“So much closer to 2 minutes than the lengrh of these videos...” I'd watch to as many minutes of these as I could-!
@gunsunnuva83463 жыл бұрын
Two Hour Papers, The Motion Picture
@oskarjung67383 жыл бұрын
Nice Dp
@mrpedrobraga3 жыл бұрын
@@oskarjung6738 Thanks
@NeoAnguiano3 жыл бұрын
This is procedural or parametric modeling, it's actually harder to model and you need a 3d artist that also understands programming (although very basic for the procedural i.e., Houdini) or very specific and technical knowledge like Fusion or SolidWorks to name a few. The main advantage are speed if you need multiple models, alternate shapes or iterations, or adjustable models that adapt to specific needs
@HailScreaM773 жыл бұрын
as someone that know a little of parametric desing , i think the same as you, this isnt too impresive, in catia or any cad software is very easy to relate the parameters of an object to another and then let it modyfy when you move the other, the only great work done here appears to be the point and click selection of what you want to mod
@orbatos3 жыл бұрын
@@HailScreaM77 TLDR; This is the equivalent to introducing inverse kinematics to the kinematic development pipeline. I might have agreed, but I read the paper because it was obviously made with this knowledge :-) . The demo is underwhelming trip use because parametrics themselves are not new, but this is a platform agnostic approach to addressing a key problem with generalising parametric modelling, the lack of back-propagated changes. From the abstract: ``` In this paper, we introduce an amendment process of the underlying direct acyclic graph (DAG) of a parametric shape. This amendment enables a local differentiation of the shape w.r.t. its hyper-parameters that we leverage to provide interactive direct manipulation of the output. ```
@Peter-ik9fz3 жыл бұрын
Yeah, somebody still created these 3D models and defined the parametric parts and they conveniently not showing that in the video. I don't see how this is "3D Modeling Without An Artist" and how it is different from other parametric modelling software.
@TheLumberjack19873 жыл бұрын
And that's not even mentioning that parametric modeling is not really a thing in animation or vfx. Sure there are procedural systems for cities etc., but all the individual pieces are all modeled and textured by using polygons and UVs, not nurbs or anything.
@animaToy3 жыл бұрын
I've got the same thought, it's misleading video, This method still requires all the data to be prepared by human (via modeling, scanning etc.) and also marked as editable, looks more like modeling + rigging routine to me.
@CybranM3 жыл бұрын
The cursor knowing which parameter to change is neat but parametric modelling has existed in 3d programs for a long time. Changing the parameters is the easy part, setting up the model with parameters is a lot harder and requires an artist
@FPRobber3 жыл бұрын
Thanks, I thought I was going crazy. The impressive part would be an AI building a parametric model. You don't need an AI to do inverse kinematics on a parametric model
@milanstevic84243 жыл бұрын
@@FPRobber well I actually thought they were using AI to build some kind of procedural graph, after the fact. in the case of IK I thought they were using AI to actually build the IK with the automatic constraints. if not, then I'm not sure what's the point.
@JrIcify3 жыл бұрын
So this is taking a normal non parametric model and making it parametric? How does it know that I want the toaster slots to be symmetrical and what if I don't?
@milanstevic84243 жыл бұрын
@@JrIcify no, it's not making anything. I figured out in the meantime, what this was about. it's a direct user interface between the geometry and the sliders what was done here. (the positional dragging without supplying that context beforehand, the brush size in one of the examples, but also the connectivity between two or more interconnected sliders as shown in the IK example)
@milanstevic84243 жыл бұрын
@@JrIcify I can imagine Karoly didn't invest himself too much in this area of life, or got too hyped himself, but the video is misleading all of us into believing this has to do something with object generation or automatic parametrization.
@zayag35433 жыл бұрын
Hate to say this, but nothing about this actually looks new for those of us who have to model for a living. You can achieve most of this with a good rig and shape keys + boolean modeling techniques. For other things you can use Houdini or blender geometry nodes.
@Recycled_Trash3 жыл бұрын
I think it does Everything automatically such as you import a design with no define geometry’s and it recognizes them automatically. I’m not sure tho this is my best guess
@karakter2223 жыл бұрын
While this is awesome, someone is still needed to make the base model before someone else can play with it and tweak it for what they need.
