How to Restore Old Family Photos with AI: A Step-by-Step Guide

How family photos deteriorate over time

Understanding how photos deteriorate helps you know what to look for and what to fix. Black-and-white prints from the early and mid-twentieth century are among the most stable photographic materials. They are not immune to damage. Silver gelatin prints develop silver mirroring — a bluish metallic sheen on the surface that obscures shadow detail. They fade when exposed to light, losing contrast in the midtones first so that a portrait that once had rich tonal depth becomes a flat, gray image. Physical handling creates scratches, creases, and fingerprint oils that etch into the emulsion over decades.

Color prints deteriorate faster and more visibly. The three dye layers in chromogenic color prints. Cyan, magenta, and yellow — fade at different rates depending on the specific chemistry used by the manufacturer. Kodak prints from the 1970s and 1980s are notorious for shifting toward magenta as the cyan and yellow dyes fade first. Fuji prints from the same era tend to shift yellow. Polaroids develop a trait warm brown tone as all three layers fade toward their base color. The result is that a color family photo from 1985 may have accurate skin tones in one region and a heavy magenta cast in another, depending on local exposure levels.

Environmental damage adds a third layer of deterioration. Water damage creates stains, warping, and in severe cases causes the emulsion to separate from the paper base. Mold appears as fuzzy spots or filament patterns, often in white, green, or black. Adhesive residue from photo album pages, tape repairs, and scrapbook mounting creates permanent marks. And insects — silverfish mainly — eat gelatin-based photographic emulsion, leaving irregular holes and surface erosion.

• Silver gelatin B&W prints develop silver mirroring and lose midtone contrast over decades.
• Color prints shift toward magenta (Kodak) or yellow (Fuji) as dye layers fade at different rates.
• Water damage, mold, adhesive residue, and insect damage add physical destruction to chemical fading.
• Most family photos are stored in non-archival conditions that accelerate every form of deterioration.

Scanning: capturing the best possible starting point

The quality of your restoration depends fully on the quality of your scan. A good scan captures all the detail that remains in the physical print. Including detail your eyes cannot easily see in the faded original. A poor scan throws away information that even the best AI cannot recover.

Use a flatbed scanner at 600 DPI minimum, 1200 DPI for small prints like wallet-sized photos and photo-booth strips. Scan in color even for black-and-white photos. The color scan captures subtle warm and cool toning in the print, paper color, and the specific hue of stains and discoloration, all of which help the AI make better restoration decisions. Set the scanner to 16-bit if available, or at minimum use the highest quality TIFF or PNG setting rather than JPEG, which introduces compression artifacts that the AI may try to sharpen rather than ignore.

If you do not have a flatbed scanner, the National Archives recommends photographing prints in diffused natural light. Overcast daylight through a window, not direct sunshine — with the camera held perfectly parallel to the print to avoid keystoning. Place the print on a dark, non-reflective surface and weigh down curled edges with clear glass. Take multiple exposures: one normal, one slightly overexposed for shadow detail, one slightly underexposed for highlight detail. The AI can work with any of these. The best starting point is the exposure that shows the most detail in the faces.

AI enhancement: recovering lost detail and contrast

After scanning, AI Enhance is the first restoration step. Boost works on the global image traits: overall contrast, tonal range, sharpness, and detail visibility. It analyzes the faded image and reconstructs what the full tonal range likely looked like when the photo was new. Deep shadows in dark clothing and hair, full midtone detail in skin and fabrics, and clean highlights in white shirts, clouds, and bright backgrounds.

For faces — usually the most important detail in a family photo — AI boost is transformative. A faded portrait where facial features have blurred into a soft smudge regains eye detail, defined eyebrows, lip contours, and skin texture. The AI does not invent features that were not there. It amplifies the signal that still exists in the faded emulsion, making subtle tonal differences visible again. The effect is like cleaning decades of grime off a painting — the detail was always there, just hidden.

Run boost before scratch removal. Boost may reveal fine scratches and defects that were invisible in the faded original. You want to catch all of these in the cleanup pass. Boost also establishes the tonal baseline that Magic Eraser uses when reconstructing areas under scratches and stains. It needs to match the surrounding restored tone, not the original faded tone.

Removing scratches, stains, and physical damage

After boost reveals the full extent of physical damage, switch to Magic Eraser for targeted cleanup. Start with the largest defects — deep creases, water stains, mold spots, and torn or missing sections — because these affect how the AI interprets the surrounding area. Once the major damage is repaired, zoom in and work through the medium and small defects: surface scratches, fingerprint marks, adhesive residue, and dust spots.

For scratches, brush directly along the scratch line. The AI reconstructs the image detail beneath the scratch, matching the tone, texture, and content of the adjacent area. Straight scratches across a smooth background area (sky, wall, clothing) are the easiest repairs. Scratches across complex areas like faces require more care. Work in short segments and verify each repair before moving to the next.

For water stains, brush over the stain boundary where the discoloration meets the normal image area. The AI blends the tones to eliminate the visible stain edge. For severe water damage where the emulsion has physically wrinkled or separated, the AI reconstructs a plausible surface texture. The result may be slightly softer than the undamaged areas. This is the one type of damage where expert restoration may produce meaningfully better results than AI.

For missing sections — torn corners, holes from insect damage, or areas where the emulsion has flaked away fully — the AI generates new content based on context. A missing corner of a sky area is a simple fill. A missing section that includes part of a face or important detail is more challenging. The AI produces a reasonable reconstruction, but for critical subjects like faces, compare the result carefully with any other surviving photos of the same person.

• Work from largest defects to smallest — major repairs inform the AI's reconstruction of surrounding areas.
• Scratches across smooth backgrounds are easiest; scratches across faces need short, careful segments.
• Water stain removal focuses on blending the stain boundary with surrounding tones.
• Missing sections are reconstructed from context — verify AI fills against other surviving photos when faces are involved.