The extinction of modelers is just a matter or time


#1

https://nv-tlabs.github.io/DIB-R/

One can say it won’t affect most jobs, but I assure it will change the landscape of what we know today as modeling. Just as watercolor painting and portrait painters is a miniscule niche now.
It’s just a matter of time until it matures, being able to model based even on concepts.
Maybe it’s a good thing, as modeling tasks are mostly mundane and not that creative. So games will fill with more objects, created faster. I’m sure it will be easy to create characters based on real humans. Maybe in a click or two.


#2

Eh. Technology will change things, of course, but more often then not with machine learning stuff, going from “ok” to “good” is many, many orders of magnitude harder than getting to “ok”. And AI tends to be very clever in some ways while being astonishingly stupid in others.

We’re probably not that far from the point where you can get a usable, production ready model from a few dozens photos from different angles. But usable models from concept art? I wouldn’t bet on it happening any time soon…


#3

Perhaps the models won’t be usable instantly, but it will cut the production time. So more people will be able to make it. Not that it’s bad. I actually don’t think that modeling itself is something, which should be kept as a thing in itself. I’ve definitevly seen AI being able to create models from concepts even today, so it’s not that far-fetched.


#4

Photogrammetry is already in use quite a bit, but it still can’t make things that don’t exist, modelers will still have quite a bit of work to do for decades to come.

(not to mention, most of the time the 3D scans require a lot of cleanup to be usable anyway)


#5

It’s very easy. Just draw the form guides, like crossections you often see in concept design. I’m sure AI won’t have difficulties interpreting it, and it takes a couple of minutes to draw by a concept designer.


#6

That would still require an artist to be part of the process to create anything at all. And, it’s not going to be precise and it would require additional work from a 3D artist anyway.


#7

It will be just like any other enabling tech- vaccum cleaners were supposed to free up time from cleaning the house- computers were supposed to reduce the need for paper and print…it ends up being a time suck- because the standards simply get higher. Lasseter had to tediously animate vertex to vertex when he started at Pixar, It gets better and deeper, and artists adapt, and we become more nuanced in discerning the boundaries of the uncanny valley.


#8

Machine learning is all the buzz, but this is 10+ years out before making any hero production models. People overestimate machine learning. In the meantime, it will create tools to make more mediocre tasks easier and quicker.


#9

for extinguishing faster :

https://www.youtube.com/watch?v=EtM3zhIqziQ


#10

mister3d said:
Just as watercolor painting and portrait painters is a miniscule niche now.

Wow! that’s the most obtuse, uninformed statement I’ve read for quite some time, care too qualify your sentiment?

…and as for an impending demise of a fundamental 3D skillset…probably not on this side of St Peter’s pearly gates.

Mate, please take my advice and grow a brain, really try to refrain from posting click baity tripe…


#11

Spot the angry water colour portrait painter.

But he’s right. In terms of paying someone to do a job, how many professional water colour painters do you see getting hired to paint background plates for animation? Photography has almost entirely eliminated demand for getting a portrait painted. They exist and people do get them done, but in an age of mass photography, portrait painters are doing it for their own entertainment, not as a job.


#12

Sacboi, I don’t know about your country (care to elaborate?), but here painters are obsolete as a class. The majority of “painters” bacame designers, because that’s the demand of the industry. Very few people are buying paintings, people are buying TV’s and Iphones, where they can witness hundreds of pictures a day. Why to buy just one?
People are the creators today. They shoot with their mobile phones, post it, they see a lot of good art and design everyday. It’s just incomparable to what people were soaking in a couple of decades ago and nowadays.
The illustrative and narrative function of paintings and a lot of traditional art has gone. I’m not arguing about its educational value for artists. Definitely it helps, albeit a bit limiting in its academical suffocating and isolated void.


