Old family photos usually fail in predictable ways: faded contrast, yellow color casts, scratches, dust, torn corners, water stains, and blurry scans made from a phone at an angle. After reading this, you should be able to make a clean digital copy, repair obvious damage with free tools, export a version that looks good on screen or in print, and avoid edits that permanently ruin detail.
Start with the best possible digital copy
Restoration is much easier if your starting file is sharp, evenly lit, and large enough. A weak scan forces you to “repair” problems that were created during digitizing, not problems in the original photo.
If you have a flatbed scanner, use it. Clean the scanner glass with a microfiber cloth first, then place the photo squarely against one edge. Scan at 600 DPI for small snapshots, especially anything around 3×5 inches or smaller. For larger prints, 300 DPI is usually enough for basic sharing and reprinting at the same size, but I still use 600 DPI if there are faces, handwriting, or fine texture I may need to repair.
Save the first scan as TIFF or PNG if your scanner software allows it. These formats preserve more editing room than JPG. JPG is fine for the final shareable copy, but it is not ideal as your master file because each save can add compression artifacts around faces, edges, and text.
If you only have a phone, you can still get a usable result. Put the photo on a flat surface near a window, but not in direct sun. Turn off flash. Hold the phone parallel to the photo, not tilted. If your camera app has a grid, enable it and line the edges of the photo with the grid. Take several shots, then choose the sharpest one at full size.
For glossy prints, reflections are the main problem. Move the photo slightly away from the window and place a white sheet of paper opposite the light source to soften shadows. If glare still appears, rotate the photo 90 degrees and shoot again; glare often shifts away from faces.
Before any repair work, make a folder with three versions:
That small habit saves you when an AI tool over-smooths a face or a manual repair goes too far.
Crop, straighten, and correct exposure before repairing damage
Do the simple structural edits first. If a scan is crooked or includes the scanner lid, desk, or phone shadow, repair tools may misread those areas as part of the photo.
Start by cropping out the scanner bed, white border, table surface, and any dark edge. If you need a quick browser-based option, use Crop Image to remove unwanted borders before doing heavier restoration. Keep a little margin if the original photo has important handwritten notes or a decorative edge, but remove blank background that only distracts from the subject.
Straighten next. Use the horizon, window frames, floorboards, or the vertical edge of a doorway as your guide. For portraits with no clear background lines, use the eyes as a reference. A tiny rotation, such as 0.5 to 2 degrees, often fixes a scanned photo that feels “off.”
Then adjust the basic tones. In free editors such as GIMP, Photopea, or your built-in Photos app, look for Levels, Curves, or Light controls.
A practical starting point:
Do not drag contrast to the maximum. Old photos often contain faint details in the shadows, and aggressive contrast can erase hair, eyes, fabric texture, or background clues.
For faded color photos, try white balance before saturation. Look for a neutral area: a white shirt, gray pavement, black suit, or paper border. Use that as the correction target if your editor has an eyedropper. If not, reduce yellow or red warmth manually. A common mistake is pushing saturation too high after removing a yellow cast. Skin becomes orange and old film grain becomes colored speckle. Increase saturation gently, usually by small increments, then zoom out and check whether the photo still looks believable.
Remove dust, scratches, tears, and stains carefully
After the image is straight, cropped, and tonally balanced, repair damage. Work zoomed in, but constantly zoom out to check whether the repair blends with the whole image.
For free desktop editing, GIMP is a strong option. In GIMP, open the working copy and duplicate the layer before retouching. Use the Heal Tool for dust, small scratches, and skin blemishes. Use a soft brush slightly larger than the mark you want to remove. For tiny dust spots, a brush between 5 and 20 pixels may be enough on a 600 DPI scan. For longer scratches, sample from a nearby clean area with similar brightness and texture.
Use Clone Tool when the Heal Tool creates mushy or smeared results. Clone is better near hard edges: eyeglasses, collars, furniture, picture frames, and text. Heal blends automatically, which is useful on skin or sky, but it can blur a sharp boundary. Clone copies pixels directly, so you control the structure.
A good repair rhythm is:
Torn corners need different handling. If the missing area is plain background, clone nearby background texture into the gap. If it cuts into a person’s clothing, copy from the other side only if the clothing is symmetrical enough to make sense. Avoid inventing facial features unless you are making a creative reconstruction and clearly label it that way. A restored family archive should not silently change a person’s appearance.
