How to Restore Old Photos: A Complete Walk-Through for 2026

What Photo Restoration Actually Means in 2026

A restored photograph is not a new photograph. It is the same image, cleaned of the damage time has added, with the original detail, grain, and character intact. Done well, a viewer cannot tell it was touched. Done badly — and this is the cautionary note for 2026 — it looks plasticised, suspiciously young, or populated by faces the AI invented. The goal is conservation, not reinvention.

This guide walks end-to-end through a 2026 restoration workflow for the three formats you are most likely to hold in your hands: paper prints, colour film negatives, and slides. It covers what to do before you open any software, which tools handle which step honestly, and where the sensible boundaries sit between human judgement and AI assistance.

If you are here because you have a shoebox of family photographs and no idea where to start — you are in the right place. If you are a professional photographer looking to add a restoration service, this will anchor the workflow before you specialise. And if you have already scanned some of your collection and are stuck on the dust-cleanup step, the links through to the detailed guides will take you straight there.

The Four Kinds of Damage, and What Heals Them

Most old photographs suffer from one or more of four distinct problems. Treating each correctly — and in the right order — is what separates a careful restoration from a muddled one.

Surface dust and scratches. Small particles, fibres, and scuffs on the scan or the original. This is the single largest time-sink in restoration, and the easiest step to get right with the correct tool.
Fading and colour shift. Dyes in colour film and colour prints break down unevenly over decades. Red layers tend to last, yellow fades, cyan goes muddy. Silver prints gain a warm cast.
Physical damage. Tears, creases, missing corners, foxing (the brown spotting common to old albums), water marks, and mould. These are reconstruction jobs.
Softness and resolution limits. The original may have been a low-resolution snapshot; the scan may be undersized; the lens may have been gentle. Sharpening and careful upscaling can recover headroom.

Work them in that order. Clean the dust first, because every subsequent step magnifies it. Correct colour and tone next, because sharpening biases the histogram. Rebuild missing content third, because you need a stable underlying image to blend against. Sharpen last, because it is the only step with no upstream dependencies.

Step 0: Scan Properly (It Is Half the Job)

The quality ceiling of your restoration is set the moment the photograph leaves the scanner. A rushed 1,200 dpi JPEG gives the retoucher almost nothing to work with. A careful 2,400 dpi 16-bit TIFF gives you dynamic range and resolution to spare. This is the single most important piece of advice in the entire guide.

The rules of thumb for 2026:

Paper prints: flatbed at 600 dpi for a standard 4×6; 1,200 dpi for wallet-size or high-detail prints; 16-bit greyscale if the image is monochrome, 16-bit colour if toned or sepia.
Colour negatives (C41): 2,400–3,200 dpi on a V600 or V850; 3,600+ on a Plustek or dedicated film scanner. Turn on infrared dust removal (Digital ICE, iSRD, VueScan IR).
Colour slides (E6): same DPI range. Infrared on.
Chromogenic C41 B&W (Ilford XP2 Super, Kodak BW400CN): infrared on — these films use C41 dye chemistry and respond normally.
Traditional silver B&W (HP5, Tri-X, Delta, FP4, Foma): infrared is not an option. These films contain metallic silver which blocks IR. Physical cleaning before scanning is the workflow.
Kodachrome: infrared off. The silver content interferes; IR produces false positives.

Save every scan as a 16-bit TIFF to a master folder you never overwrite. Every restoration step is destructive at some level, and your master scan is the reference point you return to when something goes wrong three hours in. For a deeper treatment of the scanning step itself, see the companion pillar on how to scan film at home and the flatbed-specific guide on the Epson V600 and V850.

Step 1: Dust and Scratches

Dust is where restoration projects slow to a crawl. On a single 4×6 scan you might see fifty specks. On a 60-megapixel film scan, several hundred. Doing that manually with the Spot Healing Brush is hours of zoom-and-click — and this is the step the industry has actually solved, if you know where to look.

The right tool depends on what you scanned:

Colour negatives, colour slides, chromogenic C41 B&W: infrared dust removal at scan time (Digital ICE / iSRD / VueScan IR) handles most of it. Whatever remains — and there is always some — is where a dedicated tool saves you the most time. DustMagic is built for exactly this case.
Traditional silver B&W and Kodachrome: manual cleanup in Photoshop or Lightroom. Zoom to 100%, use the Spot Healing Brush in Content-Aware mode with a brush slightly larger than the defect, and work a grid pattern so you do not miss areas.
Paper prints: a light pass of the Dust & Scratches filter at a 1–2 pixel radius is often enough, followed by manual healing of anything the filter softened the detail on.

If you are working through a meaningful archive of colour film, this step alone can absorb tens of hours. The dust and scratches pillar goes much deeper into the choices here, and the bulk workflow guide is for people sitting on thousands of scans.

Step 2: Fading and Colour Correction

Colour film and colour prints fade predictably. Yellow dyes give up first, leaving the image with a pink or magenta cast. Cyan layers in prints go muddy and brown. Reds are stubborn and linger. The colour you see in a fifty-year-old slide is almost never the colour that was there when the shutter clicked.