@mrpedrobraga3 жыл бұрын
Well, the great thing about these tools is that they're easily connectable as modules in a workflow. If we have an unrelated paper that generates these kind of models from descriptions, for example, it would just be plug-and-play
@Paulo_Dirac3 жыл бұрын
Wait for the next paper ^^
@JorgetePanete3 жыл бұрын
@@Paulo_Dirac Input: A train or something idk Output: image of train, converted to 3d, used as input for a "model to parametric shapes" converter 👁️👄👁️
@Paulo_Dirac3 жыл бұрын
@@JorgetePanete LOOOL
@HDFDuarte3 жыл бұрын
@@mrpedrobraga GPT-3 Codex already has an integration for Blender... If we join both works we might as well be already there
@kotokrabs3 жыл бұрын
Two more papers and we will make Pixar movies in our sleep 😅
@TwoMinutePapers3 жыл бұрын
That sounds great, sign me up! Thank you very much for dropping by!
@mrpedrobraga3 жыл бұрын
You say this kidding but the technology for this is coming surprisingly quickly. Brain Signals Reading + VR + parametric models
@orbatos3 жыл бұрын
@@mrpedrobraga The main overhead is generation of complex collections of parametric shapes as an object, essentially modelling in the first place. I expect a hybrid technique using description via audio, image or text, then generating models with parameters "known" but accessible. Needed components for this do seem to be available now. (ed, autocorrect)
@marka10003 жыл бұрын
@@mrpedrobraga yes we have crude BCI and pretty good VR tech, what about playing pixar movies directly to the brain??? I bet we don't have that.
@orbatos3 жыл бұрын
@@marka1000 _Actual_ low invasion nerve interfaces are still horribly inefficient and lacking in detail, but site progress is being made. The medical science is also not quite at the level that machine learning can help much, largely because of proprietary silos.
@Solizeus3 жыл бұрын
I can imagine that soon there will be a lot of online series being made by independent groups of 3 or so people, using voice changers, motion tracking and other stuff that is basically free and require little knowledge to use, that would be great
@Ch503043 жыл бұрын
That would be amazing.
@calabacita90953 жыл бұрын
@@orangestapler8729 For sure, lowering the entry barrier will make it so easy that anyone can do it, so there is going to be a lot of shit being produced, but at the same time it makes the ceiling of possibilities higher, so you have to really make something amazing to be defined as great work.
@Ikxi2 жыл бұрын
huh? series about what?
@tiagotiagot3 жыл бұрын
I feel like there's some key information that must be missing. Isn't this sorta how the character creator works in many games? Also reminds me a lot of some of the car modeling techniques in the game Automation by Camshaft Software... What's the new thing here that makes this noteworthy?
@WeedyFlash3 жыл бұрын
Yeah this was explained extremely poorly. I spent the whole video trying to read between the lines of what actually was being developed here that involves AI. Two minute paper videos have declined so much, I basically just mute and watch the demonstrations now. He talks a lot but provides no substantive information... almost like it was written by an AI.
@wateriseternal75923 жыл бұрын
Agreed ... as I understand it ... the parametric effects are given relationship attributes to other sections of model as it's built ... ... rather than having to code it to keep ratio size and position with other objects. So if I had a 3d sphere acting as a head then I add a thin cylinder to its surface acting as an eye ... I'd expect a deformation of the sphere to flatten slightly to accept the flat cylinder as its eye ... I should then be able to move the eye around ... and push it into the sphere, deforming it like a donut, or pull the cylinder away from sphere to make a snout with flat cylinder acting as an eye instead of snout. It has to just be groupings with modifiers ... memory intensive If this is machine learning ... everything I've said and more would be possible
@bradmitchell71823 жыл бұрын
This is literally what I do every day. They even missed flexible geometries based on parameters. Gosh I need to be paid way more
@joe72723 жыл бұрын
You are on the forefront of automating a huge amount of the skilled computer design field!
@alihms3 жыл бұрын
@@joe7272 I hope not. The parametric modeling is a not much of a skill really. Anybody can do it with training. Coming up with the idea, making sure it works, produceable, marketable and cost effective are the skills needed. This is what industrial designers and engineers need to excel at. This AI which simply parametrizes an existing design does not mean anything.
