Future of texturing and 3d modeling in gaming industry
polycounter lvl 3
Hi there
I have one question that is bothering me- how 3d scanning will affect 3d modeling and texturing workflow in the near future?
Take a look at this
http://www.theastronauts.com/2014/03/visual-revolution-vanishing-ethan-carter/
By 3d scanning you can achieve very good looknig asset prety quickly.(but still you steel need to do retopology, uvmapping, etc.) The problem is- what about available jobs in the industry. It already hard to find a good job withoud having any experience and this technology will definitely reduce amount of needed artists to create assets.
So there are two possible options (at least i heard about two
):
- many companies starts to use 3d scanning in their workflow (except companies that make sci fi games etc) some of the artists gets fired (or companies dont hire anymore for some time), it is even harder to find a job for 3d modelers and texture artists
-new technology will reduce the cost of making games so new companies appear (new workplaces)- it is easier to get a job
What do you think about it?
Im going to start my adventure in gaming industry in few years so this problem is important for me, btw sorry for my english xd
I have one question that is bothering me- how 3d scanning will affect 3d modeling and texturing workflow in the near future?
Take a look at this
http://www.theastronauts.com/2014/03/visual-revolution-vanishing-ethan-carter/
By 3d scanning you can achieve very good looknig asset prety quickly.(but still you steel need to do retopology, uvmapping, etc.) The problem is- what about available jobs in the industry. It already hard to find a good job withoud having any experience and this technology will definitely reduce amount of needed artists to create assets.
So there are two possible options (at least i heard about two
):- many companies starts to use 3d scanning in their workflow (except companies that make sci fi games etc) some of the artists gets fired (or companies dont hire anymore for some time), it is even harder to find a job for 3d modelers and texture artists
-new technology will reduce the cost of making games so new companies appear (new workplaces)- it is easier to get a job
What do you think about it?
Im going to start my adventure in gaming industry in few years so this problem is important for me, btw sorry for my english xd
Replies
quote from the article..
"And the end of the day, photogrammetry is a tool. Nothing less, but nothing more. It’s still up to designers and artists to decide what kind of world they are creating, and on what journey they want to invite the players."
It'll be kinda cool if an artist could load up one of thousands of 3d scanned objects, use color pickers and different view modes to get PBR data from the reference scan, and use that information on model/project they are working on. Although it'd probably be a lot more practical to compile scan data into tiling textures like DDO's mega scans.
There would still be jobs available for Zbrush artists to retopologize and do cleanup, even if the entire game industry exclusively used scan data for asset creation. You probably wouldn't be having any fun or doing anything artistically challenging though.
Anything to make our job easier I'm for.
You cannot take for granted that in 5 years you'll be working the same way in this industry. You have to be adaptable. Learn art principles that won't change but practice them with whatever tools are current.
Creating quality 3D artwork takes enough time as it is, the more tools and possible techniques that might work for a particular studio the merrier I say.
Not to mention that fear misses the idea that tools that free up some of an artist's time let's him or her focus on creating even more work = possibly more unique assets, characters, and environments that otherwise would of been scrapped in the concept phase for being too time-consuming. Imagine if some magical tool sped up the environment process for Skyrim by 50%, knowing Bethesda's previous games they probably would of tried to pack even more content and new dungeon tilesets in rather than say hey-o and not try to push themselves further.
It's not like there's a ton of spare time and a lack of what to squeeze in next in game production.. Usually you hear quite the contrary.
And such technologies actually probably create more varied kinds of jobs, said scanning equipment probably would have a company hire more technical artists as an example. Not only that but you'd need to hire location scouters just like in the film industry to actually find places interesting enough that would fit well inside your game, and probably fly a crew around with all that scanning equipment if say you wanted to scan the Roman Colosseum and found yourself stuck in New York City.
Paid traveling to distant countries to 3D scan different types of architecture and geometry to be used in a game? Sign me up!
and if you're really fearful, you can't scan a Dragon!
