As a professional 3D artist, given the current state of this technology, it primarily serves as a tool for generating a good base mesh to start from-equivalent to roughly four hours of work. However, when you begin analyzing the output and attempting to integrate it into a production pipeline, you start noticing numerous small errors in the meshes. I can see a clear application for inanimate and background objects, but textures and retopology would still require an artist's touch. Overall, it saves about half a day's work and eliminates the effort of matching a concept, but not much more than that.
Based on how far this tech has come, it's not gonna be long before A.I. can do it better. It's like when photographers started replacing painters, at first it was niche and now it's the standard.
Yes, people must accept that ai technology never stops learning, so this is only the beginning. We are probably only a year or so away from a very competent mesh generator.
I'd agree, prob less than a year. I also would bet that soon we will be able to talk with an agent AI and ask for the corrections while your spinning the model around and noticing the errors, though that to will prob pass, and it will just come out initially better than anything we can currently do. I'll be happy AF when either of those situations occur. It will be like MJ is now with 2D.
@@jac001 a 2D image is very different from a 3D mesh. This technology is nowhere near advanced enough to create a complicated, professional game character.
One thing to note is that this does not generate meshes from text, it converts mesh cloud point data into a mesh with vertices. So you still need cloud point data to input
i understand why artists are so against this technology, but in the same breathe for people like me with zero 3d experience attempting to make a game, this could be a game changer. double edged sword problem.
I mean, you could always learn how to model. It's not like it's in the realm of impossibility to pick the most free software available right now called blender3D and allow your brain to do what it does best.
If you have zero experience and not willing to learn - you should not be making a game in the first place. The market is already oversaturated with beginners and asset flip wannabes and barely anyone makes a living there.
@@FireF1y644 Not everyone has to sell their stuff, game creation can also be just for fun, not everyone is trying to make a buck from their art/games whatever
@@FireF1y644 He never said he wasn't willing to learn, you put those words in his mouth and since when were you the guardian of who can and cannot enter market. Get off your high horse.
@@LethargyPike you can't just "learn" to model. For a game designer/developer to also be able to create 3d models and textures, animations, music, or even write, at the same level as his main skill, he needs to have at least 2-3 years of experience in each. People just don't have that kind of time. I learned to use Blender in 4-5 months, and I can make basic models or edits, but creating the high quality complex models I want (especially characters and creatures) is way beyond the skill someone with only a few months of experience can achieve in a short amount of time. I could probably make a good enough character model if I spend a few months only for this, but I am not going to do that when AI will be able to generate it within a few minutes using a base image or a text prompt.
I'm a 3d modeller, why else would I be here right?, but I am one of those who adores the time saved by all this A.I. generation. I've dabbled previously and it's always been really janky and the cleanup can be longer than the modelling from scratch. But now it's getting good I would much rather have the ability to prompt my 3d modelling rather than spend months making assets for a game that sells 100 copies.
If a non 3d artist can generate any 3d model he wants with such a model by just writing a text or feeding an image, no one will pay the artists... The same with 2d artists, software developers, musicians/singers (because Suno and Udio exist which make much better music at this point), writers, etc.
Since I'm a professional 3d modeler and AI enthousiast, I've been noticing all the 3d polygon generator, and since a year or so there have been some before NVidia, such as Meshy, StableProjectorz,, Solace Vision, Latent Labs, Common Sense Machines, Spline, 3D AI Studio, DeemosTech , some of them might be cloud mesh and some might be dead by now. I think you can even generate the texture with 3dFy.
@@Dreaven.Mehendysis looks like us 3D modelers are on the chopping block. Corporate companies will consider their bottom lines and your enthusiasm for ai will be replaced by lerning to work fast food. Ai development will eventually be taken over by advanced ai, thats literally their goal, so what will you do? What would justify a good salary when they can hire anyone to do the job?
@lazerith840 @lazerith840 As any other job. The problem will be being able to not fall into leisure rather than many-parts-of-the-population possessed enterprises. Seeing the results of how AI is received in the consumer world, AI made produce do not have a good prospect. Just look at how Coka cola AI ads reception. Working with AI everything I can assure you that the results are ridiculous and time consuming. The advertisement for it does most of the job, it's all glitter and file gold.
@Dreaven.Mehendysis absence of a job just means you will be broke. You can live with that, but I got bills to pay. Your so short sighted that I kinda want you to find out these things on your own.
considering the fact that AI has very limited context length due to our hardware limitations, I heavily doubt this will be more than a gimmick where it can make relatively simple shapes with high accuracy but incapable of making any game ready or animation ready assets, and when it needs to do something semi complicated like fingers it will probably mess up way harder than 2D image generation ai's. I would rather have them work on some proper AI retopology but who am I kidding, they can't even do that and they expect to do the entire workflow, good luck with that lmao.
so they auto ran ran quad remesher after generating their original mesh. so as a 3d artist im constantly struggling with making real time meshes. this could help me become more productive and that would be cool. i could run this on a sculpted object and get a great start on a render baking mesh. so i think it wont replace a 3d artist just make us way more productive and think that would fit in an area thats needed. these models they are genning are very topologically simple to parse from a photo or many photos which is fine. itll be interesting if they can scale the complexity of these models.
