DDColor

Model

Open-source model that colorizes black-and-white and faded old photos

Price
Free, open source
Access
None, runs locally

About

DDColor is an Apache-2.0 image colorization model (ICCV 2023, Alibaba DAMO Academy) with PyTorch inference scripts, Hugging Face weights and ONNX export. It adds color only: it does not repair scratches, tears or blur. Replicate and fal also run it per call.

What you can do with it

  • Colorize black-and-white family or archive photos on your own GPU
  • Add natural color to faded historical photos in a batch script
  • Export the tiny model to ONNX and colorize images in your own runtime

Get started

  1. Clone the repo and install PyTorch and requirements.txt
  2. Pick a model, e.g. ddcolor_modelscope or ddcolor_paper_tiny
  3. Run scripts/infer.py on a folder of images

Example

git clone https://github.com/piddnad/DDColor && cd DDColor
pip install -r requirements.txt
python scripts/infer.py --model_name ddcolor_modelscope --input ./assets/test_images

Details

Hosting
Self-hosted, Runs locally
Available in
Worldwide
MCP server
None

Tasks

Alternatives

Other tools for the same tasks.

Last checked on .