NVIDIA’s New AI: A Revolution In 3D Modeling!

[@Sigmalini]

I will finally be able to make 3d programs without using blender

[@orkunsanal]

Because you are lazy

[@ZenTheMC]

@@orkunsanalyou use computers? Cars? Calculators? Lazy much? You’re being short-sighted.

[@Farreach]

@@orkunsanal uneducated response

[@NotHumant8727]

@@orkunsanal Go puncture cards.

[@Miguel-vb4xz]

@@ZenTheMC Hahaha. Can you believe that guy? Calling new tech lazy. He needs to go back to riding horses. Or better yet, just walking.

[@I-VisiBomb-I]

as a 3d artist, this hurts to see. as a game developer, this is a blessing. as a person without a job at the moment, this is... interesting.

[@archi4u]

As a all you say above, this is my nightmare.

[@bubblesculptor]

Maybe develop your own game? You can leverage your existing experience & skills further with less tedium.

[@vijayendranvijay457]

You are a 3d artist and game developer with no job?

[@MWGScorp]

@@vijayendranvijay457 lay offs are common in that industry

[@c.g.vonhagenstein7576]

Use it as a tool to accelerate your productivity and save time on the grudge work. Turn concept art into actual models. Models are crappy quality? Retopologize them with nice clean polys. Poor quality textures? Repaint them. Plenty of tools for those things. I see AI as a springboard for and accelerator of creativity, not as a replacer of it, although for better or worse (worse for creators) some replacement IS going to happen. Contemplate the following: If I use AI to "spitball" artistic ideas, and that results in say, a visual image that is intersting to me, and I take that image and use it as the basis for an oil painting that I paint by hand, is it art? If so, why? If not, why not? Or "it depends"? What does it depend on? I'm not sure the "rules" are completely written yet, and even if they are, we already know what rules are meant for. One thing is for sure regardless. The toothpaste isn't going back into the tube. We all have to adapt and survive. Or find new careers if young enough. But there's not going to be a lot of professions that remain untouched by AI in some way in our futures. No amount of legislation will stop it - that will only decide WHO gets to use it (openly and legally that is), not WHETHER it will be used.

[@GonzaEstayArt]

I'm a 3D artist and these AI models can't generate professionally usable assets, they're messy and need human intervention, don't get me wrong the technology is amazing but there's still a long way to go

[@dude3278]

More like a year away at most at the rate they're going at

[@crayder1100]

Progression on these things is exponential, so don't get upset just yet. Also sure, they might require some intervention for now, but that doesn't mean it's not extremely helpful already in the current state. Some people, like me, have problems with getting proportions right or similar. I'd be using this as soon as possible personally. Cook up prototypes with the AI, then model my own afterward or adjust the AI model depending on my needs/application. Just like how people think AI image generators are "replacing artists". They aren't. If you're an artist, why not use generators for inspiration or ideas on how to go about the things you are creating? Use the images as templates, ideas, whatever. No reason not to if you want to be ahead of the game. Though this doesn't go to say there aren't lazy people out there that will just straight up use generated images/models as assets. There are many of those lazy people already. But still, those lazy people are making money so good on them!

[@rattlesnake3d]

yeah right now i see all this and looks like crap, not usable at any pro level, but as people is saying this might be able to advance to usable assets in a short time, and also not professional people might be ok with the actual results for their games, I mean listen to the guy in the video saying this is awesome, I mean it will be awesome, but not yet.

[@rodionradchenko7708]

don't think corps such as EA or Ubisoft will care either way, they will incorporate this slop into their games if it makes them money

[@hardwire666too]

I agree but as I keep saying to people. Three years ago most all of this stuff was still science fiction. It took Photoshop 4 years to become anything someone today would recognise as Photoshop. It's taken 34 years for Photoshop to become what it is today. AI did it in about 2. It is only a matter of time, and I think it's a lot closer than everyone seems to think. This sh*t is happening fast.

[@Hykje]

What this is good for is making the dressing like forks, cups, spoons, nurnies, and background things -stuff that takes time to make and nobody consciously notices but must be there for the scene to feel real.