I wonder if they have a professional 3D Artist giving feedback on the topology, I will still say it is pretty bad topology, especially for organic models. On top of that, 3D model needs uv, characters need rigging, there is still a long way to go. 3D is way way more complex than 2D.
topology is very bad, the texture will be hard to map. and if you use a lot of these, these will bloat your game or app. you also cant 3d print these as these are point cloud generated which will have a lot of lose points and disconnected vertices
@@SmartPoly But people don't take a stand against it until it affects them. This is the great era of selfishness, individualism, and hypocrisy described in the Bible as "the great tribulation," the period that would precede the end of times. If I only cared about this issue on the day it affected me, I would consider myself unworthy of being saved.
Jobs come and go-that’s the nature of progress, and you can’t stop innovation. If AI were capable of performing heart surgeries, some people would lose their jobs, but overall, it would be a significant benefit for society.
You say that until an emp or solar flair flips the switch, or the electrical use causes more coal to be burned by power plants to power the machines on top of power our homes. Jobs do come and go, but that doesn't mean it's positive. It has negative effects to economy. Progress doesn't always bring positive outcomes. As more people lose work, then less people can afford to purchase things. The less people can afford to purchase, the less earnings a company can make. Ai is going beyond just digital work, and I lived in Detroit and seen first hand the downfall of that city which started once robots were introduced into factory work. Before jobs used to come and go, but it still involves humans, this is different. It's literally removing ourselves from the equation all together. Not all progress is good.
@@lazerith840 well said. so many do not understand that what we choose to do with technology matters. its a choice its not some autonomous destined thing. we choose. really good book on the subject, power and progress our 1000 year struggle over technology and progress.
Actually you just need a single regulatory vote to stop it, during psychedelic research in the 70's the future of psychiatric is obvious psychedelics for most in the field, where is the psychedelic research now?
@@lazerith840progress doesn't, but productivity does. Never before in human history some technology that increased productivity brought misery and unemployments, but increased our quality of life and wealthness. Let's Hope the rule keeps, if AI brings more productivity, so we can expect all those things to come again.
@starvosxant4348 Detroit was one of the wealthiest city's in the 1950's. Then the introduction of robotic machines doing factory labor caused unemployment leading to other social economic impacts in the 1960s. Over the decades it has gotten worse to the point that the city went bankrupt in 2013. You think all those people losing their jobs wasn't a contributing factor to the decline of the economy and social structure there? So in reality it seems like the only time a new tech caused issues, was in automation, just like ai. Coincidentally I lost my 3D animation job to ai in Detroit, and had to move to Louisiana for a new career path.
Recently started teaching myself UE. I figure by the time i get proficient enough to do something with it people will be able to make a full game with a simple prompt.
doesnt work that way. In order to create something u need a deep understanding of what u are doing. Ai can lend u a hand but it wont make all the job for you. Manual crafting wont be replaced. So i suggest you to stop looking too much into what the new tech is doing and put more time learning the basics.
@Damian_DH for now. AI will eventually take all jobs, its just a matter of time. Its naive to think otherwise. It might not happen in the time it will take me to learn unreal but it WILL happen.
I'm not sure that there's too much magic going on here if you understand the 3D modelling process. Once the point cloud is generated, it understands from referencing information about the object online, breakes it into components based on this info, and then uses regular and long established auto-retopology algorithms (like quadremesher or instant meshes) to generate the topology. It's like a more informed version of photogrammetry with the manual steps removed. I'm sure it will keep improving, but I also think it will reach a limitation plateau. It won't be replacing 3D artists any time soon.
Agreed, except i would replace "soon" with "in the next 2 years" . Many of my web dev jobs have been replaced because it's easier now than it was 15 years ago to build websites with the tools and Ai now available. The same thing is already happening in the already incredibly competitive 3d jobs market. It will get worse. Some artist jobs will disappear, not all.
@DirkTeucher Some jobs will be autimated, no doubt, but the tech will be most useful as an aid to artists, rather than a replacement. It might reach a certain level of quality regarding 3D meshes, but I imagine there will be a quality ceiling, and artist input will still be needed after the initial generation. The non-artist 'AI bros' will always be unskilled tourists. Otherwise, if they had the skill they would be artists.
i think its more intricate than quad remesher, at 5:12 the box is very optimized but using a quad remesher it will definitely have more polys because it treats the whole object fairly to account the small details.
hi poly. i wanted to thank you for your work. FYI, I bought your survival course about a week ago (looking forward to time with it over xmas). i was a little nervous about getting the course as you are working through the remaster, but after going on discord I can see you update the new classes for the course about every 6/7days. i think you should push info like this when promoting the course. think it will help peopleon the fence. maybe also explain the 2 options you get when buying from the webpage. hopefully it helps you with conversions. oh, and a suggestion. maybe PCG cavern system.... may be interesting. i will be starting one anyway at some point. have a good xmas, and see you next year.
Are you kidding me ?? We already have tools like this for almost a decade! We use these Auto Retopology tools all the time. I don't see this being any different from them, maybe better at avoiding infinite loops but it isn't anything new. Maybe its just for views hence the misleading title. I think now a days people can simply write AI and Replace to a video and the Algo will do the rest. Chillax guys if you know a single thing about 3D creation this is one of the most used feature that comes with almost every software you use.
