Getting Started

Desktop App

Download the latest release for your platform from the releases page:

Platform	File
macOS (Apple Silicon)	desktop-macos-arm64.zip
Linux (x64)	desktop-linux-x64.zip
Windows (x64)	desktop-windows-x64.zip

On first launch, Sirene will set up its inference environment automatically.

Docker

The quickest way to run Sirene as a server:

curl -sSL https://raw.githubusercontent.com/KevinBonnoron/sirene/main/install.sh | bash

Supply-chain note. The bootstrap script is fetched from main. For production deployments review the latest release and replace main with the corresponding tag once a release that supports your installation mode is published.

See the Docker guide for more options.

First Launch

1. Open the app (or navigate to http://localhost:5173 for Docker)
2. Go to the Models page and install the model of your choice (e.g. Kokoro v1.0)
3. You're ready to generate speech

Adding more inference servers

If you have a separate Linux machine with a GPU, you can add it as an inference server:

1. On that machine, run:

   curl -sSL https://raw.githubusercontent.com/KevinBonnoron/sirene/main/install.sh | INSTALL_MODE=inference bash

2. The script prints a URL and an auth token when it finishes

3. In Sirene → Settings → Inference servers → Add server, paste both, give it a name, save

Keep the auth token secret. It grants full control of the inference server. Do not paste it into chats, screenshots, or logs, and rotate it if you suspect it has leaked.

Rotating a token. Delete auth_token on the inference host, rerun the installer to mint a new value, then update the matching entry in Sirene → Settings → Inference servers with the new token. Until the entry is updated, every request to that inference server will return 401.

See the Docker guide for details.