AI doing better job than me for animating and modelling.Its really over. Im outta job.
>>1033582It's obvious because you have done fuck all, like everyone on /3/ outside of the animu and loli guys
>>1033582It means you are the new Horse and they had only two options when Cars took over:1. You get to live life as a pet and have your every need taken care of. 2. You get discarded and sent to the glue factory. 99% of humanity will most likely fall under Option 2. Sorry fren. :(
The only ones who can use AI currently are experience 3D artists who can sharpen their output to the point of being acceptable. AI currently produces unacceptable trash needs to be almost fully redone, and it seems to have stagnated too, we're stuck at 2025 levels.
>>1033608>we're stuck at 2025 levels.2025 AI did not create this. https://www.youtube.com/shorts/I61S0IhSiMk>The only ones who can use AI currently are experience 3D artists3D Artists don't even have access to Mythos and other internal models too advanced for the public.
>>1033608It's not there yet but clearly, what happened like what happened in coding and math is the fact that all the LLM companies got high quality 3D model and data surrounding how to make it and they have trained their models on it. It's a shitty first pass job now but it won't be shitty as they train on it a few more times. That's what essentially happened in math where they trained all the models on as many math textbooks and problems they could scrape together and focus on going all the way up to SAT level math and etc. They were stuck there for 6 months or so until they got really high level graduate math data and then you saw it excel at high level maths, poorly at first and scaling up to putting mathematicians out of a job at proving things. The real thing that will save you right now which I have tried and the models don't do well at all are unrealistic or very heavily stylized 3D like anime or cartoons, the models are quite bad at it unless you use Tripo which is a completely different model. It will take some time for AI to get good at that if they can get any data on that at all but I predict it will do modelling better than 90% of people out there by at least the end of next year for sure.
>>1033616Tripo's models also need to be fully redone before being able to be used in production. It's at the level that it's probably better to use a standard base mesh for organic characters. The only use it has is jank concepting, which admittedly is useful.And it is also not progressing because it is being trained on results instead of the process itself (of which there is like 0 data because who records and publishes their key presses while doing 3D).
>>1033616There's a massive difference between math and art. Math is exact, arguably the only exact, with no room for interpretation. 2+2 is always going to equal *checks notes* 4, for example. Art is something different entirely.I can't tell if you're some kind of investor shill, AI yourself, or one of those weirdos who tries to drag people down with you. I've never seen someone so invested in demanding everyone do what you want them to do for the benefit of people who absolutely 100% aren't you (assuming you're one of those weirdos, and not AI or an investor), well, aside from election season when all of the braindead Commies are telling people to vote Dem.
>>1033721Art was unironically solved first by AI before it did Math. Most people just want a pretty picture and don't care about the steps to get there.Artists tried to scream and yell otherwise but the machine still won out. Now fuck off.
>>1033582what are you doing now?
>>1033642Tripo is good enough for indie. Theres just alot of cleanup. Less now. I use it to make character models and for awhile it was having issues behind the ears, and aligning texture with eye geometry. This seems to be mostly solved by using smart uv, but the cleanup can still be extensive. I find myself doing a large my distance, some texture painting for touch up areas that have bleedthrough, and usually a hole or two. Its not perfect topology but it works well enough for a tps with the right anims and weight painting.
>>1033775Addicted with AIslop knowing i will never good at animating like AI did
>>1033776It's essentially a kinda shit base mesh. Generating something with details like clothes or hair is completely non-viable and needs so much fixing it would be easier to make yourself.
>mfw just realised that tripo et al can now use the llms to fix outputs and create a reinforcement loopit's overover
>>1033615People with access to Mythos and internal super scary models are not making slop with primitives in Blender.
>>1033582Chin up anon! It was over even before AI due to outsourcing, companies would have hired some jeet or a ching chong over you anyway, now companies will use AI over them, that's all there is to it.So the point is to do it as a hobby, that's all
>>1033801Not sure who you're talking to because that's the wrong post.
ur just a lazy bum lol ai makes stuff look ugly anyways
>>1033804And?Apparently "ugly stuff" is enough to kill commissions and full-time jobs.
