Open-weight model
Kokoro TTS Online
Try Kokoro-82M in your browser — no Python, no GPU, no model download. Type your text, pick a voice, and download the MP3. It is the same open-weight model that powers our free tier.
Speech studio
Natural voices, ready in seconds
5,000 free characters/day · No sign-up required · Commercial use included
Why Kokoro got popular
Most open TTS models ask you to trade something away: quality, speed, or a license that quietly forbids commercial use. Kokoro-82M is unusual because it gives you all three — genuinely natural output, fast inference even on modest hardware, and a clean Apache-2.0 license.
At roughly 82 million parameters it is a fraction of the size of most competing models, which is what makes it cheap enough to run at scale. That combination is why it shows up as the default engine in so many self-hosted setups.
We run Kokoro too — that is the point
The free tier of this site is powered by Kokoro. So when you generate audio above, you are hearing the real model, not a demo with a different engine swapped in behind the scenes.
That also means you can use this page as a straightforward capability check: if Kokoro sounds good enough for your project here, it will sound the same when you self-host it. If it does not, you can switch to a premium engine in the voice selector and compare them side by side in the same session.
When to self-host vs. use it here
A rough guide:
- Use it here when you want to evaluate the model, generate a handful of files, or avoid infrastructure entirely
- Self-host when you have steady high volume, need offline operation, or want to fine-tune
- Use a premium engine when you need strong emotional range, voice cloning, or broad language coverage beyond what Kokoro handles