SillyTavern TTS: enable it and give every character a voice
SillyTavern can read your role-play chats aloud through its TTS extension. It takes a few clicks to turn on, and each character can have a voice of its own.
This guide covers the setup, an honest look at the providers, and how to use free offline character voices.
Turn on TTS
Open the Extensions panel, find TTS and choose a provider. Tick Enabled to turn playback on, and Auto Generation if new messages should be read aloud as they arrive. If no voices appear, check the provider settings and click Refresh.
Which provider to choose
The built-in list includes System, Edge, Google Translate, Google Gemini TTS, Pollinations, Kokoro, Silero, AllTalk, XTTS, ElevenLabs, OpenAI, Azure, Novel and MiniMax. For a first try with nothing to install, start with System or Edge. Kokoro, Silero, AllTalk and XTTS use a voice server you run yourself. ElevenLabs, OpenAI, Azure, Novel and MiniMax are paid services that need an account or key, so check their current prices.
A different voice for each character
Open a chat with a character, or select your persona, and pick a voice from the dropdown in the voice map. Do this for each one, including the narrator if you use one, then click Apply. Group chats sound much better when every character has a distinct voice.
Read only the dialogue
Only narrate "quotes" limits speech to text in quotation marks, so actions are not read aloud. Ignore text inside asterisks skips *actions*; with both ticked, only quotes outside asterisks are spoken. Apply regex lets you strip anything else before it reaches the voice, and Narrate only the translated text helps if you chat through a translation.
Free character voices, offline
To use voices such as game and film characters without a per-character bill, run a small local server with the free SuperClonica engine: our guide to an OpenAI-compatible TTS server has the full code (python -m pip install superclonica-cpu fastapi uvicorn av). In SillyTavern choose the OpenAI Compatible provider, enter http://localhost:8880/v1/audio/speech as the endpoint and superclonica as the model. This provider does not fetch voices, so type the list yourself: M1,M2,M3,M4,M5,F1,F2,F3,F4,F5. Then assign them in the voice map.
Add a catalog voice
Try any voice from the SuperClonica catalog free on its page. To use it in SillyTavern, get its voice file and add a line to the server, such as VOICES["narrator"] = "/path/to/the/file". Add narrator to the voice list in SillyTavern and assign it to a character. The voice speaks 31 languages and sounds most natural in the language it was recorded in.