@llamazunited33 жыл бұрын
This is like the ultra version of creating things in spore.
@gonzalezm2443 жыл бұрын
@Anatol Yes please
@gonderage3 жыл бұрын
this will be spore jraphics in 2015
@eccentricity233 жыл бұрын
"No trained artist required. What a time to be alive!" (Unless you are a trained artist)
@orbatos3 жыл бұрын
Not quite, trained artists already use existing assets in many works. Recreating them (especially when realistic) is unnecessary, a dramatic waste of time and resources, and in the current world would put them out of work for being to slow. Now think how this and other workflow can enhance that workflow even further, perhaps allowing the client to describe and proof a scene even, then the artist can refine it stylistically as appropriate. This potentially frees the exchange from poor communication and allows the artist even more freedom to explore. Though it will also mean sweatshop labor for most media, that is literally already the case.
@FPRobber3 жыл бұрын
Except you need a very trained artist to build those models. It's like saying the ability to move stuff around in the Unity editor means you don't need modelers and programmers anymore.
@askingwhy1233 жыл бұрын
LOL!
@thomasmineo3 жыл бұрын
Man is so massive that scientists are literally warping their papers around him so that they can appear in his videos.
@TeddyLeppard3 жыл бұрын
Except the parametric model which enables these adjustments was very complicated and was built by a skilled person. I have no doubt this type of high-skill work will become completely automated one day, but not soon.
@razipour3 жыл бұрын
I'm doing this for years! Making customizable 3d objects!
@andreasv94723 жыл бұрын
With grasshopper?
@bakedbeings3 жыл бұрын
@@andreasv9472 parametric modellers like Fusion 360
@razipour3 жыл бұрын
@@andreasv9472 I use rigged mesh and unreal engine for making customizable objects.
@dereknalley3 жыл бұрын
I usually am blown away by the novel aspects of the videos here, but this one seems to lack anything new. However, if the authors were able to create an algorithm that took a 3D scan and added the parameters automatically... WOW... Loving it. Secretly, I think this video was made just because of the two minute papers model. Don't fall to pandering Dr. Károly Zsolnai-Fehér! You're better than this.
@piotrarturklos3 жыл бұрын
This video completely omits the steps taken to create such relations between elements of the object. There is no way the presented paper can be truly understood or evaluated based on this video. It is purely a cool simulation at this point.
@calorus3 жыл бұрын
Excellent excellent excellent - and, frankly I'm elated for you to see how what you do feeds and enthuses your community. Top work!
@cafeinomano_3 жыл бұрын
So intuitive...
@Noakesy3 жыл бұрын
Looks like the answer is still no. All you showed was some editing, who actually created the models in the first place :)
@sebastianjost3 жыл бұрын
I think there are AIs, that can create 3D models out of regular pictures. Though I don't know how general they are. Maybe they only work on very specific objects yet. (Or don't exist at all)
@YOEL_443 жыл бұрын
@@sebastianjost I wouldn't vote on that, this models require to be made in specific ways or else they don't work, the parametric thingy
@bradmitchell71823 жыл бұрын
Lol I do this all the time in my 3D modeling programs. Is just parameterization
@Paulo_Dirac3 жыл бұрын
+ connexions and inter-relation awareness
@CLNDSTNStudio3 жыл бұрын
you totally need some merch with the words "whatta time to be aliive!" on it... I'd totally buy that!
@sebastianzimmerhackl36863 жыл бұрын
More merch ;)
@DanielSimu3 жыл бұрын
Looks like fun software, but making parameterized models for this takes an artist much more work than creating regular 3d models. More work for artists, not less!
@clochard40743 жыл бұрын
True! On the other hand, the same model can be adapted and reused in a number of different situations!
@choo_choo_3 жыл бұрын
@@clochard4074 So can any of the other models someone makes. It's not like once a file is used it can never be used again and needs to be deleted.
@TrailAtlas2 жыл бұрын
It's remarkable how simple the UI for this is! I do this stuff all the time using a mix of Configurations, Sketches and Assemblies in Solidworks, but it's much clunkier. More like typing numbers into a table.
