Text to Speech (TTS)

AnythingLLM supports multiple text to speech providers so chat responses can be read out loud. Once a provider is configured, click the speaker icon under any chat response to hear it spoken.

Selecting a TTS Provider

Open Settings and select Voice & Speech. Under Text-to-speech Preference, choose a provider from the Provider menu, fill in its settings, and click Save changes.

The default provider is System native, which works with no configuration at all.

Supported TTS Providers

Local TTS Providers

These run on your device and require no API key.

System native

Uses your browser's built in speech synthesis, if supported. Nothing is sent to any external service. Voice quality depends on your browser and operating system.

PiperTTS

Runs open source Piper voice models locally in your browser. The voice model is downloaded once, and all audio is generated privately on your device.

Self Hosted TTS Providers

Kokoro

Connects to a self hosted kokoro-fastapi server for high quality open source voices. Set the server URL (for example http://localhost:8880/v1), an optional API key if your server requires one, and a voice such as af_bella.

Cloud TTS Providers

OpenAI

Uses OpenAI's (opens in a new tab) hosted text to speech voices. Requires an OpenAI API key and a voice selection such as nova.

ElevenLabs

Uses ElevenLabs (opens in a new tab) voices and technology. Requires an ElevenLabs API key, and you can choose from the voices available to your ElevenLabs account.

Any OpenAI Compatible TTS Service

The OpenAI Compatible provider connects AnythingLLM to any service that implements the OpenAI speech API (POST /v1/audio/speech), whether it runs on your own network or is a hosted API. It takes four settings:

Base URL: the root of the service, including the version path, for example https://api.openai.com/v1
API Key: optional, only needed if the service requires one
TTS Model: the model parameter sent with each request
Voice Model: the voice identifier sent with each request

Environment Variables

If you deploy AnythingLLM with Docker or bare metal and prefer to configure TTS through environment variables, these are the supported keys (from docker/.env.example):

# System native (default)
TTS_PROVIDER="native"
 
# OpenAI
TTS_PROVIDER="openai"
TTS_OPEN_AI_KEY=sk-example
TTS_OPEN_AI_VOICE_MODEL=nova
 
# ElevenLabs
TTS_PROVIDER="elevenlabs"
TTS_ELEVEN_LABS_KEY=
TTS_ELEVEN_LABS_VOICE_MODEL=21m00Tcm4TlvDq8ikWAM
 
# Kokoro
TTS_PROVIDER="kokoro"
TTS_KOKORO_ENDPOINT="http://host.docker.internal:8880/v1"
TTS_KOKORO_KEY=
TTS_KOKORO_VOICE_MODEL=af_bella
 
# Any OpenAI compatible service
TTS_PROVIDER="generic-openai"
TTS_OPEN_AI_COMPATIBLE_ENDPOINT="https://api.openai.com/v1"
TTS_OPEN_AI_COMPATIBLE_KEY=sk-example
TTS_OPEN_AI_COMPATIBLE_MODEL=tts-1
TTS_OPEN_AI_COMPATIBLE_VOICE_MODEL=nova

Set only one TTS_PROVIDER value. System native and PiperTTS run in the browser, so PiperTTS is configured from the UI rather than through server variables.

Language Models Speech to Text (STT)