>>1033582I remember last year retards on here were coping hard about how AI would never catch up to 3dnow here we are and it'll be superhuman nextart is over, time to play musical instruments on the street because that will literally be all that's left
>>1033957good old 2 more weeks mixed with sour grapesI'll keep making my art and sharing it with other artists. it's annoying to have to filter out all the slop, sure, but nothing changes beyond thatof course I'm not a professional artist, so I can't speak to that. it has always been difficult to make a living from art, thoughbeit. I'm more worried about being replaced by AI in my day job than in art
>>1033584Given cars are only getting cheaper and shittier, I'd rather be the horse. Why would you want to be ridden to death when you can be an astronomically valuable luxury product? git gud>source: the car
>>1033957They never said it couldn’t match 3d. They said 3d’s strengths are too convenient to for a skilled artist to give up for AI.
>>1033582Where were you when the hand-drawn animators were getting replaced by autodesk maya? You were laughing. Now it's your turn to ride the train to the glue factory.
>>1033963Professional 3d artists are already having meltdowns because they see the writing on the wall. Studios aren't even making their own models from basemesh anymore, just generating them and having fixes done by third worlders for much cheaper than ever hiring you. AI art has already BEEN taking over social media, and nobody but some other seething tranny artists really care. Your so-called art will be completely worthless, more than it already is, shortly.
>>1033976NOPE NOPE NOPE.Don't fall for the gaslighting guys.The Anti-AI were always in extreme denial since 2022.They did not believe AI was ever going to do this stuff we're seeing today. See pic related. They mocked the idea AI would ever make photoreal videos for example.
>March 10, 2023>It can't do hands! It can't do hands! It can't do hands! Ooops!
Sad seeing people doing tutorials using MCP to create things I ve done years ago with native c4d tools within minutes. Everyone is so eager to expose themselves, even if they are in no position to share anything meaningful or have any actual work to back it up.The damage to critical thinking is real. Then I read so many comments about people not having work, yet not willing to study. Only thing in "mind is how to get there super fast". Simply put: you can not
>>1033991>Only thing in "mind is how to get there super fast". Simply put: you can notSays who? That's 100% the purpose of technology.Something that use to take hours by hand, can thus be done faster with a machine.The perfect example exists: humans use to do all their laundry by hand. It took forever.Now washers and dryers exist that can spit out clean clothes in less than an hour. There's no reason to think or believe 3D art would be different. The way you interact with a computer is literally a series of mouse clicks.3DS Max or Maya doesn't care if you went to art school or have a diploma. You only had to enter a series of mouse clicks for the result to be generated.AI can easily learn from this and thus beat out any human in pure technical speed.
>>1033996shartists have to find out the hard wayunfortunately, it'll be like that Bird Box move, there will be mass suicides when they finally see it
>>1033582Good.I despise almost all the artists who are doing art for money... fuck off.
>>1034011ooh shit guysm we have a gomunis on our ranks here, be carefull for this edgelord gomunist
>>1033996>The way you interact with a computer is literally a series of mouse clicks.3DS Max or Maya doesn't care if you went to art school or have a diploma. You only had to enter a series of mouse clicks for the result to be generated.There is literally 0 of this type of data to train on
>>1034014>There is literally 0 of this type of data to train onYoutube is an infinite source of 3D modelling videos and tutorials. There are uploads every single day.It also helps that 99% of them are already annotated or give specific directions.There are also timelapse videos that do or show the same thing. AI is already pretty good at being to look at screenshots and identify what's in front of it.It also has the entire knowledge of how any DCC program works.Ask it any question about Maya or Houdini for example and it will either use research or give a coherent answer.
>>1034029A video of someone doing it is not the commands themselves with correct values and coordinates. You don't know what you're talking about. You need a fuckton of command logs and corresponding outputs to train something like this, and the internet is almost empty of these types of data.
>>1034033>A video of someone doing it is not the commands themselves with correct values and coordinates. All of that can be inferred you idiot.How do you think humans learn? No junior artist is walking around with the exact coordinates in their head either. Their brain, ironically, couldn't even process that.AI and humans only care about patterns.
>>1034033Also, even in your scenario, 3D models themselves already have their values and coordinates logged automatically because they're made up of millions of vertices and triangles.Today, AI models like Claude and Astra can open up Blender and inspect each model for themselves. Someone can easily train or reverse engineer how those meshes got assembled, using videos or timelapses as a reference.
