Modelpublic

vanch007/Audio8-TTS-MLX-BF16

sourceHugging Faceapache-2.0updated 20d agoView on Hugging Face

0likes42downloads

Model Card

Audio8-TTS-MLX-BF16

Native MLX BF16 baseline of Audio8/Audio8-TTS-Preview-0.6b for Apple Silicon.

Inference code, installation, API documentation, tests, and benchmark evidence: vanch007/mlx-audio8-tts.

Artifact

—1,147 MiB language-model weights; 2.39 GiB complete repository download.
—Bundled 1.26 GiB, 44.1 kHz neural codec with 10 acoustic codebooks.
—Baseline for comparing the recommended 8-bit release.

M3 Max benchmark

Measured RTF across the 11-language real-device matrix: 1.245–1.484. The matrix also covers zero-shot voice cloning and streaming. See the evaluation report.

Usage

bash

git clone https://github.com/vanch007/mlx-audio8-tts.git
cd mlx-audio8-tts
pip install -e '.[server]'

mlx-audio8-tts generate \
  --model vanch007/Audio8-TTS-MLX-BF16 \
  --text "Hello from the MLX BF16 baseline." \
  --output output.wav

This is an independent Apache-2.0 MLX conversion. See the upstream project for the original architecture and checkpoint.