Connect Fish Audio TTS to Coze

Use an HTTP request node in a Coze workflow to turn text generated by your bot into Fish Audio speech. The endpoint follows the OpenAI-compatible TTS request format.

Before you start

1. Create a dedicated Fish Audio API key.
2. Copy a Fish Audio Voice ID from the voice library.
3. Create or open a Coze workflow that receives a text input.

Configure the HTTP request

Add an HTTP request node after the text-generating node and use these settings:

Use this JSON body:

{
  "model": "fishaudio-s21pro-flash",
  "voice": "00a1b221-6137-4b73-ad62-b0cbce134167",
  "input": "{{text}}",
  "response_format": "mp3"
}

Map {{text}} to the workflow variable that contains the bot response. Configure the node response as a file or binary output, then return that output from the workflow.

Test the workflow

Run the workflow with a short sentence first. Confirm that the HTTP node returns an MP3 file before connecting the workflow to a published bot or app.

Troubleshooting

• 401: verify that the Authorization header starts with Bearer and uses a Fish Audio API key.
• 402: add API credits; membership credits and API credits are separate balances.
• No audio output: configure the HTTP response as a file or binary value instead of JSON.
• Invalid voice: replace the example Voice ID with a valid system or custom Fish Audio Voice ID.
• Request timeout: test with a shorter input and keep each request focused on one paragraph.

Related documentation

• OpenAI-compatible TTS
• Create a custom voice
• Coze workflow and chatflow guide