Quick answer
Colorize B&W adds predicted color to black-and-white photographs using the Evoke AI mobile app for Android and iOS. The colors are plausible reconstructions based on what the model has learned about skin, sky, foliage, and fabric — they are not the true original colors, which no longer exist anywhere in a monochrome file.
Are Colorized Photos Historically Accurate?
No, and this is the most important thing to understand before you colorize a family photo. A black-and-white image records brightness only. The color information was never captured, so it cannot be recovered. What the model does is guess, using patterns learned from a very large number of color photographs.
Those guesses are often convincing because the world is predictable. Skin falls in a narrow range. Grass is green, skies are blue, and wood is brown. Where the model has no reliable signal — the color of a dress, a painted wall, a car — it will pick something plausible rather than something true. If your grandmother's coat comes out navy, that is a statistical inference, not a memory.
For family keepsakes this is usually fine, and often moving. For archival, journalistic, or historical work, a colorized image should always be labelled as an interpretation.
Why Is This an App and Not a Web Tool?
Colorization uses a generative model that is too large to download into a browser tab for a single photo. Running it in the Evoke AI app lets it use your device's hardware acceleration and keeps processing time to a few seconds.
How to Colorize a Black and White Photo
- Install the Evoke AI app from Google Play or the App Store.
- Open Colorize B&W from the tools list.
- Choose a scanned print or an existing monochrome photo.
- Wait a few seconds for the model to run.
- Compare with the original, then save at full resolution.
Getting a Better Scan First
Colorization quality depends heavily on the input. A phone snapshot of a photo lying on a table, shot at an angle under a yellow bulb, gives the model bad information to work from — the color cast confuses its estimate of what is actually neutral.
- Photograph or scan the print flat and square, not at an angle.
- Use indirect daylight rather than a warm indoor bulb, or use a real scanner.
- Avoid glare and shadows falling across the print.
- Repair heavy scratches and tears first, then colorize. Damage confuses the model into painting artifacts.
- Scan at a reasonable resolution — very small scans give the model little to work with.
Where Colorization Tends to Struggle
- Clothing and painted surfaces, where there is no natural color to infer.
- Photos with heavy sepia toning, which the model may read as real color.
- Unusual lighting, such as stage lights or firelight.
- Crowds, where individual small faces may get inconsistent skin tones.
- Fine patterned fabric, which can come out muddy.
Skies, foliage, wood, and skin are the most reliable. If your photo is mostly outdoors, expect a good result.
Frequently Asked Questions
No. Colorize B&W runs only in the Evoke AI mobile app for Android and iOS, because the model is too large for a browser tab.
No. A black-and-white photo contains no color information at all. The model predicts plausible colors from patterns it learned during training. Skin, sky, and foliage are usually close. Clothing and painted objects are educated guesses.
Yes. It is free in the Evoke AI app with no watermark on the image you save.
Repair first. Scratches and creases can be misread as real features and coloured in, which makes them harder to remove afterwards.
You can, but results vary. The model may interpret the sepia tone as genuine color. Converting to neutral greyscale first usually gives a cleaner starting point.
The tool applies colorization automatically rather than offering per-object color pickers. For a specific known color, such as a military uniform, a manual editor gives you control the automatic model cannot.