Skip to content

Kokoro

Kokoro is a fast, lightweight (82M parameter) multilingual TTS model with 54 built-in voice presets. It delivers high-quality speech synthesis with minimal resource usage, making it ideal for quick generation tasks on Apple Silicon.

Model Variants

Model Format HuggingFace
mlx-community/Kokoro-82M-bf16 bfloat16 Model Card

Installation

Kokoro requires misaki for text processing in all languages:

# Base dependency
pip install misaki

# Optional language support
pip install misaki[ja]
pip install misaki[zh]

Usage

# Basic generation (American English)
mlx_audio.tts.generate \
    --model mlx-community/Kokoro-82M-bf16 \
    --text "Hello, world!" \
    --lang_code a

# Choose a voice and adjust speed
mlx_audio.tts.generate \
    --model mlx-community/Kokoro-82M-bf16 \
    --text "Welcome to MLX-Audio!" \
    --voice af_heart \
    --speed 1.2 \
    --lang_code a

# Play audio immediately
mlx_audio.tts.generate \
    --model mlx-community/Kokoro-82M-bf16 \
    --text "Hello!" \
    --play \
    --lang_code a

from mlx_audio.tts.utils import load_model

model = load_model("mlx-community/Kokoro-82M-bf16")

for result in model.generate(
    text="Welcome to MLX-Audio!",
    voice="af_heart",  # American female
    speed=1.0,
    lang_code="a",     # American English
):
    audio = result.audio  # mx.array waveform

Language Codes

Code Language Extra Dependency
a American English pip install misaki
b British English pip install misaki
j Japanese pip install misaki[ja]
z Mandarin Chinese pip install misaki[zh]
e Spanish pip install misaki
f French pip install misaki

Available Voices

American English

Female Male
af_heart am_adam
af_bella am_echo
af_nova
af_sky

British English

Female Male
bf_alice bm_daniel
bf_emma bm_george

Japanese

Female Male
jf_alpha jm_kumo

Chinese

Female Male
zf_xiaobei zm_yunxi

Note

Kokoro ships with 54 voice presets total. The voices listed above are a representative sample. Check the model card for the full list.