Natural text to speech from a tiny 82M model. Made by hexgrad.
pip install kokoro
Kokoro TTS is a speech or music model from hexgrad. Natural text to speech from a tiny 82M model. Running it locally needs about 2 GB of memory.
The least memory you need, by Apple chip:
Runs well on: RTX 3060 12GB, RTX 3070, RTX 3080 10GB, RTX 3090, RTX 4060, RTX 4060 Ti 16GB, RTX 4070 / 4070 Super, RTX 4070 Ti Super, RTX 4080 / 4080 Super, RTX 4090, RTX 5060, RTX 5060 Ti 16GB, RTX 5070, RTX 5070 Ti, RTX 5080, RTX 5090, Radeon RX 7800 XT, Radeon RX 9070 XT, Radeon RX 7900 XTX.
As an Amazon Associate I earn from qualifying purchases.
About 2 GB of video memory or unified memory. On a Mac, the smallest configuration that runs it is 8 GB.
A 16 GB M4 Mac runs it runs well, about about 11 s per minute of speech. An 8 GB M1 Mac runs it runs well, about about 20 s per minute of speech.
Yes, the model is free to download. Its licence is listed as Apache 2.0. That is a permissive licence that allows commercial use.
pip install kokoro.
| Machine | Verdict | Speed |
|---|---|---|
| M1 8GB | Runs well | about 20 s |
| M4 16GB | Runs well | about 11 s |
| M4 32GB | Runs well | about 11 s |
| M4 Pro 48GB | Runs great | about 6 s |
| M4 Max 128GB | Runs great | about 4 s |
| M3 Ultra 512GB | Runs great | about 4 s |
| RTX 4060 | Runs well | about 11 s |
| RTX 4070/4070 Super | Runs great | about 6 s |
| RTX 5070 Ti | Runs great | about 4 s |
| RTX 3090 | Runs great | about 6 s |
| RTX 5090 | Runs great | about 2 s |
| Ryzen AI Max+ 395 128GB | Runs great | about 6 s |
| CPU only 32GB | Runs well | about 2 min |