OldPhotosLab

AI photo colorizer

AI Photo Colorizer: Colorize Black and White Photos Online

Upload a black-and-white photo and get a colour version in seconds — realistic skin tones, clothing and backgrounds, with controls for what to emphasize and how vintage the palette should look.

Image Upload

Prompt0/500

Style

Focus

Tone

Try these examples

Aspect Ratio

Image Resolution

Model
GPT Image 2.5 Flare

10 welcome credits · No watermark · JPG, PNG or WEBP up to 5 MB

Your photos stay in your account and are never used to train models.See how we handle your files →

Colorize Photos in Three Steps

Sign in once, then upload and go. The editor is the same tool that powers the rest of the site.

  1. 01

    Upload your photo

    JPG, PNG or WEBP up to 5 MB. A colour scan of a black-and-white print keeps the paper grain, which usually gives a better result than a greyscale scan.

  2. 02

    Choose focus and tone

    Emphasize people, scenery or the whole frame, then pick natural, vintage or vivid. The model colours faces, clothing and background separately before blending them.

  3. 03

    Review and download

    If the result is too strong or too muted, change the tone and run it again. The output has no watermark, and it is saved in your account.

Before and After

Illustrative examples of the kind of result the model produces. Drag the handle to compare.

Illustrative original — Illustrative — 1900s studio portrait: sepia original, colorized skin and draperyIllustrative result — Illustrative — 1900s studio portrait: sepia original, colorized skin and draperyBeforeAfter
Illustrative — 1900s studio portrait: sepia original, colorized skin and drapery
Illustrative original — Illustrative — 1920s wedding party: bridal ivory and period fabric tonesIllustrative result — Illustrative — 1920s wedding party: bridal ivory and period fabric tonesBeforeAfter
Illustrative — 1920s wedding party: bridal ivory and period fabric tones
Illustrative original — Illustrative — 1940s military portrait: olive-drab uniform and outdoor lightIllustrative result — Illustrative — 1940s military portrait: olive-drab uniform and outdoor lightBeforeAfter
Illustrative — 1940s military portrait: olive-drab uniform and outdoor light
Illustrative original — Illustrative — 1950s family snapshot: pastel Kodachrome-era clothingIllustrative result — Illustrative — 1950s family snapshot: pastel Kodachrome-era clothingBeforeAfter
Illustrative — 1950s family snapshot: pastel Kodachrome-era clothing
Illustrative original — Illustrative — 1960s school class: mixed skin tones in one frameIllustrative result — Illustrative — 1960s school class: mixed skin tones in one frameBeforeAfter
Illustrative — 1960s school class: mixed skin tones in one frame
Illustrative original — Illustrative — 1970s wedding: warm browns and orange decorIllustrative result — Illustrative — 1970s wedding: warm browns and orange decorBeforeAfter
Illustrative — 1970s wedding: warm browns and orange decor
Illustrative original — Illustrative — 1980s ID photo: flat studio light, accurate skinIllustrative result — Illustrative — 1980s ID photo: flat studio light, accurate skinBeforeAfter
Illustrative — 1980s ID photo: flat studio light, accurate skin
Illustrative original — Illustrative — faded 1990s colour print: saturation restored, no hue shiftIllustrative result — Illustrative — faded 1990s colour print: saturation restored, no hue shiftBeforeAfter
Illustrative — faded 1990s colour print: saturation restored, no hue shift
Illustrative original — Illustrative — archival portrait with creases: damage repaired, detail recoveredIllustrative result — Illustrative — archival portrait with creases: damage repaired, detail recoveredBeforeAfter
Illustrative — archival portrait with creases: damage repaired, detail recovered

How does the colorizer choose colors?

Automatic colorization is not a single global filter. Different parts of the frame are estimated separately, then reconciled.

Skin is estimated in regions

A face is not one flat tone. Forehead, cheeks, nose, chin and the shadow under the jaw are estimated separately and blended. That is why faces do not come out looking like plastic, and why freckles and slight redness survive instead of being averaged away.