The practical workflow in Photoshop:

Neutralise the cast first. Open a Curves adjustment layer. Use the grey-point eyedropper on something you know was neutral — a concrete wall, a white shirt, a sheet of paper. If the original has no neutral reference, use the Auto button with the “Find Dark & Light Colors” option and adjust to taste.
Rebuild the tonal range. A Levels adjustment layer, pulling the black point in to where the histogram actually starts and the white point in to where it ends, restores contrast that decades of fading have drained away.
Layer-specific colour work. Use Hue/Saturation or Selective Color to nudge individual channels. A faded blue sky is almost always desaturated and shifted toward grey-green; pulling saturation up in the blue channel and rotating hue slightly restores it without making the whole image cartoonish.

For prints and photographs where the colour has degraded beyond the reach of simple curves work, the faded photo guide covers the more invasive techniques — channel-by-channel rebuilds, frequency separation for paper texture preservation, and when to accept a tasteful black-and-white conversion instead.

Step 3: Tears, Creases, and Physical Damage

This is the step where Photoshop earns its price. Clone Stamp, Healing Brush, and Content-Aware Fill are the workhorses; the judgement about which to use where is what makes the difference between a clean repair and a blurry smudge.

A quick field guide to the tools:

Clone Stamp: when you need to copy exact texture from a source region — paper grain, skin texture, fabric pattern. Use a soft-edged brush at around 80% opacity and build up in passes.
Healing Brush: when you need texture from one area blended with the luminosity of another — small scratches, blemishes, specks. Hold Alt to sample, then paint.
Spot Healing Brush (Content-Aware): for small, isolated defects on a simple background. Do not use it on edges or detailed areas; it will smear.
Content-Aware Fill: for larger missing regions — a torn corner, a missing strip. Select the damage with a margin of a few pixels, then Edit > Content-Aware Fill. Review the suggested sample area; override if the algorithm picks up something unhelpful.
Generative Fill (sensibly): Photoshop’s 2024+ generative fill can do remarkable work on backgrounds and textures. Do not use it on faces, hands, or identifying detail — the output is plausible, not true.

Work on a duplicate layer set to Normal so you can mask back the original if a repair reveals a mistake. Never paint directly on the master scan.

Step 4: Sharpening and Final Output

Sharpening is the last step, after every other decision has been locked down. The standard technique for a scanned photograph is a two-pass approach: a mild capture sharpen at the start to compensate for the scan’s softness (usually Smart Sharpen at around 100% amount and 1–1.5 pixel radius), and an output sharpen at the end tuned to the target — print, web, or archive.

A reasonable output sharpen for web is Unsharp Mask at 80–100% amount, 0.8–1.2 pixel radius, threshold 2–4. For print, crank the radius up slightly to compensate for the dot gain. Always view at 100% while sharpening; the previews at other zoom levels lie.

Save the final in two formats: a full-resolution 16-bit TIFF alongside your master scan, and a web JPEG at 85% quality for sharing. The TIFF is the permanent record; the JPEG is disposable.

Where AI Sits in a 2026 Restoration Workflow

AI tools have crossed a real threshold in the last couple of years. Used carefully, they add genuine value. Used sloppily, they produce the plastic, uncanny results you have probably seen on social media.

Two broad categories matter in 2026:

Face enhancement and restoration. Tools like Topaz Photo AI, Remini, and various GAN-based services can rebuild detail on degraded faces that is genuinely more convincing than anything a manual retoucher can reach in reasonable time. The caveat: they invent detail. What the tool shows you is plausible, not real. For a family archive where the goal is recognisability, this is usually fine. For a forensic or historical record, it is not.
Dust, scratch, and defect removal on scanned film. This is a different problem — purely a classification and masking task over honest pixel data. There is no “invent a face” step. The output is provably the same image, minus the dust. DustMagic sits in this second category.

The practical advice: use AI face restoration at the end of the workflow, on a duplicate layer, masked only to the area that needs it, at lower opacities if the result feels too aggressive. Use AI dust removal at the start, because it makes every subsequent step easier. Treat the two as different tools for different jobs.

DIY or Pay Someone: An Honest Decision Tree

Not every restoration is worth doing yourself. The table below gives a realistic view of when to roll up your sleeves and when to hand it over.

Situation	Best route	Why
One or two prints, moderate damage	DIY in Photoshop or a free tool	The learning curve pays back within the first image.
Box of ~50 family prints, similar damage pattern	DIY with a templated workflow	Most of the time cost is figuring out the approach once; reuse it.
Hundreds of scans, heavy dust, little other damage	DIY with a batch dust tool (DustMagic) + manual finishing	The dust step is the bottleneck; automate it and the rest is manageable.
Severely damaged single image, emotional value	Professional restoration service	A skilled restorer will beat DIY on tear/crease work every time.
Large archive, mixed damage, mixed formats	Hybrid — DIY the easy ones, send the problem cases out	80% of the archive will be straightforward; the 20% is where a pro saves the image.