@PrinceWesterburg3 жыл бұрын
In 3D animation, when you rig a character, you also create a 'schematic view' which is a web page or interactive map of the character so an animator (simple minded but nice people) can animate the rig. Its the same thing and has been around for decades. How does this differ from that? Further, in most 3D software, you can create parametric objects (Cinema 4D is great at it) that are very easy for end used to break, err, I mean, work with.
@ruthlessguide65183 жыл бұрын
for those of us wanting to make charming or unusual animated films (that disney etc won't be interested in financing), this amazing paper is at least a big step and maybe even a big leap forwards. thank you, karoly and the authors of the paper!
@choo_choo_3 жыл бұрын
How? Someone still needs to make the model for you to play around with. This isn't some magical "wow, a customizable 3d model appeared out of thin air!" type thing. Some dude made a model, and exposed some variables like height/width and all of that. That's all this is, and it's been done for decades. The only way you're going to make any big leaps to making animated films is to actually buckle down and do it, learn the process and software, and stop looking and waiting for shortcuts that'll let you one-click the entire creative process so you can continue to be a useless "idea guy".
@ruthlessguide65183 жыл бұрын
@@choo_choo_ not so long ago, you would have been quite correct to suggest that the only way to get a model was either to get one from a library or marketplace or make one. however, that is no longer the case - at least for non-parametric models. i definitely take the point, also made by many other commenters here, that the parameters have to be set up. whether they still have to be set up by an artist or can be done by an AI algorithm, i do not know. however, for models in general, photogrammetry has improved quite a bit - though it is still far from perfect and can be tricky to get right. and AI itself is now producing alogorithms which can create models ex nihilo (i think Karoly has done one or more videos illustrating this). the reason for wanting to make models quickly is not laziness or just being a useless idea guy (not a very wise insult to throw around - after all plato and aristotle were ideas guys, but i don't think they can be called useless since much of our western civilization, like it or not, is based on their ideas) - it is the problem of budget if one is trying to make a larger scale animated film. it can take a very long time to make a really good and realistic model and if you have hundreds or even thousands of them, that either takes a lot of money or a lot of time - potentially more than a lifetime. if you are doing something at hobby level then part of the point and pleasure is quite probably in the learning of blender (or whatever DCC you are using - though if it has to be parametric, that isn't blender). but if it is just a tool on the way to a higher goal, then one needs to find faster ways of making good and malleable models.
@TheZenytram3 жыл бұрын
how this keep coming, every paper is more revolutionary than the last one.
@tim40gabby253 жыл бұрын
If less revolutionary, it would soon be a retro channel? just joshing :)
@ReynaSingh3 жыл бұрын
These papers never disappoint
@Jandodev3 жыл бұрын
This is exactly what I need!!! Im going to try to get codex to interface with this!
@sirforcer3 жыл бұрын
To be fair, you can get a similar effect to this in Houdini if you setup your object/scene correctly (driving the length of a chair leg based on the position of the seat for example). It doesn't have the brush stoke part, but instead a list of parameters to mess with, so less intuitive and requiring expensive software and an object specifically made for it. I'm excited to see how this paper evolves and hope it gets implemented into some modeling programs soon.
@Khether00013 жыл бұрын
Seems a bit more like a rigging than necessarily a modeling development, and it's nice to see how it is evolving into having what looks like a much better interface, both modelers and animators will benefit a lot from this! I'm really excited to see what this will end up being developed into! Thank you once again for sharing!
@serta57273 жыл бұрын
Wow works like a charm
@ShadowsDML3 жыл бұрын
This is game changing for 3d Printing. We are going to have fully customizable models soon!
@ntwadumela_jadu97473 жыл бұрын
This amazing piece of design software in combination with a 3D printer and anyone can be a one-man product design firm. When can I pay a monthly subscription to use this software?
@camilotorres98353 жыл бұрын
Sincerely interested in funding a project like that
@bakedbeings3 жыл бұрын
None of this covers design or modelling, it's just final variation with nice controls. It also doesn't say anything about the format of the models or exporting capabilities. The good news is you can do this sort of thing with Fusion 360 as a one person design firm, but there's just a lot more work and learning than is portrayed here. This demo has much more in common with the workflow of an animator - manipulating controls set up (with considerable work) by a rigger, on a model created by an artist.
