live changes.

Tools

Let the assistant act during the conversation. Changes apply live.

Web search Look things up on Google, via Serper.
Camera Let the assistant see through your webcam.
A live preview shows bottom-left while the camera is on.

That's a wrap for now