@@jenkem4464 The source model is not made by this, If I am correct. BTW Source models have been done through Photographs for a long time. Regardless I will actually be happy that a tool like this comes. Right now topology is a nightmare no 3D artist likes to do and all we wanna do is get rid of it as soon as we can, because its not art its just a roadblock we have to go through.
@@sauravbhagat4150 Oh Gotcha. Yeah, if the source model isn't made by this then I get it. It still seems to do a pretty good job. I don't know much software that can separate geo like that and still have as much mesh-topo control as this. This + Microsoft's new Trellis software will be pretty cool, at least for getting a decent base for people that hate doing all that. I love the sculpting and detailing part, but not the mesh parts. Can do it, but just not as efficient as other production artists.
@@lazerith840 And those replaced Portrait Painters later became the concept artists & other types of Designers, so they weren't totally replaced. Similarly there is still hope.
@@manvendrapratapsingh1920 there’s a difference between getting replaced by a object doing part of the work, and the object doing all the work. I’ve already lost a job to ai, so has others. Maybe you will understand once it happens to you. Working in 3D is already an uphill battle, and needing less just means less jobs for people. These companies are not going to stop until you can type what you want, and it makes it. Anyone can type, so they will hire less, and pay less. A career, has turned into a job, and y’all can’t see it.
When I look up MeshAnything v2 examples online the results they show are a lot better than what's represented here. The triangles for MA2 are also a lot more regular and less crushed, which make them better for UV mapping. The quads output for Meshtron is pretty intriguing though, but I think everyone is just cherry-picking whatever examples makes themselves look good.
3D modelers for lowpoly stuff are basically obsolete now. I suggest getting into Houdini and learn creating procedural systems. That´s what game development, VFX and advertising needs the most. Worldbuilding, Procedural animation and modeling. But static meshes and even characters can now be prompted or photoscanned, autorigged and even procedurally remeshed and optimized, auto UV mapped and procedurally textured. With the exception of very specific requirements like product visualisation, architecture or character design, a lot of tasks nowadays take much less manual labour, than even just a few years back. Only putting all those assets together into dynamic systems that adapt to the users requirements and save a lot of work are of significant value nowadays. Because that´s the only stuff AI struggles with. For now. And on the other hand, it is as easy as never before to access complicated software like Houdini because of having AI learning assistance and AI assisted code generation. It will not have an answer to every question, especially when prompted for something complex in one go, but it helps breaking down processes, helps understanding basics and functions, can be fed with additional information and documentation and even scientific papers. I have tested that by myself intensely in the last year with some very interesting results. Go procedural!
Ai developers should consider the social economic impact of its use and if it's causing more damage than good. Imagine going to college to learn all this, then still having to work fast food, until they get bots to do that too. They are pretty much forcing communism because who will buy products if more and more people are pushed out of a job. Replacing ourselves seems like a bad idea. It ignores the social impacts on society for corporate greed. Even the developers of ai will eventually become replaced by advanced ai, and then what. The willingness to accept long-term loss just for short-term gain and ease of work is not only short sighted, but also counterproductive in the long term. Imagine building our society around ai, then a solar flair or emp flips the switch, and nobody knows how to do the work on their own or how the process works. People tend to disregard these questions and sometimes behave hostile twords these questions when, in reality, someone needs to ask it, and it needs to be addressed. I personally like ai tools, but if it reaches the point of replacing a human job, then is it really beneficial.
Not impressed at all. Most complex stuff in 3d modelling for gamedev is not to create a low poly mesh. It is about finding the correct art style, the balance between details and game performance, it is about creating proper topology (meshoptimizer/Nanite/LODs), it is about UV unwrapping, it is about texturing, it is about animating. This tool does the simplest and least time consuming part of all 3d modelling. Good luck for all the folks in comments that went to "wow, I can create a 3d game now without any knowledge in 3d modelling"! Maybe you should have some, so you won't get so easily excited.
4:00 - Because AI looks at references on the web. I'm guessing first it decides what the object is, then he uses the internet to figure out anything it's not 100% sure of.
I do 3D modelling since 20 years, At early times we are the 2nd generation who created 3D modelling, texturing, rigging algorithms, I do support AI do these boring modelling things ,Even all stuff... Because every game, Every movie Starts with an Idea, So AI may make my Idea accomplished, If AI will replace all human with Idea creation too and will decide to get rid of Humans out of their way then it's OK too for me...
@@alteredcarbon3853 It's not moving super fast. It's just moving. It'll take years until it's production ready. But I couldn't carel ess. I do my own models and am not afraid of being pushed out by AI. It's meant to support artist not to replace them.
Smart poly prev. video: lmaooo. If you think this tool will replace artists, you’re cooked. Smart Poly in this video: Will This AI Replace 3D Artists? So what's your real opinion?
No, they're not this meshron thing looks terrible, and why use images to generate the model when a professional can make one with complete control. This type of AI is going to be used for people who can't model but want to make their "dream game."
@@LethargyPike lol, you clearly have not seen how fast AI models are getting better. OpenAI released the o1 model a couple of months ago which was the best AI model in terms of logic/reasoning, and a few days ago they released the o3 model (they skipped o2 entirely) which crushes every benchmark that existed in order to measure AI model capabilities, often being on the same level if not better than even human EXPERTS. By the end of 2025 3d model generating models will crush the best humans, and they will do that at a fraction of the time and cost required.
