Instructions to use moonbox-dev/moss_sfx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use moonbox-dev/moss_sfx with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download moonbox-dev/moss_sfx --local-dir moss_sfx
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
MOSS-SoundEffect v2.0 for Moonbox
The files the Moonbox desktop app downloads for these models:
- MOSS-SoundEffect (OpenMOSS): Sound effects and ambiences from a description, 1 to 30 seconds.
They are the MOSS-SoundEffect v2.0 transformer and its audio decoder converted to MLX for Apple silicon, with the text encoder, tokenizer and configs it was published with. Moonbox checks every file against a signed list before using it. You do not need this repository to use Moonbox, and it is not an official release of the original models.
What is here
| Path | What | From |
|---|---|---|
models/mlx/, models/hf/MOSS-SoundEffect-v2.0/.gitattributes, models/hf/MOSS-SoundEffect-v2.0/README.md, models/hf/MOSS-SoundEffect-v2.0/model_index.json and 1 more |
MOSS-SoundEffect v2.0 DiT (MLX) + DAC VAE decoder (vae_128d_48k) + configs | https://huggingface.co/OpenMOSS-Team/MOSS-SoundEffect-v2.0 |
models/hf/MOSS-SoundEffect-v2.0/text_encoder/, models/hf/MOSS-SoundEffect-v2.0/tokenizer/ |
Qwen3-1.7B text encoder + tokenizer (byte-identical copy inside MOSS-SoundEffect) | https://huggingface.co/Qwen/Qwen3-1.7B |
models/torch/moss/ |
the transformer (bf16) and the audio decoder (fp32) exported for the Windows engine | https://huggingface.co/OpenMOSS-Team/MOSS-SoundEffect-v2.0 |
Every part was converted, quantized or repackaged from its source unless its row says "as published": treat none of these files as the original release.
Licences
| Part | Licence | Notice |
|---|---|---|
| MOSS-SoundEffect v2.0 DiT (MLX) + DAC VAE decoder (vae_128d_48k) + configs | Apache License 2.0 | Copyright 2026 OpenMOSS Team, Fudan University, SII and MOSI |
| Qwen3-1.7B text encoder + tokenizer (byte-identical copy inside MOSS-SoundEffect) | Apache License 2.0 | Copyright 2024 Alibaba Cloud |
| the transformer (bf16) and the audio decoder (fp32) exported for the Windows engine | Apache License 2.0 | Copyright 2026 OpenMOSS Team, Fudan University, SII and MOSI |
The full text of each licence is in LICENSE. The models are the work of their authors, linked above. Moonbox converted and packaged them and claims no rights in them. Use is subject to each licence above.
Quantized
Model tree for moonbox-dev/moss_sfx
Base model
Qwen/Qwen3-1.7B-Base