Phonon

Phonon is a text-to-speech model supported by MLX Audio. It generates 24 kHz speech on Apple Silicon, with four bundled voices and support for streaming audio generation.

Installation

pip install -U mlx-audio

Usage

Python API:

import numpy as np

from mlx_audio.audio_io import write
from mlx_audio.tts.utils import load

model = load("nativ-community/phonon")

result = next(model.generate(
    text="Hello, this is Phonon running on Apple Silicon.",
    voice="Freya",
))
write("phonon.wav", np.array(result.audio), result.sample_rate)

CLI:

python -m mlx_audio.tts.generate \
  --model nativ-community/phonon \
  --text "Hello, this is Phonon running on Apple Silicon." \
  --voice Freya \
  --play

Add --stream to the CLI command for streaming playback, or pass stream=True to model.generate() to receive audio chunks in Python.

Voices

Available voices: Freya, Harper, Sterling, and Toby. The default is Freya.