@OnigoroshiZero For $200 dollars a month, you can have access to ur AGI, which wasn't necessary to begin with. I'd like to see where your predictions go into next year I guess everyone's just going to get laid off and GTA 76 will come out because these AI'S are going to be able to create every single thing because they are "Experts" even if this occurs, I dunno why you're even remotely okay with that, why are you okay with AI replacing you?
Sincerely, that's leaning a bit too much towards delusional thinking. OpenAI floated the idea of charging $2,000 a month for the top models. They're burning through billions and billions of dollars to bring everyone the latest and greatest technology. Sora is $200 a month and people are already complaining about that. Devin is $500 a month. The idea that these companies will freely release the very best "job replacing" models for a fraction of the cost is silly.
I just have to wait a few more months to be able to use such models in order to finally make the assets I want to complete my game. AI is amazing, and it only gets better.
Wow, NVIDIA's AI turning text into 3D models is mind-blowing! The future of design and creativity just got a huge upgrade-imagine the possibilities for games, movies, and even VR. Truly next-level tech!.
@@islandonlinenews tell that to concept artist. Oh wait they are working fast food now because midjourny does their entire job. There's a difference between a tool, and a replacement. Ai to help with topology or UV unwrapping is a tool, ai that creates the model makes the textures and has multiple LOD's just makes someone need to find a new job after years of study. Nobody hates the ai, it's coding, but the developer need to consider the social economic impact of it. Short term ease for long term loss does not seem beneficial in the long run.
All you doom fearing never working a day in this industry worms. All I see is more room for work with real talent, you think people with no talent make anything of quality. All we are gonna see is an over saturation of Crap on art Station.
have to be naive to believe that in the coming years AI will not replace 3D artists , because the pace of AI development is accelerating and will only get faster but faith dies last of course. And to speak out against such technologies because of the possible loss of work and money is too selfish, you can't slow down technologies if they are for the benefit of society, otherwise it looks like aggressive capitalism where monopolists temporarily bury technologies for the sake of profit, and life goes on and we are not getting any younger, the more things are created that simplify our lives or save time, the better for us, society is more important than the desires of individuals who are afraid for their jobs, many professions have sunk, you can't knock them out and this process cannot be stopped
the more I follow this, I'm starting to get convinced we are possibly at least 5-10 years away from actually usable 3d models as a professional modeller might do and that's a big IF. While these look "promising" they are nowhere near usable for actual production work. I think modellers don't have a lot to worry about. It's easy to make a 2D image with a girl by analyzing a million other pictures of a girl, but making a production ready 3d model that's high quality is goint to still have to be made by humans. If Autodesk or Nvidia or even Google can't make this work so far very few others have the capability of doing it resources wise or smarts wise.
In answer to the video title: No, there are too many avenues of specialization and bespoke use cases that cant be brute forced or relegated to pure math in game dev. None of the models put out by this even shade correctly, the topology is awful. Even if the AI could do perfect wonderful ideal perfect topology recreation, it would still need a solid high detail authored mesh to work from, not this blobby point cloud shit thats currently spat out. If anything this tech might be a decent retopology system used in conjuction with photogrammatry or automatic lodding. I do this for a living, and I wouldn't considering anything spat out here anything than slightly below student work. "But itll keep getting better!" Yeah, at retaining detail from a bad source mesh, and then once the source mesh is gone, it'll retain detail with what level of control, with what level of optimization, how do you tell this to UV things in a way that lays the UVs out appropriate for a flow map, or for mirroring, or to prioritize surface area texel density based on approximate visibility within the engine, then comes art style and exaggeration. There are too many variables that you NEED control over to leave this in the hands of an AI, unless you want unoptimized messy shit.
As a professional 3D artist, given the current state of this technology, it primarily serves as a tool for generating a good base mesh to start from-equivalent to roughly four hours of work. However, when you begin analyzing the output and attempting to integrate it into a production pipeline, you start noticing numerous small errors in the meshes.
I can see a clear application for inanimate and background objects, but textures and retopology would still require an artist's touch. Overall, it saves about half a day's work and eliminates the effort of matching a concept, but not much more than that.
Based on how far this tech has come, it's not gonna be long before A.I. can do it better. It's like when photographers started replacing painters, at first it was niche and now it's the standard.
Yes, people must accept that ai technology never stops learning, so this is only the beginning. We are probably only a year or so away from a very competent mesh generator.
I'd agree, prob less than a year. I also would bet that soon we will be able to talk with an agent AI and ask for the corrections while your spinning the model around and noticing the errors, though that to will prob pass, and it will just come out initially better than anything we can currently do. I'll be happy AF when either of those situations occur. It will be like MJ is now with 2D.
@@jac001 a 2D image is very different from a 3D mesh. This technology is nowhere near advanced enough to create a complicated, professional game character.
@@jac001how is mj with 2d right now though? Last time i checked it was quite subpar and everyone hated the intrinsic style attached to it...
One thing to note is that this does not generate meshes from text, it converts mesh cloud point data into a mesh with vertices. So you still need cloud point data to input
i understand why artists are so against this technology, but in the same breathe for people like me with zero 3d experience attempting to make a game, this could be a game changer. double edged sword problem.
I mean, you could always learn how to model. It's not like it's in the realm of impossibility to pick the most free software available right now called blender3D and allow your brain to do what it does best.
