Practical overview
When it fits
Use Quillman when you want a concrete starting point for a Moshi-powered streaming voice application, or want to explore its React frontend, Python server and WebSocket audio flow.
Who it is for
- Developers exploring speech-to-speech and continuous audio interactions.
- Teams adapting a React and Python starter for a browser-based voice assistant.
- Developers who want an included command-line client for streaming voice experiments.
Documented uses
- Build and try a browser interface for spoken, back-and-forth conversations.
- Develop the Moshi inference module and expose its WebSocket interface.
- Exercise the streaming connection with the included terminal client.
- Deploy the combined frontend and Moshi server to Modal.
Requirements & limits
The documented development and deployment workflow uses Modal and requires a Modal account and configured token. The repository is MIT-licensed; its README advises checking model licenses before commercial use.
- The application is built around Kyutai’s Moshi model, with Mimi and a speech-text foundation model in its audio and response pipeline.
- The React frontend and Python/FastAPI backend are separate parts of the project, with the frontend served as static files.
- The documented workflow relies on Modal for development and deployment.
- The README presents the code as an illustrative starting point and calls for checking model licensing before commercial use.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Develop and open the combined voice-chat app
Start the frontend and Moshi WebSocket server in Modal’s development environment and open the app.
Before you start
- The repository
- Python with Modal installed
- A Modal account and configured token
Steps
From the repository root, run
modal serve src.app.Use the URL printed in the terminal to visit the app.
Expected outcome
The development server serves the frontend and starts the Moshi WebSocket server; file changes are applied while the process runs.
Guide limitations
- Frontend changes may require clearing the browser cache.
- A Modal account and token are required for the documented development workflow.
Connect with the included command-line voice client
Use the repository’s terminal client to speak with the Moshi WebSocket server.
Before you start
- The repository
- Python and the Modal development environment configured
- A running Moshi development server
- Microphone and speakers
Steps
Create and activate a Python virtual environment.
Install the client dependencies with the command below.
pip install -r requirements/requirements-dev.txtRun the command below from the repository.
python tests/moshi_client.pySpeak using an enabled microphone and listen through the speakers.
Expected outcome
The included client connects to the WebSocket server for a spoken streaming interaction.
Guide limitations
- The client requires non-standard development dependencies.
- The Moshi server must be running separately.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- modal-labs/quillman · Official homepage title · retained in its source language