Fashion meets ArchViz (3D models) - what do you think?

[danielschramm]

Hi

 

I am one of the co-founders of IONOMO.

 

We will be offering ready posed 3D human still models for close up views beginning of next year. Our approach is somewhat different: Our 3d models (captured using a full body photogrammetry system) are dressed with virtual outfits created by real fashion designers.

 

Through this we can not only create very versatile models for different purposes but also hope to provide real value to ArchViz renderings.

 

You can check out images here: http://www.ionomo.com

 

Would be really great to get your feedback & thoughts on this.

 

Thx in advance!

 

Daniel

[RyderSK]

The idea is great, scanned model plus simulated clothes, custom added hair meshes,etc...that's correct way to high-end photorealistic hummies.

The execution...is not.

 

3D people lately swarmed market, although in quality that's just notch-better than the ubiqutois 3D from Axyz and Evermotion. The realness of them including these, is so far from reality that it's not about uncanney valley, it's just poor CGI altogether. They probably will have their place far away from camera but, for close or medium range, they literally scare me with their uglyness.

 

I am not really sure clients will want their renders populated with scary plastic dolls without emotions. But I already see it happening on internet...I can only guess what horrors will the next year bring to us.

 

Though I am glad for the tech progressing down, it's not easy nor cheap to produce high-end hummies, even the top scanned content from Infinite-Realities with top shaders, still have the uncanny valley and there is lot of know-how/effort/money going into producing those.

 

I don't think the tech is ready for mass-market, and if, the price for the effort would be almost unnatainable for most.

 

I think it might have much market outside of archviz though in current stage.

 

Again, no offence, just being honest what I think of these right now.

[John Dollus]

I would disagree with Juraj to a degree. i can replace heads with photography super fast but painting the proper lighting/shading onto the clothing can be a huge pain. This would take a lot of that pain away. The downside is you are obviously locking yourself into a particular style of fashion and our clients can be extremely particular about such things. Error on the side of what the rich and famous wear in the fashion magazines and you will probably be okay. Just look at the products currently offered by the companies Juraj referenced and price accordingly.

[beestee]

One thing that I have found distracting with this type of solution is the poses that are available. These seem much better than what I have used in the past.

 

 

Although, for the bigger clients that I have worked for they request a much more obscure approach to the appearance of human