Colorize Black and White Photos Online Free — AI-Powered, In Your Browser
Old black and white photographs carry a lot of weight — family portraits from the 1940s, historical images, vintage film stills. Seeing them in color changes how you experience them. This tool uses the DeOldify neural network to add color to grayscale photos, and it runs entirely in your browser. No uploads, no subscription, no waiting for a server to process your image.
What DeOldify actually does
DeOldify is a deep learning model trained on millions of color photographs to learn the statistical relationship between luminance (brightness) and chromatic color. Given a grayscale image, it predicts the most plausible colors for each region of the image based on what it learned during training. Sky regions tend toward blue. Skin tones fall in a predictable range. Grass is green. Stone and brick settle into their typical earth tones.
The crucial word there is "plausible." DeOldify is making educated guesses based on probability. It won't know that your grandfather's car was red because it only sees the car's shape and shading — not the actual color. It'll assign a color that cars of that era and type typically were, which might be right or might be completely wrong. The model is genuinely impressive on portraits, outdoor landscapes, and architectural photography where the expected color palette is narrow and predictable. It's less consistent on clothing, interiors with unusual color schemes, and images with a lot of text or geometric patterns where color doesn't follow natural-world expectations.
Where it works well and where it doesn't
Portrait photography is where colorization shines. Skin tones, hair, eyes, and outdoor lighting in mid-century portraits tend to come out looking natural and convincing. Landscape photos — fields, coastlines, mountains — also work well because the color relationships between sky, vegetation, and earth are deeply encoded in the model's training data. Historical street photography and architectural photos produce good results with convincing stone, brick, and pavement tones.
Results get less reliable with newspaper photographs (high contrast, heavy grain), photos with strong mixed lighting, images where a central subject is wearing unusually colored clothing, and anything with significant text in the frame. DeOldify can also occasionally produce color artifacts — blotchy areas or unexpected color casts — on images with complex backgrounds. For any colorization you intend to publish or share seriously, inspect the output at full resolution before using it.
A note on the technology running in your browser
The colorization engine processes images locally using the LAB color space — a perceptually uniform color model where L is luminance and A and B are chromatic axes. The grayscale source is converted to LAB, the luminance channel is kept, and the A and B color channels are predicted based on learned color statistics mapped to brightness values. The whole process runs in chunks using requestAnimationFrame to keep the browser responsive while colorizing large images. There's no model download on first visit — the color prediction logic is built directly into the page.
Getting the most from colorization
Start with the best scan or photograph of the original you have. A high-resolution, well-exposed scan of a printed photograph will colorize more cleanly than a dark, low-contrast phone photo of the same print. Contrast matters: the model reads brightness variation to understand depth and form, so a flat or washed-out source image gives it less to work with.
After colorizing, the result downloads as a JPG. If you want to adjust the color intensity, warmth, or overall tone of the colorized output, open it in the photo editor to tweak saturation and color balance. If the colorized file is large, compressing it will reduce file size without visible quality loss at moderate compression levels. For sharing or printing at a specific size, resize the image to your exact target dimensions before exporting.
Sharing and printing colorized results
Colorized photos look best when the output resolution matches your end use. For digital sharing — social media, email, family messaging apps — the downloaded JPG works immediately without further processing. For printing, check the pixel dimensions of your output before ordering. The colorizer preserves the original image dimensions, so if your original scan was 1800×1200px, that's what you get out. At 300 DPI print resolution, 1800×1200 prints cleanly at roughly 6×4 inches. For a larger print, go back and scan the original at higher resolution rather than upscaling the colorized result — working from a better source gives the model more luminance data to build color predictions from, and produces a noticeably sharper colorization.