Kokoro and its versions
Kokoro is the small open-weight voice model that showed 82 million parameters could sound good enough. Version 1.0 from January 2025 is the release on fal and has the page here, reading the same four scripts as every voice model on this site.
hexgrad · facts checked
Which one to pick
Side by side, measured
| Model | Price | Seconds per reading | 4 scripts cost | Voices |
|---|---|---|---|---|
| Kokoro | $0.02 per 1,000 characters via fal | 0 to 1 | $0.0195 | Nineteen American English voices on fal's endpoint, 54 across eight languages in the weights; we used af_heart |
Seconds and cost are from our own runs of the same 4 prompts through each model; open a model for every reading and the request behind it.
Every version
| Version | Released | What changed |
|---|---|---|
| Kokoro v0.19 | First release, topped the TTS arena among open models. | |
| Kokoro v1.0 kokoro-82m | 54 voices across eight languages, Apache 2.0, pip package. |
Frequently asked questions
Is Kokoro open source?
Yes, Apache 2.0 weights and code.