Image Colorizer
Add realistic color to black & white, grayscale, and vintage photos with AI. The DDColor model predicts plausible colors for faces, landscapes, and everyday scenes. Run it free to preview the result; download the clean full-resolution image with credits.
Drag & drop a file here, Ctrl+V to paste, oror
Free preview HD download 1 credit Pricing
AI processing on our server
AI colorization engine
Uses DDColor (ICCV 2023), a deep learning model with two decoders: one reads the scene, one picks colours for it. It predicts plausible colours for grayscale input in about a second.
One upload, no settings
Drop the photo and the model picks the colours. Compare it with the original using the slider, then keep it or try another photo.
Private & Secure
Your photo is sent to our AI server, colorized there, and deleted immediately after processing. It is never shared or used for anything else. The clean full-resolution result is held only briefly so you can download it, then removed.
How it works
How AI photo colorization works
This tool uses DDColor, an open-source model from Alibaba DAMO Academy (ICCV 2023). It analyses a grayscale image and predicts plausible colors for every pixel — skin tones, sky gradients, foliage, fabrics, metal, wood. One decoder reads the scene and a second one picks colors for it, which keeps colors from bleeding across edges. It learned from ImageNet photos, so it knows what colors real-world objects tend to be. The model needs about a second on our AI server, and the original photo is deleted immediately. Historical accuracy is a known challenge — the AI guesses likely colors, not what was actually there. To inspect a photo's metadata before colorizing, use the EXIF Checker.
When the colors come out wrong
Skies, grass and skin are the easy cases: they have typical colors and the model gets them right most of the time. Clothing, cars and painted walls have no typical color, so the model guesses, and the guess can be bold and wrong. Fur and skin can also come out warmer than they were. Treat the result as a plausible version of the photo, not a record of it. If the colorized result needs resizing afterward, try the Image Resizer or Social Resizer.
Colorize vs restore — what's the difference
Colorization adds color to grayscale images — it does not fix blur, damage, or low resolution. If your old photo is both black-and-white and blurry, colorize first, then run the result through the Photo Unblur to recover edge detail. Scanly runs no face model, so faces are never reconstructed. If it only needs to be made bigger, the Image Upscaler (Real-ESRGAN) is the better fit. For background removal from the colorized portrait, use the Background Remover.
Common use cases for image colorization
Bringing old family photos to life for framing, gifting, or sharing online. Colorizing historical photographs for articles, documentaries, and school projects. Adding color to scanned negatives or archive prints. Enhancing black-and-white portraits for genealogy records. Previewing how a grayscale illustration would look in color. After colorizing, optimize file size with the Image Compressor, convert formats with the Format Converter, or check metadata with the EXIF Remover. Explore all 161 image tools.
Frequently Asked Questions
Is this image colorizer free?
What is DDColor?
Why do some colours look too strong?
How long does it take?
Are the colors historically accurate?
Can I colorize a photo and then sharpen it?
Are my images stored on the server?
Scanly is an independent product. DDColor is a third-party model we call through Replicate — an integration, not our brand. Scanly is not affiliated with, endorsed by or sponsored by Replicate or the model authors, and all model names and trademarks remain the property of their respective owners. AI Disclosure.