sag@lemm.ee to 196@lemmy.blahaj.zone · 1 年前AI Image Rulefile.coffeeimagemessage-square25fedilinkarrow-up1673arrow-down11
arrow-up1672arrow-down1imageAI Image Rulefile.coffeesag@lemm.ee to 196@lemmy.blahaj.zone · 1 年前message-square25fedilink
minus-squareAVincentInSpacelinkfedilinkEnglisharrow-up70·1 年前Seems to me it’d be pretty easy to tell. If the footage was AI generated, fingers would be appearing and disappearing.
minus-squarePregnenolone@lemmy.worldlinkfedilinkEnglisharrow-up34·1 年前It’s a solved problem now. Most good AI models generate correct fingers these days.
minus-squareRiskable@programming.devlinkfedilinkEnglisharrow-up25·1 年前Well, no, actually: The AI image model will generate bad fingers constantly it’s just become easier to fix via a secondary step (e.g. img2img) or you just tell it to generate 50 images and just pick the ones that don’t have messed up fingers 🤷
minus-squareWaluigis_Talking_Buttplug@lemmy.worldlinkfedilinkarrow-up14·1 年前I mean, probably not in 3-5 years let alone 10.
minus-squareAVincentInSpacelinkfedilinkEnglisharrow-up10·1 年前And by that time the neural nets will have figured out anatomically correct hands anyway making this product doubly moot
Seems to me it’d be pretty easy to tell. If the footage was AI generated, fingers would be appearing and disappearing.
It’s a solved problem now. Most good AI models generate correct fingers these days.
Well, no, actually: The AI image model will generate bad fingers constantly it’s just become easier to fix via a secondary step (e.g. img2img) or you just tell it to generate 50 images and just pick the ones that don’t have messed up fingers 🤷
Cries in Artbreeder credits.
I mean, probably not in 3-5 years let alone 10.
And by that time the neural nets will have figured out anatomically correct hands anyway making this product doubly moot