Restore Old Photo

What photo restoration actually does

A photo scanner captures whatever physical damage exists on the print: grain from fast film, scratches from mishandling, fading from light exposure, and compression blocking from decades-old JPEG saves. Restoration uses AI models trained on large image datasets to identify those artifact patterns and replace them with plausible recovered detail — the same approach used in professional archival digitization, now available without specialist software.

For isolated noise problems, the dedicated denoise image tool gives direct control over noise strength. For soft scans or slightly blurry prints, sharpen image targets edge clarity. For very small original dimensions — common with older prints scanned at low DPI — upscale image can increase resolution before the restoration pass so the model has more pixels to work with.

Restoration outcomes at a glance

Grain and noise

Film grain, scanner noise, and high-ISO artifacts reduced without smearing fine detail

Scratches and dust

Small linear scratches and dust spots filled using surrounding texture — larger tears need manual work

Faded colors

Contrast and color balance recovered on faded prints; optional colorization for black-and-white scans

Soft focus

Mild sharpening pass recovers perceived clarity on slightly blurry or out-of-focus originals

What the controls will cover

• Restoration strength: light, medium, and aggressive presets — use light for already-decent scans and aggressive only for heavily damaged originals where artifacts are clearly visible.
• Colorization toggle: opt into AI colorization for black-and-white or severely faded prints as a separate step after structural restoration.
• Face enhancement: an optional face-aware pass that improves clarity specifically around portrait subjects, common in family photo restoration.
• Output format: PNG for lossless output (no new compression artifacts added on top of the restored file) or high-quality JPEG for smaller share-ready exports.

Tradeoffs and safe use

AI restoration is generative at the pixel level — it adds detail that was not captured, it does not recover information that was never there. Restored textures look realistic but are not the original; that distinction matters for archival work, legal evidence, or any context where authenticity is required. Always retain the unmodified scan as your master file.

Aggressive restoration on a good-quality scan can introduce its own artifacts: over-smoothing, slightly plastic skin texture, or halos around high-contrast edges. Match restoration strength to the actual damage level — more is not always better. For export format decisions after the restore pass, the guide on image compression and reducing file size help balance quality against byte budget.

When not to use photo restoration

• Legal or archival masters: any modification — even noise reduction — can affect admissibility or archival integrity. Keep untouched originals and process only working copies.
• Images with large structural damage: missing areas larger than a few percent of the frame, severe water staining, or fire damage need professional manual retouching, not an automated pass.
• Already heavily processed files: running restoration on a photo that has already been sharpened and noise-reduced risks over-processing. One clean pass at native resolution is almost always better than multiple stacked passes.
• Scientific or medical images: restoration models alter pixel values; diagnostic or measurement data in the image must not be modified.

Recommended workflow for scanned prints

Scan at the highest resolution your equipment supports (600 DPI minimum for prints, 1200 DPI or more for small originals). Save as TIFF or PNG — not JPEG — so no compression artifacts are baked in before restoration. Run restoration at native scan resolution, verify the result at 100% zoom, then downscale or export as JPEG for sharing. If the original is smaller than your target display size, upscale after restoration, not before.

Restore Old Photo questions, answered