@ntwadumela_jadu97473 жыл бұрын
@@darrennew8211 I'm pretty sure they already have an AI that can take a rough drawing and make a 3D model I'm 51% sure he already did a video on this. I want to say it was gpt3 or the latest iteration of it
@ntwadumela_jadu97473 жыл бұрын
@@bakedbeings pretty sure there's an algorithm that can take a rough crappy sketch someone hand drew on a napkin and turn it into a 3D model.
@theblackbaron41193 жыл бұрын
If we can use these models for 3D printing slicers aswell, then it's a lot easier to print differently sized models for prototyping.
@dhasfhadngsdgsdgrwg3 жыл бұрын
I'm not sure how different this is from using constraints in another CAD software
@Klaster_13 жыл бұрын
Second. If the models are parametric already, the novel part is the interaction mode?
@Zenheizer3 жыл бұрын
Yeah, this, and much more, can be done with Houdini or even Blender GeoNodes and other equivalents
@cornevanzyl58803 жыл бұрын
Thank you for making me excited about the future!
@JSZ694203 жыл бұрын
This is like Houdini or grasshopper but slightly more advanced, with the brush size feature being the advancement
@Rodroelmelon3 жыл бұрын
This made me think about all the parametric assets made with Houdini fx, maybe I will try to do that toaster with it! Cool video and fun to watch as always! keep doing it please!
@ollllj3 жыл бұрын
this was already used well on the Spore game., where it was purely cosmetic. With more detail (and inverse kinematics) in some shadertoy demos.
@ollllj3 жыл бұрын
@@kicka55 to generate good sharp corners on dynamic meshes (with lower polygon count for better mapping/reflection), you must minimize polygon-normals for leastSquares, using QuadraticErrorFunction==qef()-minimizer (and care for continuity, which may even need 2 derivatives/integrals for better results). DualContouring is tricky to do in real Time, quickly ends up being only 5 fps and neeeds compute-GPU. A simpler approach seems to be to just add lots of greebles and roughness, and to just do what UnrealEngine does best; high fill-rates while not being the best for reflections, and mostly only reflecting screenSpace.
@Bjarkediedrage3 жыл бұрын
Isn't that what Houdini also does? And how much effort is involved in producing these modifyable models?
@zsoltoravecz18153 жыл бұрын
There is a blender build, witch can do similar things with kitchen furnitures. Easy to use for a "noob" but harder to add a patametric door window etc for it.
@playonce41863 жыл бұрын
a complete start to end modelling would be nice to see
As others have pointed out, this isn't really significant. But if an AI could convert a picture into this it would be. But that's a very different thing.
@YuvrajSingh-rd4tf3 жыл бұрын
I literally said what a time to be alive out loud right before he said it and my morning is super charged now!
@tando62663 жыл бұрын
Summing up the Comments: The VFX side: This is awesome, really going to break down walls for small teams. The Parametric modelers: Nothing really that special going on here, I guess the selection based on mouse size is a nice feature. You still needed someone to design the first part and assign relationships. The real value is something in the middle.
@rantingrodent4163 жыл бұрын
When two groups disagree on their perceptions, the truth does not necessarily (and often does not) lie in the middle.
@noutram10003 жыл бұрын
The hardest thing about 3D modelling is remembering what can be a very complex interface to do stuff. That's why I like TinkerCad because its so limited (although for sure the output models are not going to win any design awards). I can see this being useful for 'tinkering' easily with much more complex models or objects but the user interface needs to be kept simple, that would be the key to its usability IMO.
@MSkwar3 жыл бұрын
Anyone who uses blender or solidworks is barely impressed.
@DreckbobBratpfanne3 жыл бұрын
This could be used in some VR context, you don't need to program interactions with objects anymore and instead provide something like this 🤯
@jordanscarrott37493 жыл бұрын
We desperately need a Two Minute Papers episode on Deepminds AlphaFold2 results!
@excitedbox57053 жыл бұрын
this is already possible in CAD tools like Solidworks or Inventor. You can group objects and you can set motion points, You could also do this in blender but it would take more setup time.
