| --- |
| license: openrail++ |
| base_model: layerdifforg/seethroughv0.0.2_layerdiff3d |
| base_model_relation: quantized |
| library_name: diffusers |
| --- |
| |
| # See-through: LayerDiff 3D (NF4) |
|
|
| A 4-bit NF4 quantisation of [layerdifforg/seethroughv0.0.2_layerdiff3d](https://huggingface.co/layerdifforg/seethroughv0.0.2_layerdiff3d) for GPUs with less memory. The |
| UNet and both text encoders are quantised with bitsandbytes (NF4, bfloat16 |
| compute dtype). The VAE and the transparent VAE are unquantised and identical to |
| the full model. |
|
|
| It is used by `inference/scripts/inference_psd_quantized.py` (the default |
| `--quant_mode nf4`), which needs `requirements-inference-bnb.txt`. Read our |
| [GitHub repository](https://github.com/shitagaki-lab/see-through) for usage and details. |
|
|
| ## Licence |
|
|
| The See-through code is licensed under Apache-2.0. These weights are released |
| under Apache-2.0 for our own contributions, and they also inherit the licences |
| of the models they are derived from: |
|
|
| - [Animagine XL 4.0](https://huggingface.co/cagliostrolab/animagine-xl-4.0) and [Stable Diffusion XL 1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0): CreativeML Open RAIL++-M License |
| - [LayerDiffuse](https://huggingface.co/lllyasviel/LayerDiffuse_Diffusers): CreativeML Open RAIL-M License |
| - [SDXL-VAE-FP16-Fix](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix) (the VAE): MIT License |
|
|
| Commercial use is permitted. The use-based restrictions in paragraph 5 and |
| Attachment A of the Open RAIL licences apply to every use of these weights. If |
| you distribute the weights or a derivative of them, or host them as a service, |
| you must include those restrictions as an enforceable provision in the terms |
| that govern that use, and tell your users about them. |
|
|
| The `license` field above reads `openrail++` because that licence sets the |
| conditions of use; our Apache-2.0 grant applies on top of it. See |
| [LICENSE](LICENSE) for the full terms and [NOTICE](NOTICE) for attributions and |
| the changes we made. |
|
|
| This quantised copy carries the same licence terms as the full model. |
|
|
| ## Citation |
|
|
| If you find this work useful, please cite: |
|
|
| ```bibtex |
| @inproceedings{lin2026seethrough, |
| author={Lin, Jian and Li, Chengze and Qin, Haoyun and Chan, Kwun Wang and Jin, Yanghua and Liu, Hanyuan and Choy, Stephen Chun Wang and Liu, Xueting}, |
| title={See-through: Single-image Layer Decomposition for Anime Characters}, |
| booktitle={Proceedings of the Special Interest Group on Computer Graphics and Interactive Techniques Conference Conference Papers}, |
| series={SIGGRAPH Conference Papers '26}, |
| publisher={Association for Computing Machinery}, |
| address={New York, NY, USA}, |
| year={2026}, |
| pages={1--11}, |
| doi={10.1145/3799902.3811209}, |
| url={https://doi.org/10.1145/3799902.3811209} |
| } |
| ``` |
|
|
