100% Free In-browser AI AI Powered

Colorize Black & White Photos

Bring old photos to life with AI colorization. Upload a B&W photo and get a realistic colorized version in seconds — all in your browser, completely private.

Smart AI colorization ready — instant processing, no downloads needed.
Drop black & white photos here

JPG, PNG, WebP • Batch processing supported

Accepts: any image
Output: .JPG
Never uploaded·How to use this tool

Colorize photos in 3 steps

1

Upload your photos

Drop one or more black & white or faded photos. The AI model loads in the background on first use.

2

AI adds color

The DeOldify AI model analyzes each photo and applies realistic, historically plausible colors entirely in your browser.

3

Download results

Preview original vs colorized side-by-side and download individual photos or all as a ZIP.

Color history with AI

Machine learning colorization for black-and-white photos. Realistic color prediction based on shape, texture, and context — no manual selection required.

100% Private
Files never uploaded
Always Free
No account needed
Instant Results
No upload wait time
No Limits
Batch process freely
Feature JustDownSize Others
Price Always free Paid plans
File uploads Never uploaded Sent to server
Daily limit Unlimited 5–20/day free
Account needed No signup Registration
Watermarks None, ever On free tier

Colorize Black-and-White Photos with AI

AI-Powered Colorization

Uses a machine learning model trained on millions of images to predict realistic colors for grayscale photos. The model assigns colors based on learned associations between shapes, textures, and common color distributions.

Automatic Color Detection

No manual selection required. The AI identifies objects, skies, skin tones, foliage, and building materials automatically and applies historically plausible colors.

Private, Browser-Based Processing

Colorization runs using the AI model loaded in your browser. Your old family photos never upload to a server.

JPG and PNG Output

Download colorized photos as JPG for web use or PNG for lossless quality. The color prediction enhances the entire image in one pass.

Side-by-Side Preview

Compare the original black-and-white photo with the colorized version before downloading, so you can assess the quality of the colorization.

Multiple Passes

Reprocess the same image to see if different colorization passes produce better results in specific areas. Useful for older photos where the AI has more uncertainty.

Who Benefits from Photo Colorization

Family History and Genealogy

Black-and-white photos from grandparents and great-grandparents gain new life when colorized. Genealogists and family historians use colorization to make old photos more relatable for younger family members who grew up with color photography.

Historical Research and Education

Educators presenting historical photographs in color help students connect emotionally with historical events. Colorized photos of historical figures, wartime scenes, and early 20th-century life are more engaging than grayscale in classroom contexts.

Photography Archival Projects

Photo restoration projects that include digitizing old prints benefit from colorization as a final step. A digitized, cleaned, and colorized family archive is significantly more impactful than a collection of grayscale scans.

Artistic and Creative Projects

Filmmakers, graphic novelists, and digital artists use colorization as a creative technique — applying unexpected colors to classic images, exploring alternate color interpretations, or creating vintage-modern composite artwork.

Frequently asked questions

This tool uses DeOldify, a deep learning model trained on millions of images to predict realistic colors for black and white photos. It runs via ONNX Runtime Web directly in your browser using WebGL acceleration. The model downloads once (~61 MB) and is cached for future visits.

DeOldify produces realistic, plausible colors based on its training data. Colors like sky blue, skin tones, and foliage are typically very accurate. For specific historic colors (exact clothing, car colors) the AI makes educated guesses — results vary by photo.

No. The AI model runs entirely in your browser. Your photos are never sent to any server. The only network request is downloading the model weights on first use.

The tool works best on true black and white or heavily desaturated photos. For color photos, the AI will attempt to enhance and re-colorize them, though results may differ from the original colors.

The first run downloads the 61 MB model (cached after). Each photo takes 5–20 seconds for AI inference depending on your device's GPU/CPU. WebGL acceleration is used when available for faster processing.

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.