LEARNING HUB
PREDICTIVE TREND INSIGHT
Best free apps to colorize and repair old smartphone photos Illustration

Best free apps to colorize and repair old smartphone photos

Written by Nikhil Badjatya
Direct Summary:

DeOldify is the best-known free, open-source tool specifically for colorizing black-and-white photos, available as a free Google Colab notebook with no installation required. Its important limitation: DeOldify colorizes and does not repair scratches or physical damage — the AI is guessing plausible colors based on context (it doesn't know Grandpa's suit was actually charcoal grey, not navy blue), and damage repair is a separate task requiring different tools.

"Whether you think you can or you think you can't, you're right."

— Henry Ford

Key Insights

  • Colorization and damage repair are two separate problems: DeOldify adds plausible color but doesn't fix scratches, tears, or fading — treat these as two different tasks even if you want both done to the same photo.
  • Colorization is a plausible guess, not a historical fact: the AI infers likely colors from context and training data — it has no way of knowing the actual, true color of a specific object in a specific photo.
  • No installation needed for casual use: DeOldify's Google Colab notebook runs the model in your browser for free, which is the easiest way to try it without setting up Python locally.

"Colorize and repair" is really two distinct restoration tasks that happen to often apply to the same old photo. Colorization (adding plausible color to black-and-white images) and repair (removing scratches, tears, and damage) use different techniques, and DeOldify — the most established free, open-source tool in this space — explicitly handles only the first.

Doing colorization and repair as separate steps

1. Colorize first using DeOldify's free Colab notebook. No local installation is required — upload your photo to the Colab notebook in a browser and it applies AI-based colorization at no cost.

2. Treat the resulting colors as a plausible guess, not ground truth. If historical accuracy matters (a specific uniform color, a known object), the AI's guess may be wrong — it's inferring from patterns in its training data, not from any actual knowledge of your specific photo's history.

3. Handle scratch/damage repair as a separate editing step. For physical damage (scratches, tears, missing sections), you'll need a different tool or technique — this isn't something colorization models address, and combining the two into "one free app" oversells what any single tool actually does.

deoldify_workflow.md 
# Free colorization workflow using DeOldify's Colab notebook
# 1. Open the official DeOldify Colab notebook (no install needed)
#    https://colab.research.google.com/github/jantic/DeOldify/blob/master/ImageColorizerColab.ipynb
# 2. Upload your black-and-white photo
# 3. Choose a model: "stable" or "artistic" (video model is for footage)
# 4. Run the notebook cells to generate the colorized output
#
# Note: this step does NOT repair scratches or damage --
# that requires separate photo-editing/inpainting tools
Task Does DeOldify Handle This?
Adding plausible color to a black-and-white photo Yes — its core function
Repairing scratches, tears, or physical damage No — a separate restoration task requiring different tools

Being precise about what a free tool actually does prevents disappointment: DeOldify is genuinely excellent at colorization and genuinely free, but expecting it to also fix a torn or heavily damaged photo is expecting a tool to do a job it was never built for.

Practical Challenge

Find an old black-and-white family photo, run it through DeOldify's free Colab notebook, and compare the "stable" and "artistic" model outputs to see which one you prefer for that specific image.

Concept Check

Why can't DeOldify be relied on to know the true, historically accurate color of an object in an old photo?
Correct! The model makes a statistically plausible guess based on patterns it learned — it has no ground-truth knowledge of what color a specific object in a specific photo actually was.
Incorrect. Try again! DeOldify works on any black-and-white photo and is free — the limitation is about the nature of the AI's color inference, not the photo's age or a paywall.

Sources & Further Reading

  • DeOldify (official site) — background on the project and its NoGAN colorization technique.
  • DeOldify (GitHub) — the open-source repository, including the free Colab notebook link and explicit notes on its scope/limitations.
Previous Guide Dashboard Next Guide