Scripts in. Narration out.
Write a script, pick a voice or design your own, and hear it in seconds. rat.audio is 100% free, forever: the voice model runs on your device, so nobody pays anything. Your first generation needs no account. After that, a free Google sign-in keeps it going, with no limits.
↓ scroll
01 — Studio
Write it. Hear it.
Output
02 — Voice Lab
Make a voice that's yours.
Mix up to three of Kokoro's 33 English and Japanese voices, then tune pitch and pace. Your voices are saved in this browser, and you can share any of them with a link.
Record your voice
betaRead the line below out loud. rat.audio measures your pitch, tone and pace right here on your device, then builds the closest voice it can from Kokoro's voices. It's a match, not an exact clone, and your recording never leaves this browser.
“The quick brown fox jumps over the lazy dog. I am recording my voice so rat dot audio can learn how I sound. Pretty cool, right?”
My voices
Nothing saved yet. Your saved voices show up here and in the Studio's voice menu.
03 — Connect
Use it from ChatGPT or Claude.
Add rat.audio as a connector, then just ask: “narrate this in the Rocco Spinelli voice.” The AI writes the script and hands you a link that plays it instantly. No sign-in for the connector, and it's free.
https://rat-audio-mcp.skelr.workers.dev/mcp
ChatGPT
Settings → Apps & Connectors → Advanced settings → turn on Developer mode. Then Create a connector, paste the URL and choose No authentication.
Claude
On claude.ai or Claude Desktop: Settings → Connectors → Add custom connector, name it rat.audio and paste the URL.
Claude Code
Run this once in your terminal:
claude mcp add --transport http rat-audio https://rat-audio-mcp.skelr.workers.dev/mcpPut it in your project
Voice-overs that land right in your code.
- Ask your AI for the audio and where it goes, e.g. “make an intro voice-over in Rocco Spinelli's voice and add it to
assets/audio/intro.wav.” - Open the link it gives you. rat.audio renders the voice in your browser, plays it, and sends the finished WAV back to your AI automatically.
- Your AI downloads the file into the project and wires it up: a game sound, a site's intro, an app's onboarding or a video's narration.
Works with any AI that can save files: Claude Code, Codex, Cursor, ChatGPT agent. In plain chat apps you get a download link to drop into your project yourself. Files stay available for 7 days.
# 1. rat.audio → generate_speech → you open the link # 2. rat.audio → get_audio → file URL curl -L -o assets/audio/intro.wav \ "https://rat-audio-mcp…/audio/<job>.wav" // 3. use it in your project const intro = new Audio("assets/audio/intro.wav"); startButton.onclick = () => intro.play();
Tools: generate_speech · get_audio · create_voice · list_voices. Audio is rendered in your browser. The finished WAV is kept for 7 days so your AI can download it into a project. Coding agents can also use the local connector to write MP3 or WAV files straight into a repo.
04 — Free, for real
100% free. Forever.
Free forever, for everyone
The voice model runs on your own device, so there are no servers to pay for and nothing to charge you. There are no plans, credits or trials, and there never will be.
Private by default
Your scripts and voices never leave your browser. No account, no uploads, no tracking.
Per word
On WebGPU, rat.audio renders faster than you can listen, and playback starts after the first sentence.
Your first generation needs no account. After that, a free Google sign-in keeps generations unlimited. The first visit downloads the Kokoro-82M model (90–330 MB depending on your device). Your browser caches it after that.