#13

Well, prior too transitioning across to CG I’d made a modest living as a Portraitist (pastel) on commission, throughout the 90s here in Australia, where traditional fine arts still thrive. The digital medium is built upon foundational artistic tenets, also many renown concept artists and matt painters usually are traditionally trained. In addition tech generated content by it’s very nature is ephemeral, whereas a tactile Painting for example rendered on a canvas support can typically last centuries plus contemporary artworks are still purchased today not just for a financial investment but an intrinsic value of perception by the viewer which the same can’t be said for a hundred random CG images on social media.


#14

Browse the top tier portfolio selection on ArtStation, I think you’ll find those attributes have indeed crossed over too digital.


#15

I think that for modelers, as for many other field of 3d we are going towards a more clearer distinction between CREATIVE and TECHNICIAN artists.

The creative ones create their own design from scratch, maybe even paper and pencil, sametimes they make a clay model to scan, and then they uses softwares as Zbrush to develop and explore the design. For them automation and more powerful software are just another tool to turn faster their ideas in models. There are plenty of example here on CGspciety, modellers who turn their own ideas in models and images.

The technicians instead create their models starting from a blueprint or an existing model following the instructions of same one else.

The tech are the ones that are more affected by automation and by creative designers learning 3d becoming autonomous. But usually designers are more interested in developing their idea then in making a really polished model, so they need a tech to clean it, and usually when a tech scans or copy an existing model, then he has to modify it on request of the client.

I think the pure tech to remain valuable should learn a bit of designer stuff: design, materials, production process,… to became more then just executive, to give more inputs and reviews to the process and pipeline.

It’s the same as growing number of architects learning rendering, but many still need pure renderists because they want to spend more time on the idea or the project. But a renderist who knows interior design, composition, and house building is more independent and valuable.

Or animators, many 2d classical prefer focusing on the main performance, poses and movements, and leave all the technical curves cleaning, constraints, contacts, etc… to a tech. Even mocaps are heavily modified to made them more interesting. And there are still we have creative 3d animators who create performances from scratch on their own.

So my idea is, as long as you are creative and learn new stuff, as an artist or tech you will never extinguish, maybe you will lose just the most tedious and repetitive tasks.


#16

If to talk about depiction of the surrounding world, the photography does it better. Photos of people depict it more accurately, as not a single master could portray. HDRI displays and photos reproduct sunsets, dusks and waterwaves much more vividly.
Once something has been digitized, it’s no longer a “traditional” medium, so Artstation is kind of useless here.
Traditional medium is etching, oil painting, watercolor.
It posits a question, though, where the line lies between digital and traditional, which is vague. As digital can be printed, even having oil painting finish.


#17

You may have seen the video comparison between the FX work on the Irishman versus the DeepFakes version.

It’s pretty amazing how much better the DeepFakes DeNiro is.

Would they have done a digital portrait of DeNiro etc before they did the de-aging? They should re-do the Trom Legacy Jeff Bridges.


#18

That’s astonishing!
I recall they utilized deepfake technique as well. I guess it would be very difficult achieving glitch-free effect with deepfake approach, provided they needed at least 2k resolution.


#19

How exactly do they look ‘much better’ compared to the original?

The only difference is that they trained the deepfakes to younger faces of the actors. The younger faces are probably what you expect to see when you think of ‘DeNiro, Pacino and Pesci’.

And as they are tracking it to the re-worked shots from the show, the ‘uncanny valley’ problem remains exactly as it was before.

In the deepfakes, Al Pacino’s eyes are way too big. He looks like he’s wacked out on superglue or something…


#20

Haha superglued Pacino.
In the original, DeNiro’s eyes are too shrunken and old-looking. That is where they really improved it–I remember years back there was a discussion about the de-aging in X-men 3 and someone said the area they didn’t change was the eyes, and it seems this is where the most impact is. It’s impressive (in the case of most of the DeNiro shots).
Then again-they used the Netflix template, if they had tried doing it with the 70-year old DeNiro footage, would it have worked so well?