For a full breakdown of the service landscape — high-street chains, local photographers, AI-only SaaS products, and DIY software — see Where Can I Get Old Photos Restored?. It covers turnaround times, prices, and what each route is actually good for.

Tools and Software: What Actually Matters

The field is crowded. Here is the honest shortlist for 2026.

Adobe Photoshop: still the gold standard for single-image restoration. Non-negotiable for tear repair and fine detail work.
Adobe Lightroom: excellent for global colour correction, tonal work, and mild dust healing. Weaker than Photoshop on damage repair. Strong for managing a large restoration catalogue.
DustMagic: specialist for dust and scratches on scanned colour film and chromogenic C41 B&W. Handles the volume problem — hundreds to thousands of scans — that Photoshop cannot. Does not touch traditional silver B&W, Kodachrome, tears, creases, fading, or faces.
VueScan and SilverFast: scanner software with built-in infrared dust removal. Do the dust work at scan time where the film supports it, and you reduce the cleanup burden significantly. Our explainer on Digital ICE covers the underlying technology.
Topaz Photo AI: strong on face enhancement, upscaling, and general sharpening. Does not offer film-specific dust removal — generic denoise on film grain tends to destroy the grain you wanted to preserve.
GIMP: free Photoshop alternative. Slower workflow, steeper learning curve, but capable of everything in this guide.

For the Photoshop-specific workflow in more detail, see How to Restore Old Photos in Photoshop. For the case where Photoshop is more than you need, the faded photos guide covers lighter tools.

A Worked Example: A Faded 1978 Colour Print

To make the workflow concrete: imagine a 4×6 colour print from a family album, 1978, significant pink cast, some surface dust, a crease running diagonally across a sky.

Scan. Flatbed at 600 dpi, 16-bit colour TIFF. One scan, archived as master.
Duplicate for work. Save a working copy. Open this one in Photoshop.
Dust. Light Dust & Scratches filter at 1 pixel radius, applied to a duplicate layer masked back to only the sky and background — not faces. The dust here is paper-surface specks; it is a quick pass.
Colour. Curves layer, grey-point on a white shirt in the image. Follow with Levels to rebuild the black point. Follow with a gentle Hue/Saturation pull on the blue channel to lift the sky.
Crease. Sample just above the crease with the Clone Stamp at 60% opacity, soft brush, and paint along the line. Any remaining unevenness, hit with Healing Brush.
Faces. At 100% zoom, inspect. One uncle’s face is slightly soft. A duplicate layer through Topaz Photo AI’s Face Recovery, at 30% opacity, brings back the eyes without plasticising the skin.
Sharpen. Smart Sharpen at 80%, 1 pixel radius, noise reduction 10%.
Save. Final as 16-bit TIFF next to the master. Web JPEG at 85%.

Elapsed time: around 25 minutes for a restorer who has done this before, maybe 90 minutes for a first-timer. The per-image time drops fast across a batch; most of it is in figuring out the approach for the first one.

Common Mistakes That Ruin Restorations

Over-smoothing skin. Running the entire image through AI face enhancement at full strength erases the grain and texture that made the photograph feel photographic. Use the effect masked, at lower opacity.
Over-saturating colour. Faded photographs feel low-contrast because they are. Pulling saturation up past what the scene would plausibly have had makes the restoration look printed on a mug.
Cloning over faces. Never clone facial features. If a face has damage, heal it or rebuild with care, but do not paint over it.
Working on the master. Everyone does this once. Keep a master scan untouched. Always.
Sharpening before colour correction. Sharpening bakes noise into the channels; colour work after sharpening amplifies that noise. Correct first, sharpen last.
Ignoring the scan. Half of restoration quality lives in the scan. A bad scan cannot be saved by good software.

When You Have a Thousand Photographs, Not One

Everything above assumes a single-image workflow — careful, considered, twenty minutes a frame. If you are inheriting a full family archive, that approach becomes impractical. A thousand frames at fifteen minutes each is 250 hours. You need a different strategy: batch what can be batched, hand-finish only the hero frames, and accept that not every image needs to be museum-quality.

The archive workflow guide covers the full approach: folder structure, naming conventions, batch dust removal, global colour correction, triage logic for which frames get finished by hand. It is a different discipline to single-image restoration, and getting it right is the difference between finishing your archive and abandoning it halfway.

The Honest Summary

Restoring old photographs in 2026 is part craft, part software, and increasingly part AI judgement. The craft has not changed in twenty years: scan well, clean the dust, correct the colour, repair the damage, sharpen once, save the master. The software has. The dust-cleanup step — once the most tedious, now the most automated — no longer needs to consume the evening. The AI tools have reached the point of being useful, provided you know where their honesty limits sit.

Pick one photograph this week. Scan it properly. Walk through the five steps. Save two copies. You will have learnt more about photo restoration in an evening than any amount of reading can teach, and you will know whether the next hundred are worth your time or somebody else’s.