Water stains are difficult because they often change both color and brightness. Select the stained area softly, feather the edge, and adjust color temperature or levels only inside that selection. If the stain crosses a face, correct it in small sections: forehead, cheek, shirt, background. One broad adjustment usually leaves a visible patch.
For cracks across faces, zoom in to 100% or 200% and repair in short strokes. Do not drag the healing brush across an entire cheek or forehead. Short strokes preserve pores, wrinkles, and film grain. If the face starts looking plastic, undo and use a smaller brush with more samples from nearby areas.
Use AI repair tools, but do not let them rewrite the photo
Free AI restoration tools can be helpful for sharpening faces, reducing scratches, and colorizing black-and-white images. The risk is that they may invent details: different eyes, smoother skin, fake teeth, changed hairlines, or clothing textures that were never present.
Use AI as one step, not the whole workflow. Upload a cropped, straightened, reasonably exposed version. If the tool offers strength settings, start low or medium. Avoid “HD face” or “beauty” settings for archival work unless you only need a casual social media version.
Compare the AI output to your original at the same zoom level. Look closely at:
If the AI version improves the overall image but changes a face, blend it manually. Put the AI result as a layer above your manually edited version, then mask in only the useful areas: maybe the background cleanup, a cleaner shirt, or reduced scratches. Keep the original face if the AI face looks too modern or too smooth.
Colorization deserves extra caution. A black-and-white wedding photo may look charming in color, but the colors are guesses. Save colorized versions as separate files with names like `grandma-wedding-colorized.jpg`. Keep the black-and-white restored version too. For family records, the faithful restoration is usually more valuable than the imagined color version.
AI sharpening can also exaggerate film grain and scanner dust. If a tool makes edges crunchy, reduce sharpening later or choose the less enhanced version. For portraits, natural softness is better than a harsh artificial outline around eyelashes and hair.
Export the right version for sharing, printing, and archiving
Once the photo looks clean, export different versions for different uses. One file rarely fits every purpose.
For archiving, keep a high-quality master:
For email, messaging, or online sharing:
For printing:
If you scanned a tiny photo at 600 DPI, you may be able to print it slightly larger. If you photographed it with a phone and the file is soft, making it larger will not create real detail. It may still look fine from a distance, but faces will not become sharper just because the file dimensions are bigger.
Add metadata to the filename if you do not have a separate archive system. Useful details include year, location, names, and whether the image was colorized or AI-enhanced. Example: `1946-maria-luis-back-porch-restored-bw.png`. Avoid names like `scan_final_final2.jpg`; they become useless later.
Keep the original scan in the same folder. Future tools may repair damage better, and your taste may change. A restoration that looks impressive today may look over-smoothed later.
Common problems and how to fix them
If the restored photo looks waxy, you probably used too much noise reduction, AI face enhancement, or healing on skin. Go back to the previous version and reduce the effect. Leave some grain. Old photos should not look like phone selfies.
If scratches keep reappearing after export, check whether you repaired a low-resolution preview instead of the full image. Some phone apps and cloud viewers display compressed copies. Download or export the full-size file before editing.
If the image looks gray and flat, set proper black and white points with Levels. Old paper haze often makes blacks look milky. Move the black point slightly inward, but stop before dark suits and hair turn into solid blobs.
If faces turn orange, reduce saturation and adjust white balance first. Many faded prints have yellowed paper, not naturally warm skin. Correct the cast before adding color intensity.
If the crop cuts too close, leave more breathing room. For printing, borders may be trimmed slightly by the lab or printer. Keep important heads, feet, handwriting, and edges away from the very outside of the image.
If a repair patch looks obvious, sample from closer areas. Texture changes across a photo. Skin from a forehead will not match a cheek shadow, and wallpaper from the left side may not match the right if lighting falls off.
If the file becomes too large to send, create a separate JPG copy rather than compressing your master. Keep the restored master untouched, then make a smaller share version at 1600 or 2000 pixels on the long edge.
A good free restoration workflow is simple: make a high-quality scan, crop and straighten it, correct tone and color, repair damage by hand, use AI only where it genuinely helps, then export separate archive and sharing files. Start by cleaning up the edges with the BestAIFinds Crop Image tool, then keep your original scan safe so every edit remains reversible.