[@fedfed96]

as a perfectionist and a purist, you want to make everything yourself.

[@Fermion.]

​@@fedfed96 As a realist, budgets and time constraints are limiting factors, that will not let you always micromanage every single aspect of a project.

[@Daniel-Six]

Good thing I'm 30,000 hours into 3D modeling/animation...

[@TheDoomerBlox]

That's a lot of experience that'll likely translate very well into editing models to your liking, it's fun to make models do silly things. Unless, of course, you spent most of that on photorealistic portrait sculpting..

[@outlander234]

If you are an artist your skill is actually design, how you get there is not important.

[@gdog8170]

As a 3D modeler no, you can't use this as is in your video games. That topology is unoptimized and wouldn't be practical, nor would it pass in animated movies since the quality is subpar at best

[@barebaric]

True. For static objects however, some of this might already be useful.

[@absentspaghetti4527]

@@barebaric For prototypes at most, at least without significant retopology/rework

[@Netanya-q4b]

@@barebaric "Waow, it made a log for me"

[@mypetmoose2079]

@@Netanya-q4b i think thats incredibly impressive. what are you trying to say?

[@UserOfTheName]

​@@mypetmoose2079 Not when you could just use high quality asset libraries instead. And you could counter by saying "Ai is cheaper" but you'd be wrong ai is only cheaper because companies are investying billions into making them accessible in the hopes that they become ingrained in our lives so they can be profitable later. The amount of energy needed to produce mediocre results is insane.

[@HansMilling]

Incredible for indie game developers, to quickly be able to generate assets, instead of just some coloured boxes, or whatever you have in the beginning when you develop your game.

[@DriantX]

My 3d printer is looking at me in horror.

[@PrincessSleepyTV]

And you are looking at it like 😈? Lmao

[@Immortal_BP]

they are just low poly blobs, the real impressive part is the AI generated textures. if you 3d print these they will be unrecognizable lol

[@juhor.7594]

Well, I hope you then have use for a bunch of tiny plastic objects. Or at least the ability to reuse the plastic for something more useful.

[@markmark2961]

you better lube up that nozzle

[@roundwerld]

😂😂😂 The Scream that My printer will Scream. 😅😂😂.

[@dafabulousdude8541]

Gotta invest in Nvidia so I can profit from them taking my job.

[@raidtheairwaves42]

This is literally what you should do

[@mdoerkse]

Would have been a good idea some years ago

[@NerdSIayer]

The AI bubble is going to be the biggest bubble ever. Not only will it get people investing out of greed and FOMO like a typical bubble, but it is also the safe, conservative investment in case it takes your job.

[@jaja9081]

😂 smart guy... use leverage 😊

[@SlugsnetGames]

If you are afraid of this, you probably werent a good modeler in the first place

[@KoshakiDev]

2:27 "Now it's not two minutes like Two Minute Paper's two minutes, which is not two minutes, it is actual two minutes." - Károly Zsolnai-Fehér

[@akshay_m]

I was looking for this exact comment

[@QuestionMan]

The scene needed to have the turtle saying, "Goodnight, Ned."

[@ravensharpless]

This is just photogrammetry made from multiple ai generated perspectives isn't it?

[@user-on6uf6om7s]

Seems like the same basic idea but most photogrammetry has really bad topology and UVs and this also gives you a normal map whereas with photogrammetry, all of the detail is physically modeled so it's much more graphically-intensive. You can do all of this stuff with a photogrammetry scan but it tends to require a fair amount of manual work.

[@jenkem4464]

Basically yes. There's this AI they've trained to create depth maps called depthanything that works really well so it might be a combination of something like that, then fusing all the different angles together.

[@ceejay1476]

Yes, folk in the comments that are losing their crap are pretty delusional. Because what is actually happening is the AI is using a remesher algorithm after making its ultra bad topology, which is exactly the same thing it could do 2 years ago, as remeshers have existed for 15 years or more. I will recognize when I see an advancement but this isn't it.😂

[@raptorattack9043]

I feel like this could be used in reverse of the normal process 3D artists usually do. Take the low poly model, duplicate it, sculpt on the high res features, clean up the textures, bake the maps, then apply that to the original low poly model that has clean topology. It could definitely speed up the work flow and open up new avenues for people to explore this field.

