Fish Audio to MCP: Text to Speech, Voice Library, Voice Cloning and Transcription in any AI Agent

💡 No install? Use cloud.anythingmcp.com directly. Sign in, click Connectors → Fish Audio, paste your credentials, mint an MCP API key and you're done. No Docker, no git clone, no local server.

Fish Audio

Text to speech, voice cloning and speech to text with Fish Audio: search a library of community and licensed voices, generate natural multilingual speech with emotion tags as audio files, transcribe recordings with speaker turns, create, update and delete custom voice models and check the API credit. Speech and transcription are paid from prepaid API credit; a free text to speech model exists.

Setup

Setting up takes about 2 minutes. You need a Fish Audio account.

1. Sign up at https://fish.audio and open Developers, Get started (https://fish.audio/app/developers/get-started/).
2. Click Get your API key and copy it: it is shown only once. Paste it into FISH_AUDIO_API_KEY. It is sent as Authorization: Bearer.
3. Install the connector and run fish_audio_get_api_credit: it shows the developer API credit in US dollars and confirms the key.

API credit: new accounts start with no API credit. Text to speech with the model s2.1-pro-free works at once; every other model and transcription answer 402 until you add credit under Developers, Billing (https://fish.audio/app/developers/billing/). The developer API credit is separate from a web app subscription, which fish_audio_get_package shows.

What costs money: two tools draw from the API credit.

• fish_audio_text_to_speech: $15 per million UTF-8 bytes of text with s2.1-pro or s2-pro (about $0.015 per 1,000 Latin characters; Chinese and Japanese characters take 3 bytes each), $50 per million with drama-3, nothing with s2.1-pro-free.
• fish_audio_transcribe: $0.36 per hour of audio, billed per second.

Failed requests are not billed, and every other tool is free. Before a long text or recording the AI tells you the length and the cost and waits for your yes.

Finding a voice: fish_audio_list_voices searches hundreds of thousands of community and licensed voices by title, tag (male, female, young, old, narration, calm, deep ...) and language (en, de, zh ...). sort_by task_count puts the most used voices first, and licensed true returns only voices licensed by Fish Audio from their owners, safe for commercial use. Each voice has a sample_audio link to listen to. self true lists your own voices, including private ones. fish_audio_get_voice shows one voice; state trained means it is ready.

Speaking: fish_audio_text_to_speech takes the text and the voice id as reference_id (without one, a default voice is used) and attaches the audio file to the result. model is s2.1-pro by default; on an account without credit the AI uses s2.1-pro-free. format is mp3 (smallest), wav, opus or pcm; speed from 0.5 to 2 and volume in dB adjust the delivery; language helps read numbers, dates and units aloud. Cues in square brackets in the text set emotion and delivery: [happy], [sad], [excited], [whispering], [laughing], [sighing], [break]. Long texts are split into chunks automatically and come back as one file, but one result holds about 10 MB of MP3, so the AI splits a book into chapters.

Transcribing: fish_audio_transcribe takes audio_url, a public link to an audio or video file (mp3, wav, m4a, ogg, flac, webm, mp4 ...) of up to 10 MB, which AnythingMCP downloads and uploads. The default model transcribe-1-pro handles up to 60 minutes and several speakers: the text carries markers such as <|speaker:0|>. With ignore_timestamps false you also get word segments and speaker_turns with speaker, text, start and end. num_speakers or min_speakers and max_speakers help the speaker split; tag_audio_events false removes cues such as [laughter]. Speaker labels only tell speakers apart within one transcript.

Cloning a voice: fish_audio_create_voice takes a title and voice_url, a public link to 10 to 30 seconds of clean speech from one speaker, plus its transcript in text when you have it. The new voice is private unless you choose unlist, its _id works at once as reference_id, and it counts towards the voice limit of your plan. Only clone voices you have the right to use. fish_audio_update_voice changes the title, description or visibility of your own voice; tags are edited in the web app.

Data formats: ids are 32 hexadecimal characters. Dates are ISO 8601 in UTC. credit and cumulative_top_up are US dollars as decimal strings. Transcript timestamps are in seconds.

Paging: voice lists return total (capped at 1,000 for library searches) and has_more. The AI asks for the next page_number while has_more is true; page_size is 20 by default.

Switched off at install: fish_audio_delete_voice, because deleting a voice cannot be undone and no speech can be generated with it afterwards. A workspace admin turns it on when the team needs it.

Limits: concurrent requests depend on how much you have prepaid: 5 below $100, 15 from $100 and 50 from $1,000. A long transcription holds a slot until it finishes.

Common errors: 401 "Invalid token" means a wrong key. 402 means no API credit, or a paid model on an account without credit. 404 means an unknown voice id. 400 or 422 points to an invalid field, for example a speaker hint with diarize false. 429 means too many concurrent requests: the AI retries shortly after.

The connector does not design voices from a description, stream speech or transcription in real time over WebSocket, run Fish Audio's voice agents or use its image and video tools. Fish Audio also runs its own MCP server at https://api.fish.audio/mcp, which signs in with OAuth and bills the web app plan instead of the developer API credit.

Prerequisites: FISH_AUDIO_API_KEY

Install the connector locally

curl -fsSL https://raw.githubusercontent.com/HelpCode-ai/anythingmcp/main/docker-compose.quickstart.yml -o docker-compose.yml
printf 'JWT_SECRET=%s\nENCRYPTION_KEY=%s\n' "$(openssl rand -hex 32)" "$(openssl rand -hex 32)" > .env
docker compose up -d

Open http://localhost:3000/connectors/store, pick Fish Audio and paste in the env vars listed under Prerequisites.

Available tools

Example prompts

• "How much developer API credit is left on my Fish Audio account?"
• "Find a calm, deep male English narrator voice among the most used voices and give me its sample link."
• "Turn this product announcement into an MP3 with a warm female English voice, slightly slower than normal. Tell me the cost first."
• "Transcribe the interview at https://example.com/interview.mp3 with speaker turns and timestamps."
• "Create a private voice clone named Studio Narrator from my 20 second sample at https://example.com/my-voice.wav."

Next steps

• Connect to Claude
• Connect to ChatGPT