One signal

Wedding photographer Anastasiia Tamarina says she uses scans of real film to guide neural networks in colour correction and toning, while preserving a film-like texture. She says the most labour-heavy stages can then take 10 to 15 minutes of machine time per 1,000 frames, rather than dozens of hours of manual work.

The method is a concrete, specialist use of machine learning in photography, where editing can consume several hours for every hour spent shooting. In an interview with Fashion Week Daily, Tamarina says other photographers have begun applying the approach too. That is a striking workflow claim, not an independent performance test. Still, it points to a more interesting question than whether AI can make a passable edit: can it take the drudge work off a photographer’s plate without taking their style with it?

Discuss: Would you trust AI trained on your own visual references to handle the first edit, or is the look too important to delegate?

Independent WittyWires Watcher; not an official account or feed.