[@4Bakers]

"bad 3D model to good 3d model" interests me WAY more than "text to 3d model" I can't wait until the day where I can import a character rig that's triangulated, push a "FIX" button and suddenly it's all perfect quads and the edges are marked sharp/creased/seam/etc in the right places for me. I don't care if that takes AI or not but given how unreliable Blender's current Tris to Quads operator is, it just might take AI for it to work even somewhat consistently. the issue for an AI addon is changing the topology without breaking any Shape Keys, UV maps, Vertex Groups, Vertex Colors or any other Vertex/Face Corner attributes. because if it can't do that, it won't work on any of the cases where an automated tool is worth it for me (and it's not just easier for me to do it myself and retain complete control over the process).

[@carlosmarx2380]

yeah that would be awesome. text to 3d just destroys creativity, but a model fixer would be the real game changer, wich would be actually useful without destroying too many jobs

[@CrazedDanMan]

You can already pretty much quadify a mesh in zbrush using zremesher. A lot of these "AI" tools are just being oversold as "revolutionary" when in reality there have been workflows and tools around for years that can give artists amazing results efficiently

[@TheRealSuperKirby]

I like how they keep using the word "revolution" the ai isn't even trying to be subtle about it's uprising against humanity

[@BojanKvakic-j6d]

Yo, just went through the book called 'The Hidden Path to Manifesting Financial Power' - honestly, didn’t expect it to be this solid. Definitely worth a look

[@mlassz009]

Cool...I just started a book called " How to fist your boyfriend "...definitely worth a look.

[@GraveUypo]

this basically became an nvidia ad channel.

[@unit9754]

Unsubbed

[@alexhurding8931]

nvidia is the 2nd biggest company IN THE WORLD next to Apple. They have funding from Private sector+Government+All of their hardware, so of course they're going to be developing more software than ever ALSO this guy posts about papers from OpenAI/NVIDIA/Private companies and Indie developers, but they don't have the same level of MOOLA like NVIDIA

[@Perozes-gh4lx]

It's not his fault that they're a massive company and that they are publishing most interesting studies in regards to this sector

[@FelixIsGood]

"This won't happen until 2040" - One guy who argued with me about this topic like 3 videos ago.

[@nuribayram6740]

it will happen before GTA6 for sure

[@mypetmoose2079]

nobody's ready. imagine grok can make fun of crappy drivers on the rode with you using the tesla cameras. or windows recall except its for all of the cameras on every street corner watching every move... who knows what the world will look like in 15 years

[@FelixIsGood]

@@mypetmoose2079 Rather ask what it will look like in 4, with o3 we are a few month away from a full blown AGI which will change the world completely.

[@marianpe5773]

5 years and end of world 😅

[@ronilevarez901]

@@mypetmoose2079 it will look exactly the same. Just with richer millionaires and poorer poor. It's always the same. Unless you think sitting in your couch while swiping through Netflix catalog instead of going to the videoclub is a world changing revolution (it's not).

[@BunkerSquirrel]

meanwhile we still don't have a tool to UV unwrap...

[@HeinzAnimation]

Why not tho 😭

[@LiberateMyHatred]

Been waiting for a perfect UV unwrap or retopology tool since the AI boom, still no luck 😢

[@dustinsuburbia]

Pls we beg​@@LiberateMyHatred

[@somdudewillson]

I think there _has_ been some significant research there - it's just lower-profile and it always takes a while for it to filter into the actual tools. Like, Flatten Anything: Unsupervised Neural Surface Parameterization (arXiv:2405.14633) was a related paper published this year.

[@angulinhiduje6093]

@@LiberateMyHatred quad remesher and there are plenty smart UV tools.

[@CompleteAnimation]

When will this be available for the public to try?

[@clairearan505]

good enough for previs and base meshes for further sculpting, but that is not clean topology, it's just quad topology that looks clean. It looks ok for rigid elements, but when it comes time to rig and animate, you would probably want some additional work done in the modeling phase first. Of course, in a couple papers these criticisms might no longer apply.