If you have zero experience and not willing to learn - you should not be making a game in the first place. The market is already oversaturated with beginners and asset flip wannabes and barely anyone makes a living there.
@@FireF1y644 Not everyone has to sell their stuff, game creation can also be just for fun, not everyone is trying to make a buck from their art/games whatever
@@FireF1y644 He never said he wasn't willing to learn, you put those words in his mouth and since when were you the guardian of who can and cannot enter market. Get off your high horse.
@@LethargyPike you can't just "learn" to model. For a game designer/developer to also be able to create 3d models and textures, animations, music, or even write, at the same level as his main skill, he needs to have at least 2-3 years of experience in each. People just don't have that kind of time.
I learned to use Blender in 4-5 months, and I can make basic models or edits, but creating the high quality complex models I want (especially characters and creatures) is way beyond the skill someone with only a few months of experience can achieve in a short amount of time. I could probably make a good enough character model if I spend a few months only for this, but I am not going to do that when AI will be able to generate it within a few minutes using a base image or a text prompt.
I'm a 3d modeller, why else would I be here right?, but I am one of those who adores the time saved by all this A.I. generation. I've dabbled previously and it's always been really janky and the cleanup can be longer than the modelling from scratch. But now it's getting good I would much rather have the ability to prompt my 3d modelling rather than spend months making assets for a game that sells 100 copies.
I don't think it will replace them, but it will make their job a whole lot easier
If a non 3d artist can generate any 3d model he wants with such a model by just writing a text or feeding an image, no one will pay the artists...
The same with 2d artists, software developers, musicians/singers (because Suno and Udio exist which make much better music at this point), writers, etc.
Early Sunday morning content is the best. I love it!
Since I'm a professional 3d modeler and AI enthousiast, I've been noticing all the 3d polygon generator, and since a year or so there have been some before NVidia, such as Meshy, StableProjectorz,, Solace Vision, Latent Labs, Common Sense Machines, Spline, 3D AI Studio, DeemosTech , some of them might be cloud mesh and some might be dead by now. I think you can even generate the texture with 3dFy.
@@Dreaven.Mehendysis looks like us 3D modelers are on the chopping block. Corporate companies will consider their bottom lines and your enthusiasm for ai will be replaced by lerning to work fast food. Ai development will eventually be taken over by advanced ai, thats literally their goal, so what will you do? What would justify a good salary when they can hire anyone to do the job?
@lazerith840 @lazerith840 As any other job. The problem will be being able to not fall into leisure rather than many-parts-of-the-population possessed enterprises. Seeing the results of how AI is received in the consumer world, AI made produce do not have a good prospect. Just look at how Coka cola AI ads reception. Working with AI everything I can assure you that the results are ridiculous and time consuming. The advertisement for it does most of the job, it's all glitter and file gold.
@lazerith840 *population-owned-corporations
@@lazerith840 the hardest part of being replaced by AI will be not to succumb to the leisure time that an absence of job will produce.
@Dreaven.Mehendysis absence of a job just means you will be broke. You can live with that, but I got bills to pay. Your so short sighted that I kinda want you to find out these things on your own.
considering the fact that AI has very limited context length due to our hardware limitations, I heavily doubt this will be more than a gimmick where it can make relatively simple shapes with high accuracy but incapable of making any game ready or animation ready assets, and when it needs to do something semi complicated like fingers it will probably mess up way harder than 2D image generation ai's.
I would rather have them work on some proper AI retopology but who am I kidding, they can't even do that and they expect to do the entire workflow, good luck with that lmao.
so they auto ran ran quad remesher after generating their original mesh. so as a 3d artist im constantly struggling with making real time meshes. this could help me become more productive and that would be cool. i could run this on a sculpted object and get a great start on a render baking mesh. so i think it wont replace a 3d artist just make us way more productive and think that would fit in an area thats needed. these models they are genning are very topologically simple to parse from a photo or many photos which is fine. itll be interesting if they can scale the complexity of these models.
I wonder if they have a professional 3D Artist giving feedback on the topology, I will still say it is pretty bad topology, especially for organic models. On top of that, 3D model needs uv, characters need rigging, there is still a long way to go. 3D is way way more complex than 2D.
topology is very bad, the texture will be hard to map. and if you use a lot of these, these will bloat your game or app. you also cant 3d print these as these are point cloud generated which will have a lot of lose points and disconnected vertices
World mindset about AI: "as long as it doesn't affect me" ¯\_(ツ)_/¯
it affects everyone
@@SmartPoly But people don't take a stand against it until it affects them. This is the great era of selfishness, individualism, and hypocrisy described in the Bible as "the great tribulation," the period that would precede the end of times. If I only cared about this issue on the day it affected me, I would consider myself unworthy of being saved.
@@carlosrivadulla8903dude, we can't stand. Everytime in humanity history when we tried to face technology advancements we lost.
automation existed before A.i.
I create much better 3D models on my channel than any of this AI generated JUNK! 💪😎
Jobs come and go-that’s the nature of progress, and you can’t stop innovation. If AI were capable of performing heart surgeries, some people would lose their jobs, but overall, it would be a significant benefit for society.
