AI Expertise
Open Models / DeOldify
D

DeOldify

Visit website

Description

DeOldify is an open project for colorizing and restoring old photos and film footage. It ships three models: Artistic, with the most vivid detail but more tuning; Stable, better for landscapes and portraits with fewer odd artifacts but less color; and Video, tuned for flicker-free footage and the least colorful of the three. Code and pretrained weights are under the MIT license, and Colab notebooks for images and video are linked from the page. The author says it runs on Linux, not Windows, and that there is no ready app or interface. The repository was archived by its owner in October 2024 and is read-only. The author also notes that historical color accuracy remains a challenge.

Pros and cons

Pros

  • MIT license for code and weights
  • Three models for different tasks, including one for video
  • Colab notebooks for images and video
  • Needs roughly 4 GB of video memory to colorize, as stated by the author

Cons

  • Repository is archived and no longer maintained
  • Windows is not supported
  • No ready app or interface
  • Colors are guessed and may not match history

Key facts

License
MIT
Commercial use
Allowed
Hardware requirements
About 4 GB of video memory for colorizing (stated by the author); Linux
Developer
jantic (GitHub)
Checked on
Oct 21, 2026

Editorial score

How strongly the AI Expertise editors recommend this open model for each of its tasks right now, on a 0–10 scale. The score combines output quality, capabilities, ease of launch, license openness and freshness.

Photo & video enhancement

6.6 / 10
Output quality 6.5
Low High
Features 7.0
Basic Rich
Ease of launch 5.0
Hard Easy
License openness 10.0
Restricted Open
Freshness 2.0
Outdated Fresh