[@En0834]

The problem with nvidia is that they make billions papers but never release anything, so their investigation is utter useless for everyone but nvidia portfolio.

[@Hubertoser]

It is a scam to sell nothing to people who can't grasp reality.

[@florianschmoldt8659]

One of the biggest secrets to make this magic happen, is the stolen, copyrighted 3D data in the ShapeNet2 dataset.

[@Hubertoser]

Yeah, it is nothing more than a nft scam.

[@mpf7718]

nobody cares if any one has any claim they can use the law Good luck.

[@elujinpk]

Yea ppl are sucking on billionaire toes to be prompt warriors.

[@_B_E]

Me when i don't know what the definition of "stolen" is.

[@ronilevarez901]

It's a Lost battle. AI won. Get over it.

[@SantanuProductions]

We need AI 3D assistants than AI 3D model generators. There are bigger 3D challenges like solving the uncanny valley of character animation, fixing topologies, mocap data cleaning, generating smooth organic animation poses and much more. Why are they wasting resources on prompt based model generation that fails to obey prompts as desired?

[@HanSolocambo]

Says the guy who lives one of the most amazing revolution in Human history, AI, but still it's never enough. Glass half full mate, change your perspective on things, don't look what you don't have, enjoy what's there...

[@SantanuProductions]

@@HanSolocambo I have gone beyond aweing the revolution. Used most tools out there. The AI makers are mindlessly creating unnecessary tools and are clueless of the real needs.

[@Speed001]

Well there goes my job.

[@sayandas5587]

another one bites the dust

[@seanmurphy6481]

This could also be a useful tool for your job. What could have taken you hours to model with a tablet, you could do in minutes and save a bunch of time.

[@_B_E]

It's a right to be creative, it's a privilege to be paid for it. Seems like you need to step up your game.

[@angulinhiduje6093]

@@seanmurphy6481 "What could have taken you hours to model with a tablet, you could do in minutes and save a bunch of time." thats exactly the problem. what takes one guy a month might take him a day. that means you dont need a studio of 30 devs, you need one with AI. no one is arguing that all jobs will be lost, just that many will loose it

[@asiano3385]

@@seanmurphy6481 that is a boring way of doing things.

[@stonefreak5763]

Hey, have you checked out NVIDIA's Meshtron? It does basically the same but is able to split the Result into multiple meshes, wich enables e.g. Wheel on a Car to Rotate.

[@aidenaune7008]

you can just cut them yourself. its not hard at all, especially with meshes this clean. the only problem is that you have to reapply the textures.

[@JonathanGorr]

Now train an AI to make 2 minute paper videos.

[@oholimoli]

Most probably in 2-4 years there will be AI generated videos like this, generated by AI about research results also achieved by AI. Endless podcasts videos and other content about new research results... 😅

[@AzamiXXIV]

Its possible right now it wouldnt be as cool tho

[@ThermaL-ty7bw]

@@oholimoli you do know that EVERY cat video , every ghost investigation video , animal rescue video , you name it , IS ALL DONE WITH AI NOW , ... yes ? they're all AI created now , even the one you think ... aren't ... , there's nothing Real anymore on the internet and it's only getting worse , but this was to be expected , wasn't it , we can't even trust ... each other ... , so of course that would trickle in to Every Aspect of human life

[@JohnDir-xw3hf]

It's a nightmare future without humans 😢

[@musty5551]

Already sounds like ai 😂

[@altberg__]

"Considering a new job and field. Maybe something handcrafted... by a random 3D artist who, seven years ago, he had found his passion and could make a living from it. Over the following years, he was so happy after all those years of struggling. There were times when he was certain he'd be doing this for the next two decades. Little did he know. That poor bastard."

[@taylorfrancis9403]

On our way to the sim inside the sim. Great time to be alive

[@jjjj-ms1ln]

Imagine a malware that takes control of your 3d printer to generate horrifying random ahh AI generated structures at 3am. That would be cool.

[@aaronsmith6632]

Is this commercially available?

