Some Chinese students in my art class are getting good at anatomy insanely fast. One of them told me they draw human figures, run them through AI for critiques, redraw them, get more critiques, redraw them, etc., for hours at a time. I didn't think it was possible to make progress so fast, but apparently if the feedback loop is tight enough, you can. Should I give this approach a shot?
No grandpa that's not how AI works
>he thinks the AI feedback loop is the approach, when in fact it is the "redraw them hours at time"Just draw more.
>>8004547>some chinamen are using ai to get heckin good>posts bridgemanwhy not just post an example from their blog or instagram or whatever?
>>8004559Because this never happened and he's just an AI shill
>studentsthey can just ask their teacherAI would train you wrong
Ignore jeet AI demoralization shill threads
This does work for fundies but not for style. You can probably 5x your learning speed with a good AI tutor, but beyond fundies its value drops fast.
>>8004566Depends on what it is. For most cases the leading models can generate consistent forms and perspective. Problem is that it's generic and won't do specifics but artists going into that knowing that can still get value if they have an interesting design and know how to follow the forms. AI can generate 3D models that have fucked up normals for games fland printing but for a 2D artist this is irrelevant. AI won't revolutionize the world of the industry but it's definitely not going to suddenly become worse or disappear.
>>8004566Not defending AI slop but teachers can often train you wrong too
imagine being such a nodraw that you start trying to scare anons into being nodraw
>>8004597Why would it scare anyone, fag?
>>8004553this, ai is just like a normal guy talking to you with infinite patience and a fast search engine. But it can generate a picture of what your drawing should look like if its typical anatomy and you can copy and try to see your mistake.However this requires the ability to see what you draw and visualize how the lines and forms could be different. That alone already filters 99% of people and is the reason they're stuck. If people could just look at their art and see that something is too big or the wrong angle, then they wouldn't be here making threads about it.
>>8004599That's a question for OP not me, cretin.
>>8004547I did something exactly like this and I got roasted in /fig/ because the perspective was fucked up.
>>8004617even if it was, you still made much more in mileage and you can fix perspective much easier now. You're still way ahead of nodraws.
>>8004547It's not a bad idea on paper, but AI will hallucinate the shit out of what improvements you need to be doing. You would have to have a model explicitly trained on the exact linework that you want to replicate, as well as newbie artists trying to replicate it.
>>8004547why not just use your eyes and compare your drawing to reference to see where it falls short and fix it in next iteration?
>>8004647I'm just reinforcing what some other anons have said in the thread. It's a good idea to be critical and not mindlessly copy. But yeah I agree drawing more is always better.
Tried this with free deepseek. Not even slightly useful. It can identify whats in the image well enough but critiques it just seems to hallucinate.
>>8004547>for hours at a timethat's the not so secret, secret sauce. grinding for 4+ hours every day. chinks are good at this since their parents make them grind the books or grind the violin or whatever for hours every single day since they were kids.
>>8005006>4+ hoursThat is not a grind, that is just a normal dedicated person level. Asian level of grind is 10+
>>8004592avoid proko
>>8004602AI being infinitely patient is what makes it a beg trap. It says incorrect things that sound correct at a surface level, people would rather deal with that than get sneered at by human art communities.
>>8005185Which one you suggest?
>>8004547and what AI model did he use?
>>8004547Tell these "chinese students" that they are retarded, and AI doesn't work that way. Damn, those "chinese students" are colossal retards.
Dump fucking clanker.
>>8004566Many of the AI have been retooled to always disagree at some point in their messages now but default, because too many people used them as therapist-girlfriends and allowed the AI to convince them to commit suicide. The fix is basically when the AI takes in even slightly ambiguous user input, it has boilerplate response templates like "yes, but" over and over. So you could actually upload good work to an AI and it will force itself to call it shit, kinda like the threads here LMAO.
>>8004547I'd rather hone in on my own very cartoony soul.
>>8004547Why not just do them in AI in the first place?
>>8007357Holy fuck, what the shit is going on there?
