Colorize photo

Colorize a black and white photo with AI, and check it against the real colors

Upload the black and white photo, ask for natural, realistic colors, and name every color you know (“her sweater is emerald green”). Without that, the model guesses: in our tests all three models turned a pink car turquoise, and a red and white awning came back green or blue. With the colors named, Nano Banana 2.1 matched the real photo on all four tests. No result came back sepia. One edit costs 2 credits on Muse Image, 9 on Nano Banana 2.1 or 11 on GPT Image 2.5.

Every edit needs a paid plan. Plans start at $19/month.

Colorize a photo in the editor See the 4 tests

How do I colorize a black and white photo?

01
Scan or photograph the print
A flat scan is best. A phone photo works if the print is flat and there is no glare. If the photo is torn or scratched, restore it first, then colorize the restored copy.
02
Ask for natural colors
“Colorize this black and white photo with natural, realistic colors.” For a brown, faded print, add “remove the brown tint and the grain”.
03
Name every color you know
Clothes, a car, a front door, a sign: “the car is pastel pink, the front door is bright blue”. Ask relatives, or look at other photos from the same day. Anything you do not name is a guess.
04
Keep the rest, then compare
End with “Keep … exactly the same” and “Do not add or remove anything”. Then check faces, skin tones and any words against the original.

4 black and white photos, colorized and checked against the real colors

We took four color photos, removed the color ourselves, and gave only the black and white copy to each model. That is how each result can sit next to the real colors. Three models got a prompt that names no colors; Nano Banana 2.1 also got the same prompt with the real colors named. The prompt card is that second prompt; the prompt behind each image is under it.

1.A vintage car: every model picked the same wrong color

The car was pastel pink, the door bright blue, the mailbox red, the tree orange. In gray, pink and light blue look the same, and all three models chose turquoise. All three also made the door dark green and the mailbox gray; only GPT Image 2.5 kept some autumn in the tree. Naming the four colors fixed every one of them.

Black and white photo of a 1950s car parked in front of a white house with a tree — Black and white photoOur color photo turned black and white. This is all the models saw.

The car colorized light turquoise, with a dark green door and a green tree — No colors namedNano Banana 2.1 · 9 creditsTurquoise car, dark green door, gray mailbox, summer-green tree.

The car colorized light turquoise, nearly the same as the other result — No colors named, other modelMuse Image · 2 creditsThe same guesses for 2 credits.

The car colorized bright turquoise with a white roof and a golden tree — No colors named, other modelGPT Image 2.5 · 11 creditsBrighter turquoise, white roof; the tree came back golden.

The car colorized pastel pink, with a bright blue door, a red mailbox and an orange tree — Colors namedNano Banana 2.1 · 9 creditsPink car, blue door, red mailbox, orange tree: all right.

The original color photo: a pastel pink car, a white house with a blue door and a red mailbox, an orange autumn tree — Real colors (no model saw this)The photo before we removed the color, for comparison.

Prompt we ran

Colorize this black and white photo with natural, realistic colors. The car is pastel pink, the front door is bright blue, the mailbox is red and the tree has orange autumn leaves. Keep everything else exactly the same: the car, the house, the door, the tree, the street and the light. Do not add or remove anything.

Suggested model: Nano Banana 2.1

Colors that were in fashion when the photo was taken are not what a model expects today. If you know them, say them.

2.A portrait: right skin, wrong sweater

Skin, silver hair, sky and sunflowers came back natural on every model. The emerald sweater did not: dark olive on Nano Banana 2.1, beige on Muse Image, blue on GPT Image 2.5. The coral scarf came back orange or coral, close enough.

Black and white photo of a fictional woman with silver hair in a sweater and scarf in a garden with sunflowers — Black and white photoOur color photo turned black and white. This is all the models saw.

The woman with a dark olive sweater and an orange scarf — No colors namedNano Banana 2.1 · 9 creditsDark olive sweater, orange-red scarf; face kept.

