What are the best practices for photo restoration in 2026, and how should I restore old damaged photos?

Photo restoration in 2026 sits at an interesting crossroads. The core principles of physical conservation have not changed in decades, but the digital side of the craft has been transformed by AI tools that can repair scratches, reconstruct faces, and add plausible color to black-and-white images in seconds. This guide covers the definitive best practices, from handling the original print to choosing between manual editing and AI-assisted workflows, with honest notes about where AI falls short.

Start With the Original: Physical Preservation Comes First

Also worth reading: How Can Responsible AI Image Restoration Colorize Photos Fairly? · How Do I Set Up a Local AI Restoration Pipeline for Old Photos in 2026? · How Can AI Colorization Ethical Accuracy Tools Improve Photo Restoration?

Before any scanning or software work begins, the single most important best practice is protecting the physical photograph itself. The field of photograph conservation, long studied by archives and museums, recommends storing prints in acid-free boxes, folders, and sleeves. Acid-free and lignin-free enclosures prevent chemical degradation that accelerates yellowing, brittleness, and silver mirroring on older photographic papers. Standard archival storage boxes for 4x6 and 5x7 prints cost between $15 and $40 depending on capacity, which is a trivial investment compared to the irreplaceable nature of family photographs.

Handle prints by their edges only, with clean, dry hands or cotton gloves. Oils from skin transfer to emulsion surfaces and cause staining over time. Never attempt to unstick a photo from an album page by force; tearing the emulsion is permanent, whereas professional conservators can often separate stuck photos safely. Keep prints away from direct sunlight, humidity above roughly 60 percent, and temperature swings. A stable environment around 65-70°F (18-21°C) with moderate humidity is the accepted target range in conservation literature.

One point worth being blunt about: no digital restoration can recover information that was never captured. If the original print has faded so badly that detail is gone entirely, software can only guess at what was there. That is why physical preservation remains step one even in the AI era.

Scan Properly: Resolution, Format, and Settings

The quality of your restoration is capped by the quality of your scan. For standard prints up to 8x10 inches, scan at a minimum of 600 DPI; for small photos under 3x4 inches or anything you plan to enlarge, use 1200 DPI. Higher resolution gives restoration algorithms more data to work with and lets you crop or print larger later without visible softness. Scanning at 300 DPI, a common default, is adequate for reprinting at the same size but wastes the potential of modern enhancement tools.

Save scans as lossless files: TIFF or PNG rather than JPEG. JPEG compression introduces artifacts around edges and in flat areas, and those artifacts get amplified when AI models upscale or denoise the image. A 600 DPI TIFF of a 4x6 print runs roughly 50-100 MB, which is manageable on any modern computer. If your scanner offers it, disable automatic sharpening and auto-contrast during capture; you want the rawest possible capture, then apply corrections deliberately afterward.

For photographs with surface texture, embossing, or heavy gloss, flatbed scanners generally outperform phone-camera captures because they control lighting angle and eliminate glare. Smartphone scanning apps have improved substantially and are acceptable for casual projects, but for heirloom-quality work, a flatbed scanner in the $100-250 range (Epson and Canon dominate this segment) remains the reliable choice.

Digital Cleanup: Repair Damage Before Anything Else

Once you have a clean scan, the restoration workflow follows a consistent order: repair structural damage first, correct tone second, colorize last. Attempting to colorize a photo that still has scratches, tears, and dust means the AI will interpret damage as real content and bake it into the final result.

Structural repair addresses tears, creases, missing corners, scratches, and spots. In manual editors like Photoshop, the Healing Brush, Clone Stamp, and generative fill features handle most of this. Dust and specks respond well to automated filters, but tears crossing faces or detailed backgrounds need careful manual work. A useful threshold: if damage affects fewer than 5 percent of pixels, automated tools usually handle it invisibly; beyond that, expect to spend real time retouching by hand.

Tonal correction comes next. Old prints typically suffer from fading, yellowing of the paper base, and reduced contrast. Levels and curves adjustments recover dynamic range, while targeted desaturation removes the yellow cast from aged black-and-white prints before colorization. Work non-destructively using adjustment layers so you can revise decisions later without degrading the image.

Where AI Fits: Enhancement, Face Reconstruction, and Colorization

AI has reshaped three specific tasks within restoration. First, super-resolution and denoising models upscale low-resolution scans while suppressing grain, with modern tools producing usable enlargements at 2x to 4x magnification. Second, face reconstruction models rebuild facial features from heavily damaged or tiny source photos, which is genuinely valuable for genealogy work where a face might occupy only 50x80 pixels in the original. Third, AI colorization converts monochrome images to color automatically.

