Instructions to use tinytrashlabs/supertonic-3-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use tinytrashlabs/supertonic-3-mlx with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir supertonic-3-mlx tinytrashlabs/supertonic-3-mlx
- Supertonic
How to use tinytrashlabs/supertonic-3-mlx with Supertonic:
from supertonic import TTS tts = TTS(auto_download=True) style = tts.get_voice_style(voice_name="M1") text = "The train delay was announced at 4:45 PM on Wed, Apr 3, 2024 due to track maintenance." wav, duration = tts.synthesize(text, voice_style=style) tts.save_audio(wav, "output.wav")
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
| SuperTonic 3 (MLX conversion) — attribution notice | |
| Original model: SuperTonic 3 by Supertone Inc. | |
| https://huggingface.co/Supertone/supertonic-3 | |
| These weights were format-converted from the original ONNX exports (fp32, | |
| opset 19, `onnx/` directory of Supertone/supertonic-3) to MLX safetensors. | |
| They are MODIFIED FILES in the sense of the license: numerical values are | |
| unchanged, but the container format, tensor key names, and Conv weight axis | |
| order ([O,I,K] -> [O,K,I]) differ from the originals. The classifier-free- | |
| guidance scale (4.0) and the Euler integration step that are baked into the | |
| original vector_estimator.onnx graph are NOT baked into these weights; they | |
| are applied by the consuming runtime (mlx-audio-swift `SupertonicModel`). | |
| License: BigScience Open RAIL-M (see LICENSE), including its use-based | |
| restrictions (Attachment A). Redistribution must carry the license, this | |
| notice, and the use restrictions forward. | |