Historically plausible colour for family archives, with restoration and face repair in the same pass.
Colorization is a guess made from context: the model has seen enough photographs to know that skin, sky, brick and foliage occupy particular colour ranges, and it applies those priors to a greyscale image. It is not a recovery of the original colour — no such information exists in the negative — but on portraits and street scenes the result is usually indistinguishable from a period colour photograph.
ImageLabs pairs colorization with restoration, because the two problems come together: an old print needs scratches removed and faces rebuilt before colour looks convincing. CodeFormer handles most scans at 1 credit; badly damaged prints go to Nano Banana Pro.
Scan the print flat: A phone photo works, but a flatbed scan at 600 DPI gives the model far more to restore.
Enable colorize: The toggle sits next to the engine picker. Leave it off if you only want repair.
Compare and download: The before/after slider shows exactly what changed before you export.
Faces first: Face-specific restoration runs before colour, so eyes and mouths do not turn waxy.
Two price points: 1 credit for a standard scan, 5 for badly damaged prints that need generative reconstruction.
Keeps the grain honest: The goal is a restored photograph, not a plastic repaint.
Is AI colorization historically accurate? It is plausible, not documentary. Skin, sky and foliage are reliable; specific garment or paint colours are inferred.
Can it fix scratches at the same time? Yes — restoration and colorization run in the same pass.
Is it free? Signed-in accounts get 10 free credits every day, and new users can try Pro free for 24 hours — cancel anytime.