Testing coverage across 2025-2026 from outlets like PerfectCorp and Fstoppers, which evaluated apps including VideoProc Converter AI and various free enhancers, shows a consistent pattern: AI excels at global tasks like denoising, sharpening, and colorizing landscapes and clothing, but struggles with identity-critical details. Reconstructed faces can look plausible while not resembling the actual person, because the model invents features based on statistical averages rather than the individual. For genealogical or documentary purposes, treat AI face reconstruction as interpretation, not recovery.

Colorization deserves similar honesty. Modern AI colorization produces convincing results perhaps 70-80 percent of the time on typical portraits, but it guesses colors it cannot know. Military uniforms, brand-specific products, eye color, and period-accurate fabric dyes are frequent errors. The best practice is to use AI colorization as a starting draft, then manually correct known colors using historical references, family recollection, or surviving objects. Services like colorizethis.io position AI colorization as an accessible entry point for this reason: it delivers a strong first pass that users can refine, rather than pretending the machine knows what great-grandmother's dress actually looked like.

Manual Editing vs. AI Tools vs. Hybrid Workflow

Choosing an approach depends on the value of the photo, the extent of damage, and your skill level. Here is how the three main options compare:

The hybrid workflow has emerged as the practical default for serious hobbyists: scan properly, run AI cleanup and colorization for the heavy lifting, then spend focused manual effort on faces, known-color corrections, and final tonal polish. Pure manual restoration still wins for museum-grade or commercially published work, while pure AI wins when you have 200 photos from a shoebox and one evening.

Common Mistakes That Ruin Restorations

The most frequent error is over-processing. Aggressive sharpening, noise reduction, and smoothing produce the waxy, plastic-skinned look that immediately signals a badly restored photo. Skin retains pores and texture in real photographs; if your result looks like a porcelain doll, dial back the enhancement by 30-50 percent. A good rule: enhance until the improvement is visible at 100 percent zoom, then stop.

The second mistake is working destructively. Flattening edits directly onto the scan, overwriting the original file, or saving repeatedly as JPEG compounds quality loss with every save. Always keep the untouched master scan archived separately, work on copies, and preserve layered files for important projects.

Third is trusting AI output uncritically. As noted, reconstructed faces may not match the person, and colorized hues are estimates. Publishing a colorized photo of a relative with invented eye color, or a reconstructed face bearing little resemblance, spreads misinformation through family archives. Verify against other photos whenever possible.

Fourth is ignoring the scan stage. No amount of downstream AI fixes a blurry, low-resolution, glare-marked capture from a phone camera hovering over glass. Garbage in, garbage out remains true regardless of how good the software is.

Finally, many people skip backing up. Store restorations in at least two locations, ideally following the 3-2-1 pattern: three copies, two different media types, one offsite or cloud-based. Hard drives fail at inconvenient times, and a restored heirloom lost to a dead drive is a double loss.

When to Restore, When to Hire a Professional, and What It Costs

Timing matters mostly in one direction: sooner is better. Photographic degradation is cumulative and irreversible. A print showing early fading today will be significantly worse in five years, so digitize vulnerable photos promptly even if you do not plan to restore them yet. Once a high-resolution scan exists, the physical print's decline stops mattering because the information is preserved.

DIY restoration makes sense when photos have moderate damage, you enjoy the process, and the volume justifies learning curve investment. Professional restorers charge anywhere from $30-50 for basic cleanup to $150-500+ for severe reconstruction involving missing sections or extensive facial work, with turnaround typically running 3-14 days. Hiring makes sense for severely damaged heirlooms, photos intended for large-format printing or publication, and cases where emotional stakes make experimentation risky.

AI-first services occupy the middle ground economically. Per-photo pricing on colorization and enhancement platforms generally ranges from free trials to $5-15 per image, with bulk credits reducing unit costs. For a batch of 50 moderately damaged photos, an AI pass plus selective professional touch-ups on the ten most important images often totals less than full professional treatment of all fifty.

Building a Repeatable Restoration Workflow

Consistency separates good results from great ones, especially across a large family archive. Establish a fixed pipeline and apply it to every photo. A proven sequence: inspect and clean the physical print with a soft brush; scan at 600-1200 DPI to TIFF; archive the raw scan; run automated dust removal; perform structural repairs; correct tone and contrast; colorize if desired, using AI for the draft; manually verify and correct colors against references; apply conservative sharpening as the final step; export a working JPEG at high quality alongside the TIFF master; back up everything in two locations.

Document your settings and any known color information as you go. Family members who can identify uniform colors, house paint, or car models will not be available forever, and recording their input alongside the file preserves accuracy for future re-restorations as tools improve. The technology will keep advancing, but a well-scanned, well-documented archive ensures every future tool upgrade can be applied to your collection again from the best possible source.