You say that until an emp or solar flair flips the switch, or the electrical use causes more coal to be burned by power plants to power the machines on top of power our homes. Jobs do come and go, but that doesn't mean it's positive. It has negative effects to economy. Progress doesn't always bring positive outcomes. As more people lose work, then less people can afford to purchase things. The less people can afford to purchase, the less earnings a company can make. Ai is going beyond just digital work, and I lived in Detroit and seen first hand the downfall of that city which started once robots were introduced into factory work. Before jobs used to come and go, but it still involves humans, this is different. It's literally removing ourselves from the equation all together. Not all progress is good.
@@lazerith840 well said. so many do not understand that what we choose to do with technology matters. its a choice its not some autonomous destined thing. we choose. really good book on the subject, power and progress our 1000 year struggle over technology and progress.
Actually you just need a single regulatory vote to stop it, during psychedelic research in the 70's the future of psychiatric is obvious psychedelics for most in the field, where is the psychedelic research now?
@@lazerith840progress doesn't, but productivity does.
Never before in human history some technology that increased productivity brought misery and unemployments, but increased our quality of life and wealthness.
Let's Hope the rule keeps, if AI brings more productivity, so we can expect all those things to come again.
@starvosxant4348 Detroit was one of the wealthiest city's in the 1950's. Then the introduction of robotic machines doing factory labor caused unemployment leading to other social economic impacts in the 1960s. Over the decades it has gotten worse to the point that the city went bankrupt in 2013. You think all those people losing their jobs wasn't a contributing factor to the decline of the economy and social structure there?
So in reality it seems like the only time a new tech caused issues, was in automation, just like ai.
Coincidentally I lost my 3D animation job to ai in Detroit, and had to move to Louisiana for a new career path.
Recently started teaching myself UE. I figure by the time i get proficient enough to do something with it people will be able to make a full game with a simple prompt.
doesnt work that way. In order to create something u need a deep understanding of what u are doing. Ai can lend u a hand but it wont make all the job for you. Manual crafting wont be replaced. So i suggest you to stop looking too much into what the new tech is doing and put more time learning the basics.
@Damian_DH for now. AI will eventually take all jobs, its just a matter of time. Its naive to think otherwise. It might not happen in the time it will take me to learn unreal but it WILL happen.
@@Damian_DH bingo...afterall if its a passion go long and go deep
I'm not sure that there's too much magic going on here if you understand the 3D modelling process. Once the point cloud is generated, it understands from referencing information about the object online, breakes it into components based on this info, and then uses regular and long established auto-retopology algorithms (like quadremesher or instant meshes) to generate the topology. It's like a more informed version of photogrammetry with the manual steps removed. I'm sure it will keep improving, but I also think it will reach a limitation plateau. It won't be replacing 3D artists any time soon.
Agreed, except i would replace "soon" with "in the next 2 years" . Many of my web dev jobs have been replaced because it's easier now than it was 15 years ago to build websites with the tools and Ai now available. The same thing is already happening in the already incredibly competitive 3d jobs market. It will get worse. Some artist jobs will disappear, not all.
@DirkTeucher Some jobs will be autimated, no doubt, but the tech will be most useful as an aid to artists, rather than a replacement. It might reach a certain level of quality regarding 3D meshes, but I imagine there will be a quality ceiling, and artist input will still be needed after the initial generation. The non-artist 'AI bros' will always be unskilled tourists. Otherwise, if they had the skill they would be artists.
@@musashidanmcgrath
What the aritst will use this for lol?
to not become an artist
@@Pixelz_3D You may have missed the point. Are you a 3D artist?
i think its more intricate than quad remesher, at 5:12 the box is very optimized but using a quad remesher it will definitely have more polys because it treats the whole object fairly to account the small details.
hi poly. i wanted to thank you for your work. FYI, I bought your survival course about a week ago (looking forward to time with it over xmas). i was a little nervous about getting the course as you are working through the remaster, but after going on discord I can see you update the new classes for the course about every 6/7days. i think you should push info like this when promoting the course. think it will help peopleon the fence. maybe also explain the 2 options you get when buying from the webpage. hopefully it helps you with conversions. oh, and a suggestion. maybe PCG cavern system.... may be interesting. i will be starting one anyway at some point. have a good xmas, and see you next year.
nice and all but you cannot use this. where is the release website?
Feels like segmentation, classification and topology analysis with machine learning on top to guide.
Are you kidding me ?? We already have tools like this for almost a decade! We use these Auto Retopology tools all the time. I don't see this being any different from them, maybe better at avoiding infinite loops but it isn't anything new. Maybe its just for views hence the misleading title. I think now a days people can simply write AI and Replace to a video and the Algo will do the rest. Chillax guys if you know a single thing about 3D creation this is one of the most used feature that comes with almost every software you use.
Yes we have, and all the noobs that never worked in the industry will doom fear us to death. This gonna be so busted
I think you're missing the part where the source model is CREATED from nothing but text or an input image. The remeshing part is just on top of that.
@@jenkem4464 The source model is not made by this, If I am correct. BTW Source models have been done through Photographs for a long time. Regardless I will actually be happy that a tool like this comes. Right now topology is a nightmare no 3D artist likes to do and all we wanna do is get rid of it as soon as we can, because its not art its just a roadblock we have to go through.
@@sauravbhagat4150 Oh Gotcha. Yeah, if the source model isn't made by this then I get it. It still seems to do a pretty good job. I don't know much software that can separate geo like that and still have as much mesh-topo control as this. This + Microsoft's new Trellis software will be pretty cool, at least for getting a decent base for people that hate doing all that. I love the sculpting and detailing part, but not the mesh parts. Can do it, but just not as efficient as other production artists.
