- Home
- Skills
- APIs & Backend
- fish audio
fish audio
Works with the AI Agents you already use
fish audio
High-fidelity multi-speaker TTS and voice cloning using the fish.audio API with automated subtitle alignment.
$1
fish audio
Example session with this skill installed
Generate a dialogue between Dudu and Tuan using the script:
Dudu: Look at the moon!
Tuan: It is so bright tonight.
Include subtitles.
- Read your context and instructions
- Compiled the fish audio
Processed the dialogue request using fish.audio S2.1-pro model. Generated a 4.2-second audio file and a synchronized SRT subtitle file.
Dialogue Synthesis Results
| File Type | Filename | Duration | Speaker Mapping |
|---|---|---|---|
| Audio | output.mp3 | 00:00:04.200 | Dudu (id: e98e3...), Tuan (id: efddb...) |
| Subtitles | output.srt | N/A | Character-labeled (Dudu/Tuan) |
1
00:00:00,000 --> 00:00:01,850
Dudu: Look at the moon!
2
00:00:02,100 --> 00:00:04,200
Tuan: It is so bright tonight.
The audio was generated with a single request using <|speaker:0|> and <|speaker:1|> tags. Silence detection identified a 250ms gap between speakers, which was used to align the timestamps.
Next steps
- Run
python scripts/fish_tts.py plot script.txt output.mp3to visually verify the alignment. - Adjust speech rate by adding
--speed 0.9if the delivery is too fast for the animation. - Use the
voicescommand to find a different narrator voice for the introduction.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop manually stitching individual audio clips for multi-character dialogue. This skill integrates the fish.audio API to generate high-quality text-to-speech, multi-speaker synthesis, and zero-shot voice cloning for audiobooks, children's stories, and video voiceovers.
What it does
- Multi-character synthesis generates entire dialogues with multiple speakers in a single API request using
<|speaker:N|>tags. - Zero-shot voice cloning creates custom voices from a 10-30 second reference audio clip without fine-tuning.
- Automated subtitle alignment uses dynamic programming and silence detection to map script lines to generated audio timestamps.
- Voice library search queries the fish.audio database to find specific timbres like child or narration voices.
- Waveform visualization generates SVG plots to verify subtitle alignment against audio envelopes.
How it works
- Search for voice IDs using the
voicescommand with keywords like "child" or "narration". - Synthesize audio by providing a script file or inline lines, mapping characters to specific voice IDs via the
--castflag. - Align text to existing audio using the
alignordialogue --srtcommands to generate.srtsubtitle files. - Clone new voices by providing a clean reference
.wavfile and its transcript to theclonecommand.
Frameworks & tools
The skill utilizes httpx for API communication and ormsgpack for binary voice cloning payloads. It relies on ffmpeg and ffprobe for silence detection and audio processing. The client logic is contained in scripts/fish_tts.py.
Why this beats prompting it yourself
Writing raw API wrappers for fish.audio requires complex MessagePack handling for binary audio uploads and intricate tag management for multi-speaker requests. This skill automates key rotation across multiple tokens and provides a robust alignment engine that outperforms simple silence-based splitting.
Use cases
- Generating synchronized dialogue for animated children's story videos.
- Producing high-fidelity audiobooks with distinct voices for different characters.
- Creating localized voiceovers using zero-shot cloning for consistent branding.
- Batch processing scripts into audio files with matching
.srtsubtitles.
Known limitations
Multi-speaker synthesis requires S2 series models. Voice cloning requires ormsgpack and clean reference audio. It does not support the fish-speech open-source model or local inference.
Requires S2 series models for multi-speaker support. Voice cloning requires ormsgpack. Only supports fish.audio API, not local fish-speech inference.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 15 days ago
Requires
Permissions
Allowed Hosts