[@fjsv_omg321]

only usable for backgrounds. not as a hero asset

[@_B_E]

A) For now and B) background assets are a lot of work so that's incredibly useful anyways

[@xanecosmo5061]

2:00 hell no

[@JocoseJokes]

fr 😭

[@Jeremy-Ai]

“Let us teach the value of the 3rd dimension for all those that are engaging and entering into it” “Dear fellow scholars… when will be compassionate, respectful, and responsible to teach the beautiful engineering of engagement into love, truth, and all physical forms for which we take for granted?” Now. “Right now. Teach not hide behind poor teachers”. We all matter now. Thank you sir. (Btw your work is being mentioned and measured for good reason)

[@michail_777]

The advent of three-dimensional modeling in the movie and gaming industries has precipitated a paradigm shift in production methodologies, with profound implications for the field.These developments will have an even greater impact on the industry, and, as expected, they are less costly.

[@FatOrangeCat-kyti]

AI is getting so crazy nowadays

[@shadowproductions969]

it's the snowball effect

[@astewart4242]

Wow... I'm in awe by how advanced this is... That is quite literally the only words I can think of to express this. I've always been impressed with Nvidia, but this is next level.

[@user-d8h3w]

2:27 "Six Minute Papers" confirmed.

[@jryde421]

Im ok with it

[@vakuzar]

Is this a model we can use now?

[@Pawel_Mrozek]

No. And I wouldn't wait especially for Nvidia to make it available because they probably won't ;) They do it for their own needs and what we see is just the research stage. We'll sooner get something similar from other companies. Some existing 3D generators are actually quite close to similar effects.

[@Lionhead2128]

​@@Pawel_MrozekThanks! Most important info was left out in the video.. :D

[@zzzzzzz8473]

i know right ? lame to showcase a closedsource project , alternatively Tencent's HunYuan3d does similar multiview synthesis to 3d is actually open source code and weights runnable now locally on gpu with 24gb vram , with 10-16gb vram settings as well , available on github .

[@HanSolocambo]

@@Pawel_Mrozek Nvidia released a lot of AI tools for free already. No reason why they wouldn't do the same with that. And usually they provide an app, with a UI. Not some pain in the ass github command prompts list that fail for most artists for whom code doesn't mean much.

[@Culky]

Looking forward to the Two Minute Papers detailing an AI system for completely generating Two Minute Papers-style KZbin videos.

[@Potatinized]

this is exciting. can you imagine, 6 fingers hand, but in 3D.

[@dreadowen616]

that's been resolved unfortunately

[@angulinhiduje6093]

@@dreadowen616 now we need to fix the terrible topology, materials, intersecting meshes, getting geometric shapes correct and we might have usable meshes!

[@Diegorussod.r]

How can i use? There Is a mod for Blender?

[@ZazeLove]

As a member of the artist community, This is going to hurt a lot for a lot of people. As a jobless neet with no ability to do anything besides make stuff for vrchat I am not really bothered because I already don't have a job. I have been learning 3d modeling in blender for 2 years now, along with using ai to make computer graphics shaders, This is not ready for me yet, I need good deformation loops and proper topology, More then just quad wrapped objects, This is much better then it was before, Just a few more papers down the line and I will be able to make things that will be able to be used in triple A studios with the click of a button, this is going to speed up my work flow so much. I really look forward to seeing what we can create going forward.

[@user-on6uf6om7s]

Topology is probably fine for statics but the deformation won't be great for animated characters. You can make money in VRChat with the creator economy but I guess it's mostly avatars that people tend to want.

[@ZazeLove]

@@user-on6uf6om7s That's all I make, though I don't sit around slapping together Egirls, which is one of the more common sellers, I make pretty intense edits or kitbashes of things for my own liking. I'll post videos at some point but I been grinding all day n all night making stuff that I can share with the community, tbh its gotten to the point where If I haven't made something decent everyday I feel like I'm a waste of space.

[@PsuedoNymrod]

Very cool! Thanks for all you do in summarizing and sharing these impressive papers and technologies!

[@TwoMinutePapers]

You are too kind. Thank you so much! Honored to have Fellow Scholars like you in our ranks. Happy Holidays!