@irkedoff3 жыл бұрын
NICE❣
@mootack3 жыл бұрын
The COOLER geometry nodes
@vantahku72113 жыл бұрын
I come here to chill out. The world freaks me out and it's nice to see something that isn't on fire all the time.
@lucas3d6763 жыл бұрын
imagine giving this tool to one of your clients... the one that is obsessed with revisions and getting everything "perfect" Oh Lord! :D
@Dolphinoff3 жыл бұрын
Super cool
@PierreH19683 жыл бұрын
I think those tools are for a trained artists!. More for artists more than engineers... it opens the world of physics to artists more than the world of art to engineers
@ringtail50213 жыл бұрын
I'm pretty sure it's just a matter of years before 3d modelers are just gonna be replaced by robots
@TheKorabkoff3 жыл бұрын
Very misleading name for video. This tool is interface to interact with parametric models right? To make parametric 3D model of toaster like this you have to be experienced 3D modeler and semi programmer. To make use of this tool in production you have to have huge parametric model library, which is not even started. This is tool for solving unexciting problem.
@raullara90153 жыл бұрын
As a 3D modeling student, I feel at the same time amazed and scared.
@johnpapiewski70223 жыл бұрын
So you still need an artist to input the parametric shapes?
@morphtek3 жыл бұрын
from reading the comments people dont realise that this is not 3d modeling it is actualy "intuitive parameters control" of the specifically made 3d model ,someone (an actual 3d artist with knowledge of parametric modeling) will need to make the base and program the parameters and input of the sliders ,also this just looks like a tool that sideFX will implement in houdini in a couple of months for sure
@amrzakaria52903 жыл бұрын
Very good for 3d editing.
@nipunasudha3 жыл бұрын
*Suggestion for Károly* Dear Károly, it's very hard to identify your videos in the recommendation list because each thumbnail is different. Could you please add an unified footer/frame to your thumbnails. I assure you this is going to significantly improve your reach. Thank you for all the hard work that goes into these ❤
@SepehrNaserkhaki3 жыл бұрын
This could make hobbyist CAD work so much more accessible!
@lingwisyer883 жыл бұрын
And here I am trying to script a million and one specific nodes to be able to flexibly modify 3D geometry directly...
@Tom-ef1mz3 жыл бұрын
The fact that there was no artist involved definitely showed XD
@Zoza153 жыл бұрын
I swear, someone will make a pumpkin car!.
@Build_the_Future3 жыл бұрын
This is not special. Blender modifiers can do all of this easily. What would be impressive is if it was able to apply the Modifiers on its own and not have to be set up, but that is not the case in this paper :(
@rahulkatre97133 жыл бұрын
Yeah I was going to say that CAD programs like Autodesk Inventor, Fusion 360, Solidworks, etc. are already parametric and you could do things like this as long as the base model has been set up correctly to be parametrized, which again isn't hard if you're decent with CAD. And a parametrized base model is already a requirement for this software too, so other than a different UI it isn't really adding much. Maybe the only advantage is that this runs in real time, but the CAD program can also process way more complex designs than this thing.
@HebaruSan3 жыл бұрын
Still need artists, they'll just finish quicker
@YOEL_443 жыл бұрын
Not even, it needs to be parametrical, some artist might not need to think about it in their regular workflow
@desocrate3 жыл бұрын
The person manipulating the object is still considered an artist. Tools like these are often created by Technical Artists for 3D Artist to create content quicker.
@Veptis2 жыл бұрын
Setting up your constraints well or even writing models in openSCAD will give you parametric models. The customizer on Thingiverse is a great example of what we already have. I am really missing the point here - is it just about the interface interaction using a cursor instead of sliders?
@JonnesTT3 жыл бұрын
Thats very neat! You'll still need 3D artists to make the objects for a game but now, background items can be edited by level designers or QA for a little bit of extra personality in each scene :)
@gmodrules1234567893 жыл бұрын
So, if I'm understanding this correctly, it's basically like CAD, but for any 3d model?
