How to use these prompts
- Scan the print or photograph it flat in daylight, as sharp and as large as you can. Our scanning guide explains resolution and glare.
- Upload the image to ChatGPT, Gemini or another chatbot that edits images, and paste one prompt below.
- Start with the repair-only prompt. Add color or reconstruction in a second message so you can see what each step changed.
- Compare the result with the original at full size, eyes and mouth first. Run the prompt again if a face has changed.
Every prompt names what has to stay the same. A short request such as "restore this photo" leaves the model free to smooth skin, change the crop or modernize the colors, because a chatbot regenerates the photo rather than repairing it.
Seven prompts for old photo restoration
1. Repair only (the safest starting prompt)
Use this first on any old print. It asks for damage removal and nothing else.
Restore this old photograph. Remove scratches, dust, creases, stains and small tears. Keep everything else exactly as it is: the same people, faces, expressions, hairstyles, clothing, pose, background, framing and crop. Do not add or remove any objects or people. Do not smooth skin, change facial features or make anyone look younger. Keep the original black-and-white or sepia tone and the natural film grain. Output the full image at the same aspect ratio.
2. Heavy scratches, folds and cracks
For prints that were folded, cracked or scratched across faces or backgrounds.
This photo has fold lines, cracks and deep scratches. Remove every fold line, crack and scratch and fill those thin lines using the detail right next to them. Do not redraw areas that are undamaged. Keep each person's face shape, eyes, nose, mouth and expression identical to the original. Keep the original lighting, tone and grain. Same crop and aspect ratio.
3. Faded or yellowed prints
For washed-out, low-contrast or yellow and magenta color-shifted photos.
This photo has faded and turned yellow with age. Restore natural contrast and a neutral color balance, recover shadow and highlight detail, and remove the yellow or magenta cast. Do not repaint the scene, change any faces or add detail that is not visible in the original. Keep the period look of the photo rather than making it look like a modern digital image.
4. Repair and colorize a black-and-white photo
When you want color as well as repair. Colors will be a plausible guess, not a record.
Repair the scratches, dust and fading in this black-and-white photo, then colorize it with realistic, muted colors that suit the era of the photo. Use natural skin tones that match each person. Keep every face, expression, hairstyle, piece of clothing, the background and the crop exactly as in the original. Do not add jewelry, makeup, objects or text. Avoid oversaturated colors.
5. Protect faces in a group photo
For family and wedding groups where every face needs to stay recognizable.
Restore this group photo by removing scratches, spots and fading. The faces are the most important part: keep the exact facial features, proportions, expressions and eye direction of every person. Do not beautify, sharpen into a different face, or change anyone's age. If a face is too damaged to read, leave it soft rather than inventing new features. Keep the number of people, their positions and the crop unchanged.
6. Torn corners and missing pieces
For prints with a missing corner or edge. Reconstructed areas are always a guess.
Part of this photo is torn away. Fill the missing area by continuing the background, wall, sky or clothing that is visible next to it, keeping the same texture, lighting and perspective. Do not place new people, faces or objects in the missing area. Leave every undamaged part of the photo unchanged, and keep the original crop and aspect ratio.
Follow-up prompts when the result drifts
Send these in the same chat, after the first result, to correct the most common problems.
| Problem | Follow-up prompt |
|---|---|
| A face no longer looks like the person | The face of the person on the left has changed. Redo the restoration and keep that face exactly as in the original photo, even if it stays a little soft. |
| The photo looks too modern or plastic | This looks too smooth and modern. Keep the natural film grain and the period look of the original; only remove the damage. |
| Something was added or removed | You added or removed details. Restore again without adding or removing any objects, people, jewelry or text, and keep the original crop. |
| Colors are too strong | The colors are too saturated. Use softer, realistic colors that look like a color photo from the same decade. |
ChatGPT or Gemini: does the prompt need to change?
No. The prompts are plain instructions and work the same way in both. What differs is the model behind them, and that changes between versions, so test with your own photo rather than relying on a ranking. Two habits help in any chatbot: upload the largest copy of the scan you have, and keep each request to one job.
Limits on free image edits and how uploads are stored are set by each provider and change over time. Check the provider's site for current limits, pricing and data settings before you upload irreplaceable family photos.
What no prompt can guarantee
- Identical faces. The image is generated again, so features can shift even when the prompt forbids it.
- Lost detail. Missing areas and heavy blur are filled with a plausible guess, not recovered.
- True colors. Colorizing a black-and-white photo is an interpretation. Only family memory or records can confirm a dress or eye color.
- The same result twice. Running one prompt again can give a different picture. Save the version you like.
Restore old photos without a prompt
If wording and rerunning prompts is slowing you down, the old photo restoration studio runs one repair process on every photo. You can mark faces so their original pixels are copied back into the result, add color as a second version with colorize photos, and fix specific damage with scratch removal or photo repair. There are 2 free credits a month with a verified account and no watermark. Comparing options first? See the AI photo restoration tools compared.
Restore an old photo free