You also have to factor in what art in real-time games actually means. Sure, it's purpose is to provide visual feedback and give a sense of being in an environment, setting, or situation that you normally wouldn't be in or see in your day-to-day life (visual immersion)
But it also has the rather huge part of having to be functional and serve under the structure of gameplay and level design, something I'd say most real-life buildings are not really suited for by themselves without further modifications.
You wouldn't really want to hand over to your level designers a set of buildings that are basically completely static and function as-is. Hence the whole notion of modeling with modularity and tile-sets in mind; not only would you have to retopo said scanned buildings so they are optimized but you would probably have to hack away at them smartly as well.
So yeah, I view 3D scanning as a possible and welcome addition to the current selection of tools for 3D artists to use, but I don't see it taking away any jobs. (at least until they can somehow scan the visuals going on in people's brains :O)
when they can scan things out of a concepts artist imagination
Until then, how would you scan a demon, a dinosaur or just a lamp that does not exist but has to be conform to a environment ?
Larger studios probably use more of it, but for the normal modeler barely anything should change I suppose, no need to worry
Sacrifice a virgin lass on a demonic altar to summon it, then hope it has a penchant for game art and is willing to co-operate?
You got the point.
They said the same thing about artists being replaced by Mocap or 3d scanning when it came to films/adverts but it never happened.
Poeple adjusted their pipeline, tools and embraced the new technology.
I remember a decade ago seeing forum posts talking about 3D scanning taking all our jorbs.
It's risky because everything has to be captured in the right matching lighting and placed in such a way that you don't get conflicting shadow directions. Also the lighting ingame can't be too elaborate either.
Really they should have used cross-polarization to separate it out into flat albedo. But it was still a pretty game, from a distance. I enjoyed it.
The hulk for the Avengers movie was based on 3D scan data of the actor, but with manipulated proportions and detailing. If your management thinks they can save modeling time and money by making artists start with scan data, even if you're making a demon, they would likely do it if it fits the art style.
Small studios can also just purchase data from 3D scanning companies relatively cheap. I suspect it will only be the larger companies making their own $20,000 80 DSLR photogrammetry rigs though. I believe EA and Ubisoft have already had their own for a while now.
Anyways, as others have stated, all the creations such as monsters, aliens, etc will still need the horse power of an ARTIST with a hollywood/film/photo level of quality. All this, if the production is aimed for realism.
All the games with cartoonish characters will remain with the same needs. The advantage of all this new tech will be noticed more in games such as COD.
Cartoon graphics and hand painted textures can't be scanned either.
You guys have to check out the Ryse GDC Talk.
Here's a sample of some of their facial animations:
http://vimeo.com/86232748
Insane what they do. They don't just scan the model, they scan the expressions (for morphed normal maps), then they scan the entire Facial Action Coding System for triggering of the normal maps, and blend shapes.
Crazy crazy shit! So damn cool!!
$20K will buy you a 30-40 camera rig. Don't forget it's not just cameras, but stands, cables, USB hubs, computers, flashes, and some custom software development and such to drive the rig. And even with that you're on a relatively standard set of cameras and objectives.
Also, as others have already pointed out, there are issues.
Scan data needs to be processed and cleaned up. Even if you have 120+ high end cameras, or even a Lightstage.
You can't scan sci-fi or fantasy stuff at all.
Even real life stuff needs scouting, casting, time and so on.
Also, sometimes real life just doesn't work and you and up making changes driven by art direction.
And in the end, scanned assets will free artists to spend more time on the stuff that can't be photographed.
Also, FACS scans. Aligning the scans is a nightmare. Actors usually don't get FACS and you can't get them up to speed in a day. Expression scans also don't just get into your rig without serious work, either.
All in all scanning is a pretty damn useful tool, but that's all it is, it won't replace you, it won't bring team sizes down, so don't fear it but embrace it instead.
I have used many types of scanners, from laser 3d scanners, to video scanners and phometric image scanning, and established a pipeline for my studio when these are in use.
