Skip to content
ReviveLens
Get Started

Legal

Third-Party Licenses

ReviveLens restores photos with open-source models, used as published with their pretrained weights. We thank their authors. Each project's original license notice is kept with the software we run.

Last updated: October 28, 2026

Models

  • Bringing Old Photos Back to Life (Microsoft): Scratch detection and damage repair. License: MIT (code and pretrained model). Source
  • GFPGAN v1.4 (TencentARC): Face restoration. License: Apache License 2.0. Source
  • Real-ESRGAN (xinntao): Detail and upscaling. License: BSD 3-Clause. Source
  • MI-GAN (Picsart AI Research): Filling stains, tears and torn edges. License: MIT (code and weights). Source
  • DDColor (piddnad): The free color version. License: Apache License 2.0 (code and weights). Source
  • BiRefNet (ZhengPeng7): Separating people from the backdrop. License: MIT (code and weights). Source
  • MediaPipe multiclass selfie segmenter and face landmarker (Google): Finding skin, hair, clothing, eyes and lips for the color version. License: Apache License 2.0 (code and model cards). Source
  • OpenCV: Image processing and inpainting of small leftovers. License: Apache License 2.0. Source

Software

The website and services are built with open-source software including Next.js and React (MIT), FastAPI (MIT), PyTorch (BSD-style) and Pillow (MIT-CMU), each under its own license.

Training data

Some of these models were trained by their authors on public research datasets whose own terms differ from the model licenses. ReviveLens does not train models and does not use customer photos for training.