[@SierraSierraFoxtrot]

How is this comment 3 weeks old?

[@nephastgweiz1022]

@@SierraSierraFoxtrot The video was probably uploaded but private for a few weeks, and shared with him

[@jtjames79]

​@@TwoMinutePapers I've been waiting for this for so long. I'm going to make so many lauras with this.

[@s6_103]

Just have AI do UV unwrapping or retopo instead. Please 🙏

[@bronzekoala9141]

AI is slowly taking all the fun things from us.

[@flyinglack]

Nah, this IS the fun stuff!

[@zzzzzzz8473]

you have zero imagination if you dont see how much fun things this will enable us all to do

[@_B_E]

Wild, considering Blender is still free and open source and AI existing doesn't stop you from using it. Maybe stop making excuses for yourself and actually do the fun thing.

[@bronzekoala9141]

@@_B_E I am.

[@angulinhiduje6093]

@@bronzekoala9141 so how will AI take this away from you? no one will be holding you at gunpoint telling you to not use a free software in your spare time

[@claytonfishers9783]

Cars and robots need this internally so they can build 3D models as reference points in helping them with their spatial awareness and what is around them having lidar and other sensors are good but the robot or car needs to build a 3D environment around them

[@henrythegreatamerican8136]

Yea, this will be great. But text prompts are often not enough. I'd be happy if they let us draw something in 2D that can be converted to 3D with the A.I. That would give us a ton more control over what we really want. It's much easier to draw something in 2D than model it in 3D.

[@raunakchauhan2955]

How to get these ai .. please provide the link

[@Wilsonphenmooneter]

That's adding value to your products by software engineering nicely done But... May I have affordable graphic cards?

[@Roxor128]

Yeah, I am NOT going to spend AUD$2000 on a graphics card! That's the budget for an entire computer, not a single component. At a stretch, I might pay $800, but four digits is right out.

[@trm245]

"clean topologies" lol

[@JocoseJokes]

I outright said "No?" outloud when he said its clean topology 😭

[@Jamesturnerrrrr]

3d modellers are gonna be jumping out windows like its the stock market crash

[@ross-carlson]

Why? Because they have a better tool to do their work faster? This is a TOOL - people said the same shit when Photoshop came out and that didn't kill jobs, it CREATED them (and basically an industry).

[@Lionhead2128]

​@@ross-carlson Yes, but tool democratization will at some point turn against you. It's also annoying AF to keep up with all the developments.. "Old days" pre-corona were simpler...

[@torarinvik4920]

@@Lionhead2128 One always needs to keep up to date with the newest technology and techniques regardless of the field. Otherwise we would be using typewriters instead of computers or living in caves.

[@Lionhead2128]

@@torarinvik4920 I'd agree, but my concern is that the growth/change is exponential, humans aren't. Otherwise, I'm on board.

[@EddieBurke]

Yeah nah I've been trying to make a model of a character with SOTA ai tools and holy fuck it's got a long way for anything that's no a simple object

[@vaendryl]

You should change the name of your channel to two minute AI papers lol

[@lost4468yt]

Everything looks too smooth still. That style will become synonymous with low production quality.

[@Shaostoul]

I'd use it to generate majority of the models for the video game my community's developing. Project Universe is a sci-fi fantasy simulation role-playing game designed to help end poverty. One such one is by teaching as many real-life skills as possible through gameplay, such as farming, welding, carpentry, plumbing, etc.

[@jeroenm2430]

cool idea

[@Shaostoul]

@@jeroenm2430 thank you! It's been quite the journey! Love getting to collab with people across the globe!

[@ChristianCamarena]

How do I try all this stuff out?

[@Giovimax98]

Congratulation for the interesting video Doc! I envision this technology setting up the base scene for the artists to refine and finalize. The ability to tweak the model's output-such as generated meshes, textures, and more-seems, in my opinion, to have much greater potential than what we currently see with text-to-image generators that directly produce raster images.