Scanning will never replace an artist, only speed up a workflow and make them more efficient.
Scanning and getting a "good" scan is two completely different things.
Laser scanners like the "next-engine" scanner allow an object to be placed on a turntable and as it turns, scans a set distance using a combination of lasers and cameras to produce a mesh and a texture. The scan consists of one "cycle" but within that cycle it may scan the object on that particular rotation several times for clarity. The texture isn't an unwrap but merely the mesh referencing parts of the photograph. The issue I find with these sort of scanners is the alignment process and what they're actually able to pick up. Anything reflective will not be picked up; and unless there are obvious points to reference it can be difficult to align; unless you make one i.e. putting small bits of bluetac on certain sections that can be smoothed out in zbrush.
video scanners like the "artek" scan an object using its camera and flashes of flight. It works on a frame-rate and this effects the scan pickup very much. Its a handheld scanner and whilst its great; the issues I've found is that it loses tracking extremely often. By this I mean, it picks up an object and produces a 3d mesh on screen via its knowledge of where the object started in xyz space and where it ends. Should an object be consistent with no major landmarks it will struggle to comprehend where it begins and finishes.
Just to clarify on this further; if I was to 3d scan a rug from the centre to its edge; it would get lost, however if I scanned it from the edge to where it met the floor it would comprehend this fine.
Again issues with alignment can be found; it literally relies on the users ability to match several points amongst scans; it does however give a measurement of its accuracy in mm. In addition scans can give quite noisey and grainy results.
I clean scan data up pretty much on a daily/every few days basis in zbrush. Its extremely easy using a combination of morph, smooths, retops and projection. You'll get a good result IF you put the work in and only if. You still need to retop, you still need to bake and you still need to work hard to get the scan to a reasonable/aligned standard to begin with.
Hope this helps - any questions just ask.
Clearly based on opinion rather then experience as I've seen some amazing stuff using photometric scan data.
Just because you scan something doesn't mean it ends there; you can easily use this as a quick base. The exact same as you would with any base mesh.
Don't need any of that. I use this all the time. Works better;
http://www.artec3d.com/hardware/artec-eva/how_it_works/
Aligning scan data is stupid easy if you have a program made for doing it. What program were you using?
I'd suggest checking out Innovmetric Polyworks.
We use it at work and it can get great results, though it takes time. Though it costs 15k GBP....At least with that DSLR you atleast can use it for something else.
I worked at a 3D scanning company for a few years cleaning up scan data, so my opinion is based on quite a bit of experience.
The artec was sort of a gamble whether we got good data or not, and it took a really steady hand to get good results with it. Our camera rig produced much better stuff to work with.
I've been using some form of scan data to model cars for a while now.
I remember being a little worried about it taking my job when the tech first came out.
But in the end, as it's been said before, it's just a tool.
Even if you're taking the scan data and replicating it 1:1 without any artistic input, it still requires a lot of work and a good eye.
I see this tech being used to allow artist to perform their jobs quicker and to achieve more accurate results, not taking any jobs.
We've tested portable laser scanners and no thank you. The translucency of human skin wreaks havoc with it and you also can't keep a subject completely still for the time it requires to do all the passes. It's fine for maquettes with a matte finish but even a slight amount of reflectivity will also mess up the laser.
A DLSR rig is also far more flexible as well.
Er, I'm talking about aligning various expressions scans taken over a long period of time. I've yet to see proper alignment that doesn't need manual fixing.
I think our guys are using Meshlab, I'll point them to this one although I don't have high expectations. Too many moving parts of the human face, including the ears and the scalp, to get 100% perfection.
Clearly didn't look at the artek properly then because it's not a laser scanner, and seeing as we've made characters using that said scanner with little issues clarifies this
The only high quality scan data I've seen from real living humans was either from photoscans or from a light stage.
Don't worry. Its the same argument that people made for mocap - we still have animators don't we? Art direction is still needed and artists are still required to at the very least clean up the data and make it game-ready.