@milanstevic84243 жыл бұрын
okay it took me a couple of minutes to realize what I watched. the true breakthrough here is that the user interface knows how to inversely control the hyperparameters. it's not exactly trivial, but that's all. in other words, it amends the acyclic graph that is used to produce the geometry in such a way to help interpret direct access to geometry inversely, thus identifying the exact path in the graph, based on input, and streamlining the control logic, instead of requiring a user to tweak the sliders themselves. this project builds upon procedural modelling and parametrization without doing anything on that forefront. certainly no AI as far as I can tell and no changes to how procedural modelling or IK actually works. somebody still has to rig everything into a coherent whole. basically it's the Spore (2008) editor for the world of commercial parameterized objects, which are supposed to be made to behave like that by (many) artists. so: 1) the video title is misleading. 2) yes, artists are very much required to build this narrow combinatorial space of hyper-parametric freedom.
@electronresonator88823 жыл бұрын
I hope next two paper will show us converting real photo into virtual object simply by using phone camera
@willhart21883 жыл бұрын
I'd like to see this on the Sims 5, if that ever happens.
@PedanticPig3 жыл бұрын
Paralives (indie Sims-like in development) has something similar to this.
@JiteshGosar3 жыл бұрын
Which simulation tool is used to show this results? How can I create similar simulation?
@jacobellinger80273 жыл бұрын
og Startrek ai's are becoming real in my lifetime!!
@iamapie133 жыл бұрын
Meanwhile trained artists "are you threatening my job?"
@lethaldumpster26993 жыл бұрын
Artists have to make the base model, and then the artist or a coder has to setup advanced systems to connect that model to this software. This isnt modeling without an artist, as its an artists model, being tweaked with values. Which has existed for years. Houdini actually has the power to do this in Unreal Engine 4, right now. Its super cool, but the setup is super time consuming on the coder side. An artist could make the toaster taller in 60 seconds, while setting it up to use the advanced system so a non-artist can edit it, would take several hours.
@iamapie133 жыл бұрын
@@lethaldumpster2699 I know, it's a joke, he repeatedly says the phrase "without a trained artist"
@johnclark9263 жыл бұрын
The title made me think there were AI-generated 3D models. Oh well, this is much more practical and maybe there’ll be a 3D-model-from-text-prompt AI two more papers down the line! Edit: I appreciate the title change, it fits the paper much better
@Mummy5423 жыл бұрын
It is not new technology, its a usual parametric system. Its suitable only for making variations of already modelled and setupped stuff, witch easy to make by hand in most cases.
@eldrickzero48853 жыл бұрын
Non artist people don't really understand what artists do, it's kinda funny
@trafficface3 жыл бұрын
First they take my programming, now they take my 3D modelling... I have nothing left to be replaced by 😭
@niallrussell71843 жыл бұрын
interesting, but CSG pre-dates meshes, and parameters have always been there, like 3D Studio stacks. Was a pita modelling in raytracers like vivid/POV all those decades ago.
@henrychan7203 жыл бұрын
Haven't CAD programs been doing this for 20 years already?
@rexbk093 жыл бұрын
Love, Love, Love !@!
@kabirbroadcasting3 жыл бұрын
This is insane. 🙏
@Fredman55513 жыл бұрын
I'm hoping Corridor digital watches this
@abhijitleihaorambam37633 жыл бұрын
Now 3d modellers, programmers who next doctors ?🤔🤔
@sourdonkeyjuice3 жыл бұрын
Soo.. it's like inverse kinematics on a toaster 😂😂
@theatheistpaladin3 жыл бұрын
This should be a blender extention.
@evolutionxbox3 жыл бұрын
This could really help with web design/development
@neolynxer3 жыл бұрын
Congratulations, what a time to be alive! They've invented Houdini!
@MonsieurSwag3 жыл бұрын
In blender*
@Sirmenonottwo3 жыл бұрын
This will be a great new modeling software! If it is open source then someone could implement it into blender or some such other preexisting software.
3 жыл бұрын
That dislike is from a 3D studio max developer xD
@nuduw3 жыл бұрын
I mean isn't this the same thing we do in Houdini or Blender's geometry nodes? Parametric/Procedural workflow... Of course CAD softwares were a thing since a very long time, they work totally in a parametric manner.
@phailupe29413 жыл бұрын
Oh yea, it’s all moving together
@Leonardorth.3 жыл бұрын
Me who wants to become a 3D artist watching this video: "Bruh"