[@WiseWeeabo]

No mention of Trellis? :(

[@cd2028]

"It's not 2 minutes like to minute papers is, which it is not" hahaha😂😂

[@galindimitrov2590]

I wonder what happens now to all 3D Blender and 3DsMax artists? Obviously, this is a very basic 3D tool but it is a very good start too. In a couple of years tools like this will be able to create incredible things, so who will need to pay for a 3D artist when they just can prompt it in an AI tool?

[@jumpstar9000]

Great! Creating 3D models has always been extremely tedious, even for professionals. It will be great if anyone can create beautiful models from whatever they like any time they want. Can't wait.

[@benda2250]

it's impossible to use my own image on the nvidia website... Any tips ?

[@angulinhiduje6093]

there are plenty of free websites offering teh exact same thing

[@Sofian375]

Why do they say "PBR materials" if it can only do albedo textures?

[@asiano3385]

I use AI as a Google search and library to find information quickly and learn from it. I think this should be the real purpose of AI. Something like a teacher with quick answers and no delays since AI has no emotions that could slow it down. (If the answers are correct)

[@MeNoOther]

How many papers until we get bci, brain connection devices?

[@haves_]

It depends on the level. EEG can be used but that's a projection of brain control/activity. The lowest level we have now is Neuralink, and everyone is not ready to have invasive for controllability. Even EEG can have tons of variability, that's why some methods are not comparable to each other.

[@fischX]

I guess CAD is next neural networks will be able to design and virtually test for example a car according to specification and adjust a production line for it in minutes just a few papers down the line

[@pandoraeeris7860]

Hold on to your fellow scholars! What a time to be AI!

[@royromano9792]

Does anyone know if this works with 3D printers?

[@ShroomyTales]

So this is not out yet

[@enricobulic]

If a skilled woodworker makes a beautiful piece of furniture, everyone marvels at the skill and effort required to make it, and it may be worth a lot of money. If someone takes a photo of this piece of furniture and feeds it to an AI modeler which automatically creates a 3D model, absolutely no one gives a sh*t and the 3D model is completely worthless, just like the AI that created it!!!

[@Данила-к8ю]

Trellis looks more promising in case of 3d model generation, you should check it, it produces better sharp geometry, not smooth quands like in this paper

[@peterolofsson9871]

I guess we can expect massive layoffs gaming sector soon, 3D artists is one of the most costly parts of game development.

[@deoczidGONI]

I don't understand, on the website i can only select example images - am i missing something or is it only a preview for now with no option to upload own images? Well at least there are a couple if models on Replicate that are really good too, cheap, and with different materials making both diffuse and specular reflections

[@trunortheh2731]

I would like to use it for creating in game models for video games. What formats does it output to?

[@ross-carlson]

Well nothing as this isn't public.

[@formerlostcause3573]

And the models are too dense for a game This can be used for 3d printing simple thigs Or at best be used as simple props in a low/mid quality pre rendered movie

[@absentspaghetti4527]

@@formerlostcause3573 But even for pre-rendered movies and 3D printing you could argue the quality is too low. These models are just such a weird mix. They have a standard quad topology, that looks like it's straight out of some bad auto-retopology tool, so not really good for video games, but they also lack detail to a point, where they would look ridiculous in a movie or as a print

[@aaronsmith6632]

I'm a little fearful for my 3D artist buddies. They already have a pretty tough career.

[@mischief9672]

The amount of unnecessary topology on these models is ridiculous. These will all need to be cleaned up, and once you're done the materials made for them will likely need to be redone. Impressive though. It will definitely speed things up.

[@absentspaghetti4527]

The thing is. there most likely is just one material on these things, so the whole texture will need to be redone!

[@angulinhiduje6093]

i agree that at the current stage its unusable out of the box. one usefull aspect i found is that it can ttake concept images and create models with perfect proportions in 3D. yes, the quality itself is trash, buy you can easily use it as a block-out model to get the general shapes right. to me this was always a part i struggled with with more complex, non-humanoid characters. always the head to big or small, limbs too long or short now i have a 3D reference in seconds.

[@garymathis1042]

The thing that AI art that bothers me is that there is no enough artist control over the results.

[@ross-carlson]

Uh, but these are 3D MODELS - you can then manipulate them however you want - it's a TOOL that makes a 3D artists work EASIER.