It will not replace artists it will make them more efficient and their job easier, its a tool like a drill vs screwdriver.
@@ChunkyDeLumpy tell that to the concept artist who have been replaced by midjourny.
Ex: Did the iPhone eliminate professional photography ? Id argue it grew the industry significantly
@ChunkyDeLumpy because that is still a camera. That is not a good comparison. But the camera replaced portrait painters.
@@lazerith840 And those replaced Portrait Painters later became the concept artists & other types of Designers, so they weren't totally replaced. Similarly there is still hope.
@@manvendrapratapsingh1920 there’s a difference between getting replaced by a object doing part of the work, and the object doing all the work. I’ve already lost a job to ai, so has others. Maybe you will understand once it happens to you. Working in 3D is already an uphill battle, and needing less just means less jobs for people. These companies are not going to stop until you can type what you want, and it makes it. Anyone can type, so they will hire less, and pay less. A career, has turned into a job, and y’all can’t see it.
When I look up MeshAnything v2 examples online the results they show are a lot better than what's represented here. The triangles for MA2 are also a lot more regular and less crushed, which make them better for UV mapping. The quads output for Meshtron is pretty intriguing though, but I think everyone is just cherry-picking whatever examples makes themselves look good.
I wonder how well AI is able to adapt artstyle.
art style is the easiest for it to mimic, if enough reference is shown
How do you get the point cloud info from concept art?
3D modelers for lowpoly stuff are basically obsolete now. I suggest getting into Houdini and learn creating procedural systems. That´s what game development, VFX and advertising needs the most. Worldbuilding, Procedural animation and modeling. But static meshes and even characters can now be prompted or photoscanned, autorigged and even procedurally remeshed and optimized, auto UV mapped and procedurally textured. With the exception of very specific requirements like product visualisation, architecture or character design, a lot of tasks nowadays take much less manual labour, than even just a few years back. Only putting all those assets together into dynamic systems that adapt to the users requirements and save a lot of work are of significant value nowadays. Because that´s the only stuff AI struggles with. For now.
And on the other hand, it is as easy as never before to access complicated software like Houdini because of having AI learning assistance and AI assisted code generation. It will not have an answer to every question, especially when prompted for something complex in one go, but it helps breaking down processes, helps understanding basics and functions, can be fed with additional information and documentation and even scientific papers. I have tested that by myself intensely in the last year with some very interesting results.
Go procedural!
Ai developers should consider the social economic impact of its use and if it's causing more damage than good. Imagine going to college to learn all this, then still having to work fast food, until they get bots to do that too. They are pretty much forcing communism because who will buy products if more and more people are pushed out of a job. Replacing ourselves seems like a bad idea. It ignores the social impacts on society for corporate greed. Even the developers of ai will eventually become replaced by advanced ai, and then what. The willingness to accept long-term loss just for short-term gain and ease of work is not only short sighted, but also counterproductive in the long term. Imagine building our society around ai, then a solar flair or emp flips the switch, and nobody knows how to do the work on their own or how the process works. People tend to disregard these questions and sometimes behave hostile twords these questions when, in reality, someone needs to ask it, and it needs to be addressed. I personally like ai tools, but if it reaches the point of replacing a human job, then is it really beneficial.
Peoples are selfish and will only understand the impact when they themselves will be directly affected in their careers.
.. :/
It’s already started replacing human jobs…
Not impressed at all. Most complex stuff in 3d modelling for gamedev is not to create a low poly mesh. It is about finding the correct art style, the balance between details and game performance, it is about creating proper topology (meshoptimizer/Nanite/LODs), it is about UV unwrapping, it is about texturing, it is about animating. This tool does the simplest and least time consuming part of all 3d modelling. Good luck for all the folks in comments that went to "wow, I can create a 3d game now without any knowledge in 3d modelling"! Maybe you should have some, so you won't get so easily excited.
4:00 - Because AI looks at references on the web. I'm guessing first it decides what the object is, then he uses the internet to figure out anything it's not 100% sure of.
Is that published to download Nvidia Meshtrone?
I hope one day they feel a duty to protect us,
When we're a lower form of life.
Within 2-3 years, anyone can make games alone.
This is more of an ai photoscan meshing system not ground up model generation
YOU CAN'T KILL ME
Imagine how this will be in 5 years!
Amazing!!!
Without testing it difficult to have an opinion.
I do 3D modelling since 20 years, At early times we are the 2nd generation who created 3D modelling, texturing, rigging algorithms, I do support AI do these boring modelling things ,Even all stuff... Because every game, Every movie Starts with an Idea, So AI may make my Idea accomplished, If AI will replace all human with Idea creation too and will decide to get rid of Humans out of their way then it's OK too for me...
Very promising but it needs a couple years more.
No it's moving super fast, give it a year.
@@alteredcarbon3853 It's not moving super fast. It's just moving. It'll take years until it's production ready. But I couldn't carel ess. I do my own models and am not afraid of being pushed out by AI. It's meant to support artist not to replace them.
I started learning Blender at the worst time.
Bro is against himself 😭😭😭
last time I 3D modeled something was years ago
I will be impressed when they can make good textures. Until then, it’s all academic.
Get rid of the tri's make it all quads !
