Free local text to speech for AnythingLLM

1Start the speech server

With Docker, which includes FFmpeg (AnythingLLM plays MP3 audio):

docker run -d --name loudness-tts -p 8880:8880 -v loudness-tts:/data ghcr.io/loudness-ai/loudness-tts

Or with Python, after installing FFmpeg. For the desktop app on the same computer, the default address is enough:

python -m pip install loudness-tts
loudness-tts serve --port 8880

If AnythingLLM runs in Docker, start it with --host 0.0.0.0 instead, and add --api-key with a secret on shared networks.

2Choose the provider

In AnythingLLM, open Settings → Voice & Speech. Under Text-to-speech Preference, set Provider to OpenAI Compatible.

3Fill in the fields

• Base URL: http://127.0.0.1:8880/v1 for the desktop app, or http://host.docker.internal:8880/v1 when AnythingLLM runs in Docker
• API Key: leave it empty, unless you started the server with --api-key
• TTS Model: tts-1
• Voice Model: nova, or a voice id

Then select Save changes.

4Test it

Open a workspace chat and press the speaker icon under an answer. The first answer can take a few seconds longer while the voice downloads.

5Optional: set it with environment variables

For AnythingLLM in Docker, the same settings can go in its .env file:

TTS_PROVIDER="generic-openai"
TTS_OPEN_AI_COMPATIBLE_ENDPOINT="http://host.docker.internal:8880/v1"
TTS_OPEN_AI_COMPATIBLE_MODEL="tts-1"
TTS_OPEN_AI_COMPATIBLE_VOICE_MODEL="nova"
# TTS_OPEN_AI_COMPATIBLE_KEY="choose-a-secret"   (only with --api-key)

See AnythingLLM’s text-to-speech documentation for the other providers.