Qwen3-TTS-12Hz-1.7B-Base-8bit

This is an MLX 8-bit version of Qwen/Qwen3-TTS-12Hz-1.7B-Base, prepared for use with OpenVox and other applications built on mlx-audio.

The model was converted to MLX format using mlx-audio 0.3.0.

For architecture details, supported features, licensing information, and other model-specific information, refer to the original Qwen model card.

OpenVox

This model is provided for use with OpenVox, a native macOS application for local AI speech generation, voice cloning, audiobooks, transcription, dubbing, and other audio workflows.

OpenVox runs supported AI models locally on Apple Silicon Macs, allowing speech generation and voice cloning without requiring audio or text to be uploaded to a cloud speech service.

Use with mlx-audio

Install or update mlx-audio:

pip install -U mlx-audio

CLI Example

python -m mlx_audio.tts.generate \
  --model theoracleguy/Qwen3-TTS-12Hz-1.7B-Base-8bit \
  --text "Hello, this is a test."

Python Example

from mlx_audio.tts.utils import load_model
from mlx_audio.tts.generate import generate_audio

model = load_model(
    "theoracleguy/Qwen3-TTS-12Hz-1.7B-Base-8bit"
)

generate_audio(
    model=model,
    text="Hello, this is a test.",
    ref_audio="path_to_audio.wav",
    file_prefix="test_audio",
)

Credits

• Original model: Qwen/Qwen3-TTS-12Hz-1.7B-Base
• MLX inference framework: mlx-audio
• Packaged for use with OpenVox