This al helps with generic stuff so making background stuff and side characters will def be much faster...
U still need a artist to create the idea..
the tool is only as good as the user
It's not AI you hate it's Capitalism
No capitalism is the best thing ever
@mothurman it's the best at exploiting you and under servicing you and gaslighting you into endless service, yep, the best
Smart poly prev. video: lmaooo. If you think this tool will replace artists, you’re cooked.
Smart Poly in this video: Will This AI Replace 3D Artists?
So what's your real opinion?
my opinion is what I stated before. The title is to provoke thought
dam
Gg pro
All human jobs are going to go away, it's just a matter of time.
No, they're not this meshron thing looks terrible, and why use images to generate the model when a professional can make one with complete control. This type of AI is going to be used for people who can't model but want to make their "dream game."
@@LethargyPike lol, you clearly have not seen how fast AI models are getting better.
OpenAI released the o1 model a couple of months ago which was the best AI model in terms of logic/reasoning, and a few days ago they released the o3 model (they skipped o2 entirely) which crushes every benchmark that existed in order to measure AI model capabilities, often being on the same level if not better than even human EXPERTS.
By the end of 2025 3d model generating models will crush the best humans, and they will do that at a fraction of the time and cost required.
@OnigoroshiZero For $200 dollars a month, you can have access to ur AGI, which wasn't necessary to begin with.
I'd like to see where your predictions go into next year I guess everyone's just going to get laid off and GTA 76 will come out because these AI'S are going to be able to create every single thing because they are "Experts" even if this occurs, I dunno why you're even remotely okay with that, why are you okay with AI replacing you?
Sincerely, that's leaning a bit too much towards delusional thinking. OpenAI floated the idea of charging $2,000 a month for the top models. They're burning through billions and billions of dollars to bring everyone the latest and greatest technology. Sora is $200 a month and people are already complaining about that. Devin is $500 a month. The idea that these companies will freely release the very best "job replacing" models for a fraction of the cost is silly.
I just have to wait a few more months to be able to use such models in order to finally make the assets I want to complete my game.
AI is amazing, and it only gets better.
I am a 3D animator and none of these examples shown have good topology for animation. If you need to make static objects then I understand.
Wow, NVIDIA's AI turning text into 3D models is mind-blowing! The future of design and creativity just got a huge upgrade-imagine the possibilities for games, movies, and even VR. Truly next-level tech!.
Are you a bot?
Don’t hate the AI. That’s like hating a hammer. Learn to use it to your advantage. Increase your production pipeline. Get more done. Don’t complain.
@@islandonlinenews tell that to concept artist. Oh wait they are working fast food now because midjourny does their entire job. There's a difference between a tool, and a replacement. Ai to help with topology or UV unwrapping is a tool, ai that creates the model makes the textures and has multiple LOD's just makes someone need to find a new job after years of study. Nobody hates the ai, it's coding, but the developer need to consider the social economic impact of it. Short term ease for long term loss does not seem beneficial in the long run.
All you doom fearing never working a day in this industry worms. All I see is more room for work with real talent, you think people with no talent make anything of quality. All we are gonna see is an over saturation of Crap on art Station.
have to be naive to believe that in the coming years AI will not replace 3D artists , because the pace of AI development is accelerating and will only get faster but faith dies last of course. And to speak out against such technologies because of the possible loss of work and money is too selfish, you can't slow down technologies if they are for the benefit of society, otherwise it looks like aggressive capitalism where monopolists temporarily bury technologies for the sake of profit, and life goes on and we are not getting any younger, the more things are created that simplify our lives or save time, the better for us, society is more important than the desires of individuals who are afraid for their jobs, many professions have sunk, you can't knock them out and this process cannot be stopped
the more I follow this, I'm starting to get convinced we are possibly at least 5-10 years away from actually usable 3d models as a professional modeller might do and that's a big IF. While these look "promising" they are nowhere near usable for actual production work.
I think modellers don't have a lot to worry about. It's easy to make a 2D image with a girl by analyzing a million other pictures of a girl, but making a production ready 3d model that's high quality is goint to still have to be made by humans. If Autodesk or Nvidia or even Google can't make this work so far very few others have the capability of doing it resources wise or smarts wise.
Hey poly please make a tutorial about playfab for prop hunt course
you like to spam the same comments under each video and in the discord server
In answer to the video title: No, there are too many avenues of specialization and bespoke use cases that cant be brute forced or relegated to pure math in game dev. None of the models put out by this even shade correctly, the topology is awful. Even if the AI could do perfect wonderful ideal perfect topology recreation, it would still need a solid high detail authored mesh to work from, not this blobby point cloud shit thats currently spat out.
If anything this tech might be a decent retopology system used in conjuction with photogrammatry or automatic lodding.
I do this for a living, and I wouldn't considering anything spat out here anything than slightly below student work. "But itll keep getting better!" Yeah, at retaining detail from a bad source mesh, and then once the source mesh is gone, it'll retain detail with what level of control, with what level of optimization, how do you tell this to UV things in a way that lays the UVs out appropriate for a flow map, or for mirroring, or to prioritize surface area texel density based on approximate visibility within the engine, then comes art style and exaggeration. There are too many variables that you NEED control over to leave this in the hands of an AI, unless you want unoptimized messy shit.
sometimes nvidia would release these research papers but never release them for public. the tease 🥲