Clothing and period

Colour is partly a historical question. The model is trained on dated archival photographs, so it leans toward palettes that fit the era: olive and khaki uniforms, pastel 1950s dresses, warm browns and oranges for 1970s interiors. It is a strong prior, not a guarantee — verify anything documented.

Sky, foliage and indoor light

Backgrounds are coloured from brightness gradients and texture rather than a fixed palette. Overcast skies stay grey-blue, tungsten interiors stay warm, and grass does not turn neon.

You can steer the result

Focus controls what the model prioritizes (people, scenery or the whole frame); tone controls how saturated and vintage the palette is. Re-running with a different combination is the practical way to correct a result you do not like.

Colorize Black and White Photos of Every Kind

The model behaves differently on faces, snapshots, weddings and archival material.

Portraits and family faces

Frontal faces colorize best. Keep hands away from the chin and avoid heavy shadows across one side of the face.

Family snapshots and albums

Snapshots from the same roll and decade tend to receive similar palettes, but consistency is not guaranteed. Review each photo before printing an album as a set.

Wedding photographs

Bridal white, bouquet flowers and skin are estimated in separate passes so the dress does not pick up a pink cast from the faces.

Historical and archival photos

Uniforms, signage and buildings are where automatic colour is least reliable. Treat the result as a starting point and use the tone controls when you have a documented reference.

Can you colorize faded or damaged photos?

A faded colour print is a different problem from a black-and-white scan. A 1980s print that lost its magenta layer still contains hue information, so the job is to restore saturation and correct the cast rather than invent colours from scratch. Pure black-and-white and sepia scans go through full colorization instead.

AI Colorization vs Manual Colorization

Both can produce a good result. They differ in cost, speed and how they handle historical specifics.

AI colorizer (this tool)Manual / studio service
Time per photoSeconds20 minutes to 2 hours
Cost10 welcome credits, then plans from $9.99$5–$50 per photo
Historical accuracyStrong on skin, clothing and sky; weaker on uniforms and signageResearched case by case
CorrectionsRe-run with a different focus or toneIterative, by email
Best forFamily albums, gifts, social postsPublication, museum and genealogy research

Tips for Better Results

  1. Scan at 300–600 dpi in colour mode, even for black-and-white originals — a colour scan keeps the paper grain, which helps the model.
  2. Repair damage first. Scratches read as edges and can pull colour into the wrong place.
  3. Keep faces frontal and unobstructed. A profile with a hand across the cheek is the most common failure case.
  4. Use natural tone for family photos and vintage for historical scenes; vivid works best for landscapes and posters.
  5. For an album, review each photo before printing. Automatic palettes can drift between photos from different rolls.

Frequently Asked Questions

How accurate is the AI colorization?

Skin, clothing and sky are usually close. Uniforms, signage and specific product colours are the unreliable cases; treat those as a starting point and adjust with the focus and tone controls.

Can I fix a colour the AI got wrong?

Yes, by re-running the photo with a different focus or tone. Region-by-region colour painting is not available in this version, so a correction applies to the whole image.

Does it work on faded colour photos, not just black and white?

Yes. Faded prints are re-saturated and corrected rather than invented, so the result keeps the photo’s original look.

Why did my photo come out with the wrong skin tone, or too vivid?

The usual causes are a small or partly shadowed face, a strong colour cast in the scan, or a sepia original that already shifted the brightness distribution. Try natural tone first; if the whole image is too strong, lower the tone and re-run.

Does colorizing reduce the resolution or quality of the photo?

Colour is added on top of the original luminance, and the output follows the resolution you select. A low-resolution scan does not become sharper by being colorized.

I have a whole album — will the colors match across photos?

Photos from the same roll and decade tend to get similar palettes, but consistency is not guaranteed. Review each result before printing a set; do not assume one batch will match perfectly.

Which photos colorize worst?

Profiles and faces partly covered by a hand, uniforms and signage you can verify from a record, and heavily damaged prints. Those are also the cases where a wrong colour is most noticeable.

Related tools

Related guides

See exactly what each plan includes, or buy a one-time credit pack.

Compare plans and credit packs