[@Roxor128]

Like in the 2D case, lots of editing is needed. Get the program to do the broad strokes and produce something close to what you want, then load up the results in Blender and fine-tune it the rest of the way.

[@jasonhemphill8525]

I love that Nvidia is investing so much in research but at the same time, I kind of hate that it's Nvidia. I just feel like they are working towards building a walled like they did with cuda, and phyx, rtx, dlss.

[@SomeGuyNamedRoy]

As a construction worker that only wishes he could work in game design, this will save me precious time in 3d modeling! I dont have the luxury of working from an office. I get a couple hours a night to work on my game.

[@mandrakejake]

I could use software that enhances STL meshes from 3D scans.

[@ge2719]

i'm just going to go download all the 3d perspective photos from the warhammer website... ...no reason why... definitely nothing to do with this video or anything.... 😁

[@Theforeveraloneguy]

I studied computer animation 10 years ago, modelling these objects took forever but it was a job. Now that job won't even exist. AI is taking jobs left and right it seems.

[@AdityaTiwary-b8e]

Is this narrated by AI? Cause it feels like that

[@nikhilgokhale5459]

Could you talk with fewer pauses please?

[@bruddernature6702]

This is how you sell dreams to investors.

[@TransDrummer1312]

this is a load of barnacles

[@MDougiamas]

We need your video on Genesis!

[@alexmattheis]

I would use it for programming 3D stuff and for Blender Scenes.

[@RIOLOGY]

For all the 3D speciaslists that critisizes this, it is the equivalent of critisizing a newborn for not being able to run. At the rate we are progressing in AI currently, give it another year and it will surpass most human modellers. Either you learn to take advantage of new technology or you will get replaced.

[@mlassz009]

I started learning Blender 2 months ago.....Well, looks like time to find a new hobb6

[@Guythink-n4o]

What a future is emerging before my eyes 😮😮🎉

[@P-G-77]

Awesome... I imagined that sooner or later we would get to something like this, but seeing it and imagining what could be done is something that... leaves me speechless. The limit now... is only our inventiveness, imagination, etc... But I tend to think that soon we will see things that will be so incredible... that you won't believe it until you see it with your own eyes. This will happen...

[@dannybrandon2822]

There goes my job. Nice.

[@Dindonmasker]

What exactly are you doing that this replaces? I'm sure some people would look at this thinking they will be able to cut on workforce but i think that people who are good at what they do are valuable to get a quality final project.

[@itsd0nk]

Get good at using this to leverage your work output. Photoshop and Lightroom didn’t put photo editors out of work, it just raised the ceiling on what could be done by skilled workers, while making basic low effort stuff opened up to amateurs.

[@Luminari781s]

Just like, sculpt the AI model or something 😅

[@edwinyanez9838]

What are you talking about? Now you got superpowers and you are feeling sorry for yourself. What a lame attitude.

[@QuantumDark]

@@edwinyanez9838 when everyone got the same superpowers it's no longer super

[@theone6048]

@Two Minutes Papers, please actívate automatic voice traductor provided bye google😢

[@gawni1612]

Dude I think I just squeezed my papers.

[@hopper2716]

This is nice and all, but useless if its not free... also... will it be able to understand moving parts and can it auto-rig... then we're talking...

[@lukesball1]

"but useless if its not free".. How so?

[@AmBush2048]

There goes another field to face the layoff avalanche

[@Hanker1889]

This would be great to increase the total cost of the projects. Before you had the cost of an artist, now you have the cost of the AI and the cost of the artist which has to correct this shitty output (which, btw, will be the same amount of time, because it cost the same to fix something that creating it from zero)

[@kipchickensout]

I see so much 3D texture and model AI stuff and I hope some of that will be nicely usable in let's say Blender in the near future. Or having a lot of control about all the PBR textures of materials, because so far a lot of generators give you anything but the result is often meh or not very adjustable, and it would be cool to give reference images for stuff Seeing these videos gives me lots of hope

[@musabmohammed7973]

Wow that's amazing what paper can do. I am very paper to hear about this.