AI ‘restoration’ slammed by photo pros

ON1 announced Restore AI will ship in April 2026 as a new module inside ON1 Photo RAW MAX, billed to “repair damaged photos, restore faded colors, enhance lost detail, and colorize black‑and‑white images.” (on1.com) PetaPixel’s review of ON1’s materials noted the company supplied 73 before/after examples in its press kit and documented cases where faces were redrawn with added teeth, makeup and inconsistent textures. (petapixel.com) In a March 28 op‑ed, PetaPixel author Jaron Schneider argued that AI “restoration” substitutes model‑generated detail for missing evidence and contrasted that practice with conservation norms taught at Florence’s Opificio delle Pietre Dure. (petapixel.com) Archival and genealogical organizations have formalized caution: the Society of American Archivists published guidance on generative AI in the American Archivist, and the Coalition for Responsible AI in Genealogy issued a “Protecting Trust in Historical Images” statement on November 4, 2025 recommending always label, always cite, and treat AI‑modified images as illustration rather than evidence. (www2.archivists.org) Forensic researchers have documented similar risks in news contexts — UC Berkeley professor Hany Farid’s GetReal lab concluded a widely circulated Minnesota shooting image was an AI enhancement — and a June 25, 2025 case study of attempts to restore the world’s oldest photograph highlighted “unintended consequences” when generative tools invent lost features. (ischool.berkeley.edu) ON1’s marketing materials defend Restore AI as using larger cloud‑based models to reconstruct detail, while early third‑party coverage and forum comparisons have already raised questions about hallucination, texture fabrication, and quality relative to established tools such as Topaz — concerns surfaced ahead of the April 2026 rollout. (on1.com)