Summary
We'd love to use Fish Audio's cloud API (api.fish.audio) as a TTS provider in OpenClaw, a popular open-source AI gateway with 30k+ stars. OpenClaw has built-in TTS support for OpenAI, ElevenLabs, Google, Microsoft, and MiniMax — all through a standardized speech provider interface.
Adding an OpenAI-compatible TTS endpoint (POST /v1/audio/speech) to the Fish Audio cloud API would allow Fish Audio to work as a drop-in TTS provider in OpenClaw and any other application that supports the OpenAI TTS API format.
Why This Matters
- OpenClaw has thousands of users who could instantly use Fish Audio for voice messages across Telegram, Discord, WhatsApp, Signal, Slack, and 10+ other channels
- Zero integration work for end users — just point
messages.tts.providers.openai.baseUrl at https://api.fish.audio and set the voice ID
- Fish Audio's S2 Pro model achieves the lowest WER on Seed-TTS Eval among all models — this quality deserves wider reach
- Discussion How should I make it compatible with openai-api? #933 asked about this for self-hosted setups, but the cloud API (
api.fish.audio) would have much broader impact
Proposed Endpoint
POST https://api.fish.audio/v1/audio/speech
Content-Type: application/json
Authorization: Bearer <token>
{
"model": "s2-pro",
"input": "Hello world",
"voice": "98655a12fa944e26b274c535e5e03842",
"response_format": "opus"
}
Where voice maps to Fish Audio's reference_id. This follows the OpenAI TTS API spec.
Current Workaround
We currently call Fish Audio's native /v1/tts endpoint via a custom Python script with msgpack serialization. It works but requires manual exec pipelines and can't use OpenClaw's built-in TTS delivery system (which handles voice note formatting, streaming, and multi-channel delivery automatically).
Impact
This would make Fish Audio accessible to the entire OpenClaw ecosystem with a single config change:
{
"messages": {
"tts": {
"auto": "inbound",
"provider": "openai",
"providers": {
"openai": {
"baseUrl": "https://api.fish.audio",
"apiKey": "your-fish-key",
"model": "s2-pro",
"voice": "98655a12fa944e26b274c535e5e03842"
}
}
}
}
}Happy to help test if you decide to implement this!
Summary
We'd love to use Fish Audio's cloud API (
api.fish.audio) as a TTS provider in OpenClaw, a popular open-source AI gateway with 30k+ stars. OpenClaw has built-in TTS support for OpenAI, ElevenLabs, Google, Microsoft, and MiniMax — all through a standardized speech provider interface.Adding an OpenAI-compatible TTS endpoint (
POST /v1/audio/speech) to the Fish Audio cloud API would allow Fish Audio to work as a drop-in TTS provider in OpenClaw and any other application that supports the OpenAI TTS API format.Why This Matters
messages.tts.providers.openai.baseUrlathttps://api.fish.audioand set the voice IDapi.fish.audio) would have much broader impactProposed Endpoint
Where
voicemaps to Fish Audio'sreference_id. This follows the OpenAI TTS API spec.Current Workaround
We currently call Fish Audio's native
/v1/ttsendpoint via a custom Python script with msgpack serialization. It works but requires manual exec pipelines and can't use OpenClaw's built-in TTS delivery system (which handles voice note formatting, streaming, and multi-channel delivery automatically).Impact
This would make Fish Audio accessible to the entire OpenClaw ecosystem with a single config change:
{ "messages": { "tts": { "auto": "inbound", "provider": "openai", "providers": { "openai": { "baseUrl": "https://api.fish.audio", "apiKey": "your-fish-key", "model": "s2-pro", "voice": "98655a12fa944e26b274c535e5e03842" } } } } }Happy to help test if you decide to implement this!