The woman with a beige sweater and an orange scarf — No colors named, other modelMuse Image · 2 creditsBeige sweater, orange scarf; face kept.

The woman with a blue sweater and a coral scarf — No colors named, other modelGPT Image 2.5 · 11 creditsBlue sweater, coral scarf; face kept.

The woman with an emerald green sweater and a coral red scarf — Colors namedNano Banana 2.1 · 9 creditsEmerald sweater, coral scarf, lavender and sunflowers: all right.

The original color photo: an emerald green sweater, a coral red scarf, purple lavender and yellow sunflowers — Real colors (no model saw this)The photo before we removed the color, for comparison.

Prompt we ran

Colorize this black and white photo with natural, realistic colors. Her sweater is emerald green, her scarf is coral red, the flowers behind her are purple lavender and yellow sunflowers. Keep everything else exactly the same: her face, expression, hair, clothes, the garden and the light. Do not add or remove anything.

Suggested model: Nano Banana 2.1

3.A shop front with a sign

Every model kept the sign “Café Lumière” letter for letter, accents included. None guessed the colors: the turquoise front came back green or blue, the red and white awning green or blue and white, the yellow bicycle cream, the orange chairs green or tan.

Black and white photo of a corner cafe with a striped awning, a sign reading Café Lumière, chairs and a bicycle — Black and white photoOur color photo turned black and white. This is all the models saw.

The cafe colorized with a green front, a green and white awning and a cream bicycle — No colors namedNano Banana 2.1 · 9 creditsGreen front and awning, cream bicycle, green chairs.

The cafe colorized sage green with a green and white awning — No colors named, other modelMuse Image · 2 creditsSage green front and awning, cream bicycle.

The cafe colorized blue with a blue and white awning and tan chairs — No colors named, other modelGPT Image 2.5 · 11 creditsBlue front and awning, cream bicycle, tan chairs.

The cafe with a turquoise front, a red and white awning, orange chairs and a yellow bicycle — Colors namedNano Banana 2.1 · 9 creditsTurquoise, red and white, orange, yellow: all right; sign unchanged.

The original color photo: a turquoise cafe front, a red and white striped awning, orange chairs and a yellow bicycle — Real colors (no model saw this)The photo before we removed the color, for comparison.

Prompt we ran

Colorize this black and white photo with natural, realistic colors. The shop front is turquoise, the awning has red and white stripes, the chairs are orange and the bicycle is yellow. Keep everything else exactly the same: the shop front, the sign and its words, the awning, the tables and chairs, the bicycle and the light. Do not add or remove anything.

Suggested model: Nano Banana 2.1

4.An old sepia print with grain

We gave this one a brown tint and film grain, like a faded print. Every model removed both and returned full color, not sepia. The navy floral dress came back right everywhere, because dark blue reads as dark gray. The mustard cardigan did not: beige, gray or gray-blue. GPT Image 2.5 also turned his beige trousers into blue jeans.

Sepia photo with grain of a fictional older couple on a sofa in a living room — Black and white photoOur color photo turned into a grainy sepia print. This is all the models saw.

The couple with a beige cardigan and a navy floral dress — No colors namedNano Banana 2.1 · 9 creditsTint and grain gone; beige cardigan, navy dress.

The couple with a gray cardigan and a navy floral dress in a cream room — No colors named, other modelMuse Image · 2 creditsGray cardigan, cream wallpaper.

The couple with a gray-blue cardigan and blue jeans — No colors named, other modelGPT Image 2.5 · 11 creditsGray-blue cardigan; trousers turned into jeans.

The couple with a mustard cardigan, a navy floral dress, an orange sofa and green wallpaper — Colors namedNano Banana 2.1 · 9 creditsMustard, navy, burnt orange, sage: all right; skin a little warm.

The original color photo: a mustard cardigan, a navy floral dress, a burnt orange sofa and sage green wallpaper — Real colors (no model saw this)The photo before we removed the color, for comparison.

Prompt we ran

Colorize this old sepia photo with natural, realistic colors, and remove the brown tint and the grain. His cardigan is mustard yellow, her dress is navy blue with white flowers, the sofa is burnt orange and the wallpaper is sage green. Keep everything else exactly the same: their faces, expressions, hair, clothes, the room and the light. Do not add or remove anything.

Suggested model: Nano Banana 2.1

If the print is damaged as well as faded, restore it first and colorize the restored copy, so the model is not guessing through scratches.

Check before you use it

Every color you did not name is a guess. Gray only records how light or dark a color was, so pink, light blue and pale green can all come back as the same shade.
Faces stayed the same people in our tests, and skin came back natural, sometimes a little warm. Compare eyes and teeth before you share it.
Look at clothes and objects you did not mention: a model can change more than the color, as when beige trousers came back as jeans.
Words on signs stayed exactly the same in our tests. Check them anyway.
Label it as colorized when you share a family or archive photo. The colors you did not know are the model’s choice, not a record.

Which model to use

Every model runs in the same editor and draws from the same credits. Prices are for one edit in the web app.

Model	Credits per edit	On Starter	Use it for
Muse Imageone size, about 2.5 MP	2	$0.038	Cheapest; natural colors and faces kept, but its guesses were as wrong as the others’.
Nano Banana 2.11K	9	$0.171	Our pick: with the colors named it matched all four real photos. 14 credits at 2K for a larger file.
GPT Image 2.5high quality	11	$0.209	Brightest colors; it changed beige trousers into jeans, so check clothes.

The editor shows the credit cost before you generate. Prices checked October 11, 2026.

What it costs

One edit costs 2 credits on Muse Image, 9 on Nano Banana 2.1 at 1K and 11 on GPT Image 2.5. On the Starter plan ($19 for 1,000 credits a month) that is $0.038 to $0.209 per photo.

Every edit needs a paid plan. Plans start at $19/month. Unused subscription credits roll over, downloads carry no watermark, and paid plans include a commercial license.

All model prices

Frequently asked questions

How do I colorize a black and white photo?

Upload it to the editor and write “Colorize this black and white photo with natural, realistic colors”, then name every color you know and end with “Keep everything else exactly the same. Do not add or remove anything.” Compare the result with the original before you download it.

Are AI colors accurate?

Only the ones you name. Black and white keeps how light a color was, not which color it was, so the model picks a likely one. In our tests the models got skin, sky and plants right, and got a pink car, a red awning, an emerald sweater and a mustard cardigan wrong until we named them.

Can I colorize an old sepia photo?

Yes. Add “remove the brown tint and the grain” to the prompt. On our grainy sepia print, all three models removed the tint and the grain and returned full color.

Will faces change?

In our runs every face stayed the same person, with natural skin that was sometimes a little warm. The image is redrawn, though, so compare faces closely before you share it.

Should I restore the photo first?

If it has scratches, tears or stains, yes: restore it, then colorize the restored copy. Doing both in one prompt gives the model more to guess.

How much does it cost? Is there a free option?

One edit costs 2 credits on Muse Image, 9 on Nano Banana 2.1 at 1K or 11 on GPT Image 2.5. AI Media Studio has no free tier. Every edit needs a paid plan. Plans start at $19/month.

Uploads are processed by the AI provider that runs the model you pick. Retention, processors and deletion requests are covered in the privacy policy.

How do I colorize a black and white photo?

Scan or photograph the print

Ask for natural colors

Name every color you know

Keep the rest, then compare

4 black and white photos, colorized and checked against the real colors

1.A vintage car: every model picked the same wrong color

2.A portrait: right skin, wrong sweater

3.A shop front with a sign

4.An old sepia print with grain

Check before you use it

Which model to use

What it costs

Frequently asked questions