Minimal Conversation · S2S backend

Hugging Face Realtime

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

Models Nemotron 3.5 Transcribe 30B-A4B · Gemma 4 31B · Qwen3-TTS 1.7B Powered by speech-to-speech Built by Hugging Face · andito · tfrere · A-Mahla

Language

Tap to start

Applies when you start a conversation

Reasoning level

Higher levels can take longer to reply. Available levels depend on the model.

camera

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. Models can run locally or through a provider, depending on your backend configuration.

View the project on GitHub

Example pipeline

You speak
VAD detects speech silero-vad
STT transcribes it NVIDIA-Nemotron-3.5-Transcribe-30B-A4B
VLM analyses and composes the reply google/gemma-4-31B-it
TTS speaks it back Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice
The orb replies

Settings

Speech-to-speech server URL URL of your speech-to-speech server, e.g. http://localhost:8080 (the app adds /v1/realtime). Transport How audio travels to the server. Applies on the next conversation.

Voice Conversation language Sets transcription and speech for this conversation.

Show reasoning Show collapsible reasoning in Conversation. Applies immediately.

Microphone Applies on the next conversation (or Restart). Speakers Where assistant audio plays. Chrome/Edge support switching outputs. Faster interruptions Lowers the assistant's volume when you start speaking. Applies on the next conversation. Playback startup buffer (ms) WebSocket only. 0 starts immediately; larger values delay playback to absorb delivery gaps. Applies on the next conversation.

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.

Instructions

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.

Search API key

Camera Let the assistant see through your webcam.

A live preview shows bottom-left while the camera is on.