README.md

---
license: mit
tags:
  - face-aging
  - stylegan2
  - psp
  - age-manipulation
  - mirror
library_name: pytorch
---

# SAM — Face Aging (A.I.M.I Mirror)

Mirror of the **SAM (Style-based Age Manipulation)** pretrained checkpoint by Yuval Alaluf et al. (ICCV 2021), re-hosted here for stable URLs inside the A.I.M.I desktop product. Contents are unmodified — the weights in `sam_ffhq_aging.pt` are byte-identical to the official Google Drive download referenced from [github.com/yuval-alaluf/SAM](https://github.com/yuval-alaluf/SAM).

Used by A.I.M.I's image-editor face-aging feature: shift the apparent age of every detected face in a photo by -50 to +50 years while keeping identity and background intact.

## Files

| File | Size | Purpose |
|---|---|---|
| `sam_ffhq_aging.pt` | ~2.27 GB | All-in-one checkpoint: pSp encoder + StyleGAN2-FFHQ generator + VGG age regressor + training options |

## License

**MIT** — identical to the upstream release. Redistributed unchanged.

See [LICENSE](https://github.com/yuval-alaluf/SAM/blob/master/LICENSE) in the original repo for full terms. Commercial use permitted; attribution required.

## Attribution

- **Paper**: Yuval Alaluf, Or Patashnik, Daniel Cohen-Or. *"Only a Matter of Style: Age Transformation Using a Style-Based Regression Model"*. ICCV 2021. [arXiv:2102.02754](https://arxiv.org/abs/2102.02754)
- **Source code**: [github.com/yuval-alaluf/SAM](https://github.com/yuval-alaluf/SAM)
- **Builds on**: pSp encoder (Richardson et al., *"Encoding in Style"*, CVPR 2021) and StyleGAN2 (Karras et al., *"Analyzing and Improving the Image Quality of StyleGAN"*, CVPR 2020).

When using this mirror in downstream products, please cite the SAM paper above.