Photo Restoration Ethics: When to Leave the Imperfections In
The Question No Tutorial Asks
Most photo restoration content focuses on HOW to fix things. The harder question is WHAT to fix.
A 1947 wedding portrait has a thumbprint on the bride's skirt where someone touched the print before it dried. AI inpainting removes the thumbprint cleanly. Should you?
A 1965 family beach photo has color fading concentrated on faces because the photographer used a budget Kodachrome roll. AI face restoration recovers the original skin tones convincingly. Should you?
A 1989 graduation photo has a scratch through the diploma. AI inpainting fills it. The diploma now reads cleanly. The scratch was there because the original got pulled out of the album frame so many times that the surface wore. Should you erase that?
There's no universal right answer. But there are useful frameworks.
Three Categories of Imperfection
Category 1: Damage that wasn't there originally
Examples: water stains, foxing spots, scratches from album sleeves, fingerprint marks, color cast from acid in the paper.
These are post-photographic damage. The original photo didn't have them. Removing them returns the photo closer to what the photographer intended. Almost always fix these.
Tools: face restoration for portraits, fix blurry photos for general.
Category 2: Original photo limitations
Examples: low resolution from a budget camera, soft focus, color shifts from old film stock, motion blur from a slow shutter.
These were properties of the original photograph. The photographer might have wished they weren't there but they're part of what was actually captured.
Whether to fix depends on intent:
- If the photo is for showing the family today, fixing helps. Modern viewers expect modern fidelity.
- If the photo is for historical archive, leave intact. The blur and color shift ARE the photograph as it was made.
- If the photo is for a memorial slideshow, fix lightly. Heavy AI restoration can make a 1970 photo look like a 2020 photo, which feels off.
Category 3: Period markers
Examples: sepia tone (intentional, not damage), specific film grain (Kodachrome's saturation, Tri-X's contrast), dust and scratches that evidence age, faded edges from album exposure.
These are what makes a 1955 photo look like 1955 and not like a digital photo with a vintage filter. They contribute to historical authenticity.
Preserve these. Modern AI restoration tools sometimes interpret them as damage and "correct" them, which produces photos that look like Instagram filters of period content rather than the actual era.
When using colorization tools, keep the original monochrome master separately. Some viewers prefer the unfiltered version.
The "Two Versions" Default
For any restoration project on family heirlooms, save two versions:
- Restored: cleaned up, ready for sharing, framing, slideshows
- Original scan, untouched: the photo as it actually exists today
The untouched scan is the document. The restoration is the interpretation. Future generations might prefer the document over your interpretation.
This is standard practice in archival work. Museum digitization always retains the raw scan separate from restored derivatives.
The Face Smoothing Problem
AI face restoration tools have a specific failure mode: aggressive skin smoothing that erases age markers.
A 1985 photo of an 85-year-old grandmother shows real wrinkles, real life-marked skin. Some AI tools "restore" this by replacing the wrinkles with smooth 30-year-old skin. The result looks technically polished but no longer looks like the actual person.
Three signs the AI smoothed too aggressively:
- Skin texture is uniform across the entire face
- Eye crinkles are gone or reduced
- The photo looks like a Hollywood headshot
When this happens, dial back the restoration strength or use a less aggressive tool. The goal is to recover ORIGINAL detail, not invent new skin.
The Group Photo Resurrection Problem
In a 1962 family reunion photo with 30 people, three are partially obscured: blurred, behind glare, too small. AI inpainting can "restore" their faces.
The problem: the AI doesn't know what those people looked like. It generates plausible-but-fictional faces that might or might not resemble the actual person. A relative looking at the restored photo might think "Uncle Frank's nose was bigger than that" or "that's not Cousin Mary's chin."
For partially obscured people in group photos, the ethical default is to leave them as they are or to identify them with names rather than reconstruct features. AI inpainting is appropriate for damage to known faces (a scratch through the bride's eye), not for inventing details that weren't captured.
Frequently Asked Questions
Is it wrong to colorize a black-and-white photo?
It's a creative choice, not an ethical violation. Colorization adds information that wasn't in the original. Some families love the colorized version as a way to feel closer to ancestors. Others prefer the documentary truth of monochrome. Keep both versions and let viewers choose.
What if AI hallucinates a face that's slightly wrong?
This is why face restoration AI works best on photos where the face is mostly intact and just damaged or low-resolution. For heavily obscured faces, AI invents plausible details that are not the actual person. Leave heavily obscured faces alone, or use a service shop with skilled retouchers and family input.
Should I tell the family the photo was AI-restored?
Yes. Most people care about the result, not the method, but transparency matters for trust. Mentioning "I ran this through PhotoFlip's restoration" lets the family ask for the original if they prefer it.
Can I sell restored photos?
Restoration of someone else's photograph (a hired subject, a museum piece, a copyrighted image) is a copyright question, not an ethics question. Family heirlooms you own are fine to share, gift, or sell prints of. Other people's photos require permission.
What about removing ex-spouses from family photos?
This is the family's decision, not yours. Some families want history preserved as it happened. Others want photos that match the current family configuration. Check with the closest survivor and respect their choice.
Related Reading
- Wedding Photo Restoration Checklist
- Memorial Slideshow Photo Prep
- How to Colorize Black-and-White Photos
Bottom Line
Fix damage that wasn't there originally. Decide thoughtfully on photo limitations based on your purpose. Preserve period markers. Always save both restored and untouched versions. Use PhotoFlip's face restoration and colorize tools with awareness of what they're changing, not just what they're improving.