TinyTavern is a chat app you run yourself. Connect a local model or an OpenAI-compatible service, talk with characters, and create images and videos through ComfyUI.
Edit messages, try alternative replies, and explore different paths through a conversation. Your chats and gallery stay together, whether you open them on your phone or desktop.
You’ll need Docker with Compose and an HTTPS certificate.
Place the certificate in certs/cert.pem and its private key in certs/key.pem, then run:
./scripts/init-caddy.sh --media-dirs
docker compose -f docker-compose.yml up --build -d tinytavern caddy-prodOpen https://:5487 in your browser, using your server’s address in place of <host>.
- Open Settings → Model connections. Add your model’s API URL (including
/v1), enter an API key if needed, choose a model, and save the connection. - Start a new chat with Assistant and send a message.
- For character chats, create a character or import a SillyTavern PNG card under Settings → Characters.
To require sign-in, set a password under Settings → General.
Set your ComfyUI address under Settings → Generation settings, then add a workflow exported in API format under Settings → Workflows. Open Gallery → Tools → Generate media, choose a workflow, and add any input images. Prepare a prompt or write your own, then click Generate. Review the variations and Save your favorites.
Text workflows return text you can copy or reuse as a prompt. When generating from a chat, Add saves the text as a message. Avatar generation uses the same model connection settings and skips prompt preparation for workflows without a prompt input.
To update, pull the latest code and rerun the Compose command above.
Keep data/ and .secrets/; they hold your saved work and access keys.
Setup details, backup commands, and development guidance are in AGENTS.md.