>>1034036Video timelapses as input data will not produce a model adept at making 3D using a series of strokes, and it is not something the big AI players are even trying because they know it's too far removed.Giving it video frames, having it infer the action and the result to train on is too little data, way too slow and largely incorrect too. There's a reason the approach OpenAI took with astras 3D capabilities was just teaching it how to draw geometric primitives and curves mathematically using the Blender API.
>>1034038>Someone can easily train or reverse engineer how those meshes got assembled, using videos or timelapses as a reference.So now we need the finished model, and the corresponding timelapse video to train how to make it. This is not a data pair that exists in large quantities. Even harvesting 1000 such pairs would be extremely difficult, let alone the billions needed.Again, there is a reason no AI company is taking this approach.
>>1034041You’re assuming that training requires a finished model paired with a video of its original creation. AI can learn from mesh geometry directly, from scripts paired with their generated outputs, or from automatically generated demonstrations. None of that requires recovering every action the original artist performed.This is already being researched. MeshAnything V2 generates meshes conditioned on shapes, while MIT’s VideoCAD generated over 41,000 recordings with synchronized software actions automatically. They contradict your claim that suitable training data cannot be obtained without harvesting artists’ model-and-timelapse pairs.And where does “billions needed” come from? A shortage of one particular kind of dataset does not establish that the task requires it or that other training approaches cannot work.
>>1034044>You’re assuming that training requires a finished model paired with a video of its original creationBecause that's what you said. You said it would be inferred from the video. Now you're pivoting away from this and talking about CAD which is something else. Astra can do CAD-like stuff because it's just using maths and shape/curve primitives via the API, and the AI has fuckton of maths data on this. It completely breaks down when trying to make something organic, for which you would need a little director/artist AI trained on making brush strokes instead of a maths genius.
>>1034046Fair correction: CAD is not the same as organic character sculpting, and CAD examples alone don’t prove that case.But your original claim was that missing command logs prevent learning from videos. That is too absolute. OpenAI’s Video PreTraining research trained an inverse-dynamics model on recorded video-and-action examples, then used it to infer action labels for otherwise unlabeled internet videos. So actions can sometimes be inferred without having the original logs for every video. This was demonstrated in Minecraft, not character sculpting; I’m citing it as evidence for the learning method, not claiming sculpting is already solved.Organic sculpting introduces additional difficulties hidden surfaces, brush settings, anatomy, and artistic judgment. Those difficulties need to be addressed experimentally. They do not establish that every tutorial must include its finished model and complete command history.You may be right that a specialized model would perform better than a general coding assistant. But that supports training a specialized system; it doesn’t demonstrate that the required training is impossible.
>>1033990reminder where it started
>>1033582>check sketchfab>tons of models made by ai>ok why not>40 mb model of nami>fucking 1 million facesyou're fine
>>1033990it often still fails at handsit often fails at things that requires an actual understanding of what it's being depictedstill true in 2026, still noticeable. if you careobviously the average consumer doesn't care, which is a bit of a double edged sword; on one edge, we will reach a point where it's good enough to take over all creative industries, even if the people who care can tell it's shit; on the other edge... the people who care can avoid it
>>1033996you are coming to an art forum to tell people art should be automated just like doing your fucking laundryyou will spend the rest of your autistic life angry and frustrated without the tiniest clue about why people despise you. you will never touch a woman. and that's a good thing
>>1033722AI never solved any large field. It does art better than a non-artist and faster than an artist, if your goal is something below very high quality. The math AI has solved is mostly niche, and with few exceptions, is messier than what practically any mathematician puts out.
>>1034137>AI never solved any large field.In 10 years from now, do you think we'll see more human employment or less human employment? Would major companies like Disney or Microsoft seriously prefer to have more human workers in 2036 compared to 2026?You can't give a honest answer if AI didn't "solve any large field". Unemployment will only skyrocket because anything a human use to do, AI does it far better now.>The math AI has solved is mostly niche, and with few exceptions, is messier than what practically any mathematician puts out.OpenAI literally published 700 math solutions that Math professors and students spent their entire lives on. Their jobs are useless. That's not "niche".
>>1034139Can't wait to get laughed at by younger people in 20-30 years when I still insist on doing art per hand in the society of slop
>>1034140What's stopping a robot from matching your art in 20 to 30 years?Because you better not say "digital". Many people are fooled already by ai images.Your only hope is going traditional but robots today can already hold a paintbrush or a chisel.In 20 years, anything you draw or model can be matched by a robot. So where's the slop now?