But more importantly, if you want epic concept art in motion vistas and beautiful things that don't exist in reality, you're still going to need artists aren't you? Unless you're recreating a real-world environment 1:1 you are going to want artists, and at the very least you need technically trained people to clean up the data.
I forgot to mention that the Artec always ruined the color textures and was highly sensitive to lighting conditions. Sometimes a weird lighting condition would make it impossible to capture usable data. The color texture data was usually in an unrealistic hue or overly saturated and totally mismatched to our photogrammetric data.
It was primarily used as our plan B scanner or a budget option for customers without a lot of money. Do you work for artec or something?
Scanning is a nice extra tool, but it's rather limited. You need to have a real-world asset first, which may or may not cost a lot of effort to make. For Ethan Carter it makes sense since it's pretty much a real location, but for something like Halo/Skyrim?
Yes, but it's quite a mixed thing
Our first experimental scans were created from static assets using a single camera. These included props like food on the table in the AC4 Black Flag trailer or sushi in Watch Dogs. Also, Edward's naked upper body was based on a lifecast, but we ended up heavily modifying it.
There's a yet unannounced project (going public very soon) where we've used a lot of facial scans from the client, and some arm/hand scans from our newly built rig. Well actually it's quite a few months old by now
We're now using it on a lot of stuff for human scanning, props and such, and still experimenting with its possible uses, but most of these projects will only be released next year.
Yeah we've looked into Artec but we've found it to be quite expensive (a proper full body scan needs I think something like 8 units), the donwloadable samples weren't too convincing, and it's flexibility didn't seem to be good enough either.
Scanning a living human is IMHO far too problematic with a handheld device. Just the breathing itself would generate a lot of inaccuracy, and even with holding to some bars or something would not guarantee a completely motionless stance. Photogrammetry rigs or Lightstages on the other hand fire in an instant so accuracy is only a function of the number of cameras vs. coverage.
http://www.3dscanstore.com/
The problem after this point, is mapping the animation and blend shapes.
Obviously this won't work without FACs, and the performance animation from the proper actors. It likely won't kill our in-house jobs, but it might reduce the requirement for outsourcing.
Every human face has a unique individual pattern of facial wrinkles, and you have two options to cover that - either model a custom face with edge loops aligned to those patterns, or create animated normal/displacement maps. Both require a lot of manual work.
Our approach to background characters is to re-use the same base mesh. This allows re-use of the same set of facial blendshapes; it takes some tweaking to fix things like eyelids not connecting in a blink shape and so on, but it's still reasonably fast and works well for most of the cases.
The other option would be to build a bones based face rig, but I find those to be even more complicated and harder to refit. Still it's a more viable option for games as blendshapes are very expensive, both in runtime memory and in processing (bones are almost free on the GPU).
Also, background characters are usually not tasked with expressing complex emotions, so you don't need overly complicated facial shapes. And if you do, it's better to do it right with custom models and shapes anyway.
Oh and more about FACS.
First, the basic set of 40-50 Action Units or elemental shapes is already a significant effort to scan, it takes hours for the capture and days for the processing. You'll also end up with the wrong expression for 10-30% of them because the actor can't be sufficiently trained for the session, so you already have to do a lot of manual sculpting.
Second, the system describes complex facial expressions as a combination of various AUs, and simply adding together the elemental blendshapes will break the face almost every time. So you need corrective shapes that are dialed in automatically, like when you need a smile with an opened jaw. The number of possible combinations is a few thousand and it's obviously impossible to scan them all, so you have to do a LOT of extra corrective shapes manually, too. However, combinations are usually less work - except a few dozen really nasty ones...
Of course if you don't need a fully complete character, you can narrow this down significantly, but for movie VFX it's a given that you'll need hundreds at least.
So, full body scans - OK. Basic expression scans - quite expensive. Fully featured FACS library - impossible.
But the Order and Ryse has me frothing at the mouth.
^__^
Very jealous of you pre-rendered guys over this.