Kit-Bin
Donate

Text to Speech Online

Paste text, pick a voice, download an MP3. Nothing you type leaves this browser tab.

Processed entirely in your browser. Never uploaded to a server.

0 / 3000

Paste or type text, pick a voice, and click Generate to get an MP3 of it read aloud. The speech model runs on your device — nothing you type is uploaded anywhere.

Why there's no per-character limit tied to a subscription

Every hosted text-to-speech service that meters usage or requires an account does so because they're paying for server-side inference on their own hardware, every time you generate audio. That constraint doesn't exist here — the model runs in your browser, not on a server Kit-Bin pays for, so there's nothing to meter and no account to create. The one limit this tool does have — 3,000 characters per generation — is a flat cap chosen to keep synthesis time and memory reasonable on an average device, not a usage tier. For longer text, split it into chunks and generate each separately.

FAQ

How many voices and languages are there?
28 English voices — a mix of American and British, male and female. The underlying model is a small one, so this is the current tradeoff for keeping everything on-device; it does not support other languages yet.
Is there a usage limit?
Yes: 3,000 characters per generation. That's a device-memory limit, not a billing one — there's no server metering usage, so there's nothing to subscribe to or run out of. The cap exists purely to keep synthesis time and memory reasonable on an average phone or laptop.
Is my text sent anywhere?
No. The text you type is only ever handed to the speech model running in this browser tab. The one thing that does get downloaded is the model itself (about 80 MB, once per browser, cached afterwards) — that download is separate from your text and contains no information about it.
How natural does it sound?
Good for a model this size, not indistinguishable from a professional voiceover. It's an 82-million-parameter model chosen specifically because it can run entirely on-device — a cloud service with a multi-billion-parameter model will generally sound more natural, at the cost of your text leaving your device.
Can I use this on my phone?
Yes, though the first generation will be slower on a phone than a desktop while the model downloads and runs — subsequent generations in the same browser tab are faster since the model stays cached.

Related tools: MP3 to WAV, Trim Audio, Merge Audio.