hexgrad · Apache 2.0
Kokoro
A tiny 82M voice model with 54 preset voices in 8 languages. Runs on almost anything, even without a graphics card.
How much memory it needs
Memory to run each model, from its model card or the size of its files. Smaller versions trade some quality for less memory.
| Model | Version | Needs | Business use |
|---|---|---|---|
| Kokoro-82M | 8-bit ONNX | 0.09 GB | Allowed |
| FP32 | 0.4 GB |
Which computers run it
The best model in this group for each machine.
| Computer | Best version | Fit |
|---|---|---|
| Laptop or PC, no graphics card, 32 GB | Kokoro-82M FP32 | Runs great |
| RTX 4060 · 8 GB | Kokoro-82M FP32 | Runs great |
| RTX 3060 · 12 GB | Kokoro-82M FP32 | Runs great |
| RTX 5060 Ti · 16 GB | Kokoro-82M FP32 | Runs great |
| RTX 4090 · 24 GB | Kokoro-82M FP32 | Runs great |
| RTX 5090 · 32 GB | Kokoro-82M FP32 | Runs great |
| Mac with M4 · 16 GB | Kokoro-82M FP32 | Runs great |
| Mac with M4 Pro · 48 GB | Kokoro-82M FP32 | Runs great |
| Mac with M5 Max · 128 GB | Kokoro-82M FP32 | Runs great |
| Ryzen AI Max+ 395 · 128 GB | Kokoro-82M FP32 | Runs great |
| DGX Spark · 128 GB | Kokoro-82M FP32 | Runs great |
| RTX PRO 6000 · 96 GB | Kokoro-82M FP32 | Runs great |
| Mac Studio with M5 Ultra · 256 GB | Kokoro-82M FP32 | Runs great |
| Mac Studio with M3 Ultra · 512 GB | Kokoro-82M FP32 | Runs great |
What it is good and bad at
Good at
- Tiny and fast
- Runs on CPU
- Apache 2.0
Not good at
- No voice cloning
Start it
- Kokoro-82M: runs in kokoro (PyTorch) (also kokoro-onnx, Transformers.js (ONNX), MLX (mlx-community)). Model card