You pulled a shoebox out of a basement and the top photo has fuzzy violet-grey blooms across the sky and a powdery ring where somebody's face used to be. That's mold — and if you wipe it with a paper towel you'll smear spores into the emulsion and make the stain permanent.

What causes mold stains on old photographs

Mold is a biological problem layered on top of a chemistry problem. Fungal spores are already present on almost every stored photograph; they germinate when relative humidity rises above roughly 60–65% and find a nutrient source. Photographic gelatin is that nutrient source. NEDCC's guidance on moldy paper recommends dropping RH below 50% or freezing materials for 48 hours specifically because the fungus only sits dormant at that point, it doesn't disappear (NEDCC Leaflet 3.8).

Colors come from the species. Aspergillus and Penicillium tend toward blue-green and black. Some Cladosporium strains leave a rusty violet halo. The fuzzy texture you see is a hyphal mat; the darker centers are where the fungus metabolized the gelatin binder and excreted pigments into it. NEDCC's photograph leaflet notes that gelatin becomes "soft and sticky" in humid conditions — meaning the mold didn't just sit on the surface, it digested part of the image layer itself (NEDCC 5.3).

How AI handles mold stains — and where it can't

PhotoFlip's restoration model treats a mold stain as a structured occlusion problem. The pipeline does four things for this damage type specifically:

• Stain isolation. The model learns that mold blooms have soft radial edges and cluster with similar-colored siblings, so it segments them differently than it would handle a scratch or a crease.
• Color-cast compensation. Where gelatin under a stain has shifted yellow-brown from biological activity, the model pulls that region back toward neutral using surrounding skin and paper tones as anchors.
• Inpainting the lost area. On the eaten-away centers, it generates plausible content from context — cloth texture, wall texture, hair — and blends it into the mold-free surround.
• Face rescue. If mold landed on a face, the face-restore pass reconstructs eyes, mouth, and skin from its facial prior.

Honest limit: where mold fully destroyed the gelatin — deep black centers with no image information left — AI is inventing, not recovering. If the only trace of a child's face is one surviving eye, what comes back will be plausible but not guaranteed to match. The Getty's preservation guide is frank that biological damage is among the hardest to reverse because the substrate itself has been consumed (Getty Conservation Institute).

Example restorations

• Basement wedding portrait, 1967. Violet blooms across the bride's veil and a fuzzy black ring on the groom's lapel. AI rebuilds the veil lace from the clean half of the dress, clears the lapel stain, and recovers both faces.
• School class photo, 1981. Diffuse green haze over the whole image. Here AI is mostly doing color correction and micro-contrast recovery — the damage is surface-level and uniform, which is the easiest mold case.
• Great-grandmother's studio portrait. Deep black centers on the cheek and forehead. Expect a believable reconstruction of cheek and skin tone but be aware the model is filling in from learned priors, not from the original image.

How to restore a photo with mold stains

1. Scan or photograph the print in good light. Do not wipe the surface first — loose spores smear into the emulsion and make the stain worse. If the mold looks active (fuzzy, not powdery), isolate the photo in a sealed bag before handling.
2. Upload the file at photoflipai.com/restore. PhotoFlip runs in the browser and processes the image in-session.
3. Review the output. If the face was partially mold-eaten, also run it through face-restore — the face-specific model works from a different prior and often rescues detail the general restore pass smooths over.

Related damage types: if your photo has mold plus heavy brown discoloration, the yellowing from oxidation is probably compounding the stain — see yellowing. For pricing, credits never expire at photoflipai.com/pricing. Full pipeline details are at how-it-works and common questions at faq.