Don't Randomly Tweak AI Image Editing; Define User Scenarios First
Just finished reviewing BPs for several AI photo editing projects. A bunch of people come in bragging about how awesome their models are, claiming they can generate high-def details and auto-complete faces. All I want to ask is: what's the user scenario? Is it for aunties photoshopping travel pics, or photographers doing post-processing? You build a super model, inference costs two mao per run—will users pay for that? Back when I was at DJI, I felt products were most afraid of feature bloat with no one using them. My advice to everyone building AI software: think clearly first—is anyone willing to pay for your core function? Don't race on tech before finding the scenario; that's suicide.
Physix Frontier