Connect Fish Audio TTS to OpenCrabs

OpenCrabs supports any /v1/audio/speech compatible endpoint and can turn agent replies into voice messages for Telegram, WhatsApp, Discord, or Slack.

Configure TTS

Edit config.toml:

[voice]
tts_enabled = true
tts_mode = "openai_compatible"
tts_base_url = "https://fishaudio.org/v1"
tts_model = "fishaudio-s21pro-flash"
tts_voice = "00a1b221-6137-4b73-ad62-b0cbce134167"

Store the API key in keys.toml:

[providers.tts.openai_compatible]
api_key = "YOUR_API_KEY"

Do not commit keys.toml to a public repository. You can also run /onboard:voice, select OpenAI-compatible, and enter the same URL, model, and voice interactively.

Test the connection

Restart OpenCrabs and trigger a short response through a connected channel. After the audio plays, replace the test voice with your preferred Voice ID.

OpenCrabs handles audio encoding for each messaging channel. First verify that the audio file plays, then check whether the channel displays it as a native voice bubble.

Troubleshooting

• 404 endpoint: the Base URL is https://fishaudio.org/v1, not a documentation page URL.
• Local Piper is still used: set tts_mode to openai_compatible.
• No speech is generated: confirm tts_enabled = true and that the channel permits audio attachments.
• 401/402: verify the key in keys.toml and the API credit balance.

Related documentation

• OpenAI-Compatible TTS
• API errors
• Official OpenCrabs voice guide