PhotoBoost models (FP32)
The FP32 ONNX models PhotoBoost downloads for its
High quality mode. They are exported from each upstream's own weights and checked against
PyTorch by scripts/export-and-quantize.py. Every file is listed in manifest.json with its
size and SHA-256, and the app refuses one that differs. Each model keeps its upstream license:
| Model | File | License |
|---|---|---|
| realesrgan-x4plus | this repository | BSD-3-Clause |
| realesrgan-x2plus | this repository | BSD-3-Clause |
| gfpgan-v1.4 | this repository | Apache-2.0 with third-party terms |
| scunet-color-real-psnr | this repository | Apache-2.0 |
| vitmatte-small-composition-1k | this repository | Apache-2.0 |
| rmbg-1.4 | upstream | bria-rmbg-1.4 (non-commercial) |
Non-commercial terms. GFPGAN v1.4 includes StyleGAN2-derived code under the NVIDIA Source Code License and DFDNet-derived code under CC BY-NC-SA 4.0, so treat commercial use as not cleared. RMBG-1.4 is BRIA's, licensed for non-commercial use; it is not hosted here, and the manifest points at BRIA's own repository at a pinned commit.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support