AI photo restoration

Old photo. A restored interpretation.

AI photo restoration uses a reference edit to address scratches, faded tones and blur in an old photo, with optional colorization. Upload an authorized scan or photo, choose GPT image 2 in the existing image tool and describe the damage; a reference edit costs 15 credits. AI redraws missing details rather than recovering a verified original. Faces, clothing and background details can differ from the source, so keep the original and review the edit carefully.

Upload only photos you have the rights to use and edit, with consent from people pictured and guardian permission for children where appropriate. This independent website is not affiliated with or endorsed by model vendors. Restored faces and inferred colors can differ from the original; do not use the output as historical, identity or forensic evidence.

An authorized old-photo scan plus a repair instruction
15 credits per reference edit using GPT image 2
Scratch and fading edits, with colorization only when requested
How to restore an old photo

Request repairs while keeping the source

Work from a copy of your best available scan. Separate physical damage from intentional features such as grain, monochrome tones and the subject’s age.

1. Prepare a photo you have rights to use

Choose a clear scan or evenly lit photograph of the print. Keep the original unchanged. GPT image 2 accepts JPEG, PNG or WebP references up to 30 MB. Confirm your rights and any necessary consent before uploading private family photos.

2. Describe damage and optional colorization

Open the generator, select GPT image 2, upload the photo and paste a repair instruction. Specify scratches, faded contrast or blur. Ask to retain faces, age, clothing and composition. Request colorization explicitly if wanted; otherwise ask to preserve the original monochrome tones.

3. Check the quote and inspect redrawn details

Sign in and check the displayed quote: 15 credits for a reference edit with this model. Compare eyes, mouth, hands, clothing and background with the source before saving. Keep the restored interpretation alongside the original rather than replacing your archive.

Photo restoration prompt examples

Repair visible damage with a specific instruction

These are untested prompt suggestions, not measured restoration results. Change them to match your own photo. A plausible detail is not proof of what the original scene contained.

Scratches on a monochrome family photo

Edit my authorized old family photo to reduce visible scratches and dust marks. Keep the monochrome tones, natural grain, facial proportions, apparent ages, clothing and composition as close to the source as possible. Do not beautify faces, remove wrinkles or invent jewelry. Leave uncertain details restrained instead of adding sharp invented features.

Faded print with soft detail

Restore the tonal balance of my faded photo and gently improve visible soft edges. Preserve the original palette, expressions, hairlines and background arrangement as closely as possible. Avoid aggressive smoothing or invented text. Do not change the age of anyone pictured. Keep a natural photographic texture.

Optional colorization of an old portrait

Create a colorized interpretation of my authorized monochrome portrait. Use restrained, plausible colors and preserve facial shape, expression, age, clothing and composition as closely as possible. Do not add objects or modernize the styling. Colors are inferred artistic choices rather than recovered historical facts.

Restoration limits

Keep repair, enhancement and age edits distinct

Choose the workflow according to the change you want. Restoration addresses visible damage; enhancement focuses on clarity or enlargement, while age progression deliberately changes apparent age.

Inspect faces and missing detail

AI can redraw faces, erase distinguishing marks or invent background details. Compare the output with other authorized references if available. A sharper face can be less faithful to the person originally photographed.

Treat colors as an interpretation

Colorization cannot establish the true color of clothing, skin, objects or surroundings. Preserve a monochrome version when historical fidelity matters and label colorized copies as interpretations.

Preserve the original archive

Save a separate edited copy and retain your source scan. This workflow does not repair the physical print or authenticate identities, dates or events. Keep private family material within the permissions you have.

FAQ

AI photo restoration questions

Can I request repairs for scratches, fading and blur?

Yes. Upload an authorized old photo with GPT image 2 and describe the damage to address. Repair quality is not guaranteed; inspect the result for invented details or changes to faces and clothing.

Can I colorize an old black-and-white photo?

Yes. Ask for colorization explicitly. The model infers plausible colors; it cannot verify historical skin, clothing or scene colors. Keep the original and label the colorized version as an interpretation.

Will AI recover the exact original face?

No. AI redraws missing or blurred details, and faces can differ from the original person. A clearer image does not establish identity or historical accuracy.

Is restoration the same as upscaling or age progression?

No. Restoration targets damage and faded tones. Image quality enhancement focuses on clarity or enlargement, while age progression intentionally changes apparent age. Ask restoration to preserve age and inspect whether it did.

Can I upload any old photo I find online?

No. Upload only photos you have the rights and permission to use and edit. A photo being old or publicly viewable does not establish those rights.

Is photo editing free?

No. This guide does not promise free generation. A reference edit with GPT image 2 costs 15 credits. Sign in and check your balance and the displayed quote before submitting. Each retry is a new request.

Does the button automatically set up the editor?

No. It opens the existing image generator and does not prefill a prompt, select a model, upload a photo or submit a job. Select the model and add your own reference and instruction manually.