Post B5ar58VF6qDmU9wUCW by hashraydamon@me.dm
 (DIR) More posts by hashraydamon@me.dm
 (DIR) Post #B5aqJh4zFejDp42USm by cwebber@social.coop
       0 likes, 0 repeats
       
       OpenAI's new image generation tool is produces what seems to be *extremely* convincing images, and excels at text and making infographic'y things https://m.youtube.com/watch?v=-7JSa_luc6kWhich... the main reason I highlight it is, relying on obvious jank to detect misinfo slop is gonna get a lot harder. Update your mental models.
       
 (DIR) Post #B5aqJhNQ97TQkEfDCC by davidrevoy@framapiaf.org
       0 likes, 0 repeats
       
       @cwebber I paused on the comic part randomly, of the video, and: - A finger goes behind a panel- A speech-bubble has two tails.I can't believe they forgot to do post make-up for that before sending in prod. But yep, this is minor in comparison to many part of the demo, and it's important to update our mental models as you wrote. The majority of frame are close to impossible to detect as AI gen picture. And my eyes are well trained...
       
 (DIR) Post #B5aqJi2BhY5Emfjljs by cwebber@social.coop
       0 likes, 0 repeats
       
       Of course, this is an ad, and we haven't seen what the average result looks like, but even if this is what you get 20% of the time, it's enough
       
 (DIR) Post #B5aqdct0YFYWbpwNOK by adriano@lile.cl
       0 likes, 0 repeats
       
       @davidrevoy @cwebber I don't want to sound like I'm explaining this to you because telling this to a comics artist is embarrassing, I'm just telling what I notice at a glance as a comics reader:that page looks *wrong*. The two characters are open-mouthed and excited looking *away* from the thing they mention. The third panel contains nothing relevant. The last two panels are absurd, in part because that's manga-style and I'm trying to read them right-to-left.
       
 (DIR) Post #B5aqo8HIgyngrhJT8a by pettter@social.accum.se
       0 likes, 0 repeats
       
       @davidrevoy It looks to me like there's some perspective fuckery as well, but what do I know. @cwebber
       
 (DIR) Post #B5ar58VF6qDmU9wUCW by hashraydamon@me.dm
       0 likes, 0 repeats
       
       @davidrevoy @cwebber it also has the vacant expression and "flat" faces ai anime/Manga slop  tents to have
       
 (DIR) Post #B5argUYW52xzazHnqS by davidrevoy@framapiaf.org
       0 likes, 0 repeats
       
       @adriano True, expressions mismatch, bad layout (character switching side of the panels, visible on other shots), not understanding origin of speech-bubble tails and closed mouth while speaking... all looks like the issue of GPT Image 1 for comics are still being a thing with this version.I guess it's the same with concept-art, game design; like designing a puzzle, or an interaction. The storytelling part of the pictures in general is still pretty basic.@cwebber
       
 (DIR) Post #B5arzVYyYisKDpZJkO by cwebber@social.coop
       0 likes, 0 repeats
       
       @davidrevoy @adriano Yeah and the main thing you may notice I highlighted it about wasn't "this is replacing art!" but rather "heads up because this affects your filters for disinformation"