Colorize a black-and-white photo

To colorize a black-and-white photo, scan or photograph the print, upload it and run the colorizer: natural, period-appropriate color goes onto skin, clothing, sky and surroundings, and scratches, creases and fading are repaired in the same run. Compare it with the original, label it as colorized, and make it with any DreamBench plan to download the color photo full size.

No, they are inferred. A gray print carries no color information, so the model chooses plausible, period-appropriate colors from what the scene suggests: skin, sky, grass and wood tend to come back right, while a dress that was green may come back blue. Treat the result as a likely version of the day, not a record of it.

Yes. Sepia, brown-toned and faded prints are handled the same way as gray ones: the old tint is replaced with natural color and the fading is lifted as part of the repair. A heavily yellowed or very low-contrast print gives the model less to read, so its color will come out flatter and less certain.

Usually, yes. The instruction asks the model to keep every person, face, object and the composition, with nothing added or removed. It is a re-render by an image model rather than a tint laid over your pixels, so a badly damaged face is repaired by continuing what is there. Compare the before and after in the page before you download.

Run it again: every run is a fresh guess, so the colors can land differently. A sharper scan with more detail also helps the model read fabrics and surfaces. If the original tones matter more than color, the AI photo restoration tool has a Restore only choice that repairs the print and keeps its black-and-white or sepia tones.

Not on this page: the model picks likely colors for the whole picture, and there is nothing to set. To change one thing afterward, open the color photo in the AI image editor and say what should change, such as make her dress green. For one colored element on a black-and-white photo, colorize it here, then ask the editor to make the rest black and white.

Restoring a photo repairs damage, scratches, tears, stains and fading, and can keep the original black-and-white or sepia tones. Colorizing adds color to a black-and-white photo. This page does both in one run; the AI photo restoration tool has a Restore only choice that repairs the print and leaves it black and white.

They can try: both edit a photo from a prompt, and the result depends on how you ask. Here the instruction is written for the job, natural period color with the scratches and fading repaired in the same run, with faces and the composition kept close to the original, and the result downloads full size.

You don't need one: this page works in the browser on a phone, tablet or computer, Android or iPhone, Mac or Windows, with nothing to install, and takes a scan or a photo of the print straight from your camera roll. The color photo downloads to the phone like any picture, full size.

Yes, in other ways: Photoshop's Colorize neural filter, under Filter, Neural Filters, colors a photo and lets you set focal points to override a color, but it doesn't repair damage; MyHeritage colorizes old photos on its own site. This page repairs scratches and fading in the same run as the color, and the result can go straight into a DreamBench video.

No. Every run needs a DreamBench plan; you create an account on the way, and the button shows the price in credits before you press it. The tools come with every plan, alongside the video models, and a run that fails is refunded.

No. The page takes one photo per run, and there is no batch upload. Run it again for the next one: each run is priced separately, and the button shows the price before you press it. For an album, scan every print first, then run them in turn.

No visible one. A DreamBench tool adds no watermark to an image it returns, on any plan. The image model that adds the color embeds SynthID, Google's invisible mark that identifies a picture as made or edited by AI; it changes nothing you can see, and the download is the color photo full size.

The color photo is the image model's own size, not the scan's. The model renders at its 2K tier when the upload is over 1.5 megapixels and at its 1K tier below that, and a larger upload is scaled to 2048 pixels on its long edge first, so a scan over 1.5 megapixels gets the larger result. For a bigger print, enlarge it with the AI image upscaler.

A date or a caption is kept in place, but check it: the instruction asks the model to keep every object and the composition, and the picture is re-rendered rather than copied, so its maker says spelling and fine detail can still come out wrong. Compare the writing with the original before you share it; to take a caption off instead, use the text remover.

Yes, a group photo is colorized like any other: the instruction asks the model to keep every person, face and object in place, and skin, hair and clothing are each given a likely color. Each face is re-rendered from the pixels it has, so small faces in a large group are the ones to compare with the original. Scan the print at the highest setting you have.

A run starts at 7 credits from your DreamBench plan, depending on the photo, and once you have a plan the button shows each run's price before you press it. Starter, $9.99 a month, includes 1,100 credits a month, at least 122 runs of the Photo colorizer, plus AI video. A run that fails is refunded.

Yes: a free account and a DreamBench plan. The tools are part of every plan, from Starter at $9.99 a month.

A photo up to 40 megabytes: JPEG, PNG, WebP, HEIC or another format your browser opens. A photo straight from a phone is fine, an iPhone's HEIC included: the page scales a large picture down to 2048 pixels on its long edge before the run, so the result is never larger than the page says.

Yes. Under the DreamBench Terms you may use what a tool makes for personal and commercial purposes, including advertising, monetized content and work for clients, with no separate commercial-use upgrade. The Terms don't guarantee copyright in a result or clear another person's rights in your upload, so only use photos you have the rights to.

No. DreamBench doesn't use your prompts, uploads or videos to train or fine-tune AI models, and any future training program would be opt-in.

It stays in your browser until you make a result. Then it is used to make that result and nothing else, and removed from the tool's storage after seven days along with the result. A result you keep in your library stays as long as you keep it.