Instructions to use moonbox-dev/xvc with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use moonbox-dev/xvc with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download moonbox-dev/xvc --local-dir xvc
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
X-VC for Moonbox
The files the Moonbox desktop app downloads for these models:
- X-VC (X-VC authors): Turns a recording into a preset voice or one copied from a short clip, keeping the delivery.
They are the X-VC voice converter as one MLX file for Apple silicon (16-bit, and the full-precision original), with the speaker encoder it uses. 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/hf/X-VC/ |
X-VC voice converter (converter, adapters, prenet, SAC acoustic codec) | https://huggingface.co/chenxie95/X-VC (xvc.pt -> xvc_infer.safetensors); code https://github.com/Jerrister/X-VC |
models/hf/X-VC/xvc_infer.safetensors |
GLM-4-Voice tokenizer (Whisper-large-v3-derived encoder + VQ), embedded in xvc_infer.safetensors as semantic_encoder.* | https://huggingface.co/zai-org/glm-4-voice-tokenizer |
models/hf/eres2net/ |
ERes2Net speaker encoder (iic/speech_eres2net_sv_en_voxceleb_16k) | https://modelscope.cn/models/iic/speech_eres2net_sv_en_voxceleb_16k |
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
Built with glm-4. xvc_infer.safetensors contains the GLM-4-Voice tokenizer (its semantic_encoder.* weights), which is under the glm-4-voice licence, not MIT: academic use is free, commercial use needs registration with Zhipu AI (https://open.bigmodel.cn/mla/form). That is why this repository declares that licence; the full text is in LICENSE.
| Part | Licence | Notice |
|---|---|---|
| X-VC voice converter (converter, adapters, prenet, SAC acoustic codec) | MIT License | Copyright (c) 2026 Jerrister Zheng. The SAC acoustic codec inside is Apache-2.0 (Soul-AILab). |
| GLM-4-Voice tokenizer (Whisper-large-v3-derived encoder + VQ), embedded in xvc_infer.safetensors as semantic_encoder.* | glm-4-voice License (Zhipu AI) | Built with glm-4 |
| ERes2Net speaker encoder (iic/speech_eres2net_sv_en_voxceleb_16k) | Apache License 2.0 | 3D-Speaker (Alibaba DAMO Academy) |
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/xvc
Base model
chenxie95/X-VC