This model requires you to accept the Apache 2.0 license terms before download.
Description
Kokoro 82M — hexgrad lightweight 82M-parameter text-to-speech model (built on StyleTTS 2) delivering natural, high-quality voices at very low compute cost, ideal for fast and on-device synthesis. Apache 2.0.
Tags
At a glance
| Kind | text-to-speech model |
|---|---|
| Precision | float32 |
| Container size | 353 MiB |
| Weights inside | 312 MiB |
| Components | 5 |
| Licence | Apache 2.0 |
What is inside the container
A .nbx container packs every part of Kokoro-82M — 5 components with weights, graphs and runtime settings — so one file is all the runtime needs.
| Component | Blocks | Layers | Weights |
|---|---|---|---|
| bert | 0 | — | 24.0 MiB |
| bert_encoder | 0 | — | 1.5 MiB |
| predictor | 0 | — | 61.8 MiB |
| decoder | 0 | — | 203 MiB |
| text_encoder | 0 | — | 21.4 MiB |
Default settings
The values the container ships with. They apply when you do not pass your own.
- Sample rate (Hz)
- 24000
- Temperature
- 1
Memory needed
The weights of Kokoro-82M take 312 MiB in float32. That is the floor for the memory the runtime has to place, spread over the devices it finds; activations come on top and grow with the size and length you ask for. A measured peak per device class is not published for this model yet.
Verify what you downloaded
Every download carries its SHA-256, and the file endpoint sends it in the X-NBX-SHA256 header. The container is 353 MiB (370,004,316 bytes).
$ sha256sum Kokoro-82M.nbx
96502026c5e788fe47b02ea206c4a443a6234289f002e6975e41c8c3fa00c797
Questions
What is Kokoro-82M?
A text-to-speech model by hexgrad, distributed as a float32 .nbx container and licensed Apache 2.0.
How big is the download?
353 MiB, of which 312 MiB are weights.
How do I run it?
Install the runtime from the download page, then use the commands under “CLI Usage” below: neurobrix import hexgrad/Kokoro-82M, then neurobrix serve --model Kokoro-82M.
CLI Usage
$ neurobrix import hexgrad/Kokoro-82M
$ neurobrix serve --model Kokoro-82M
$ neurobrix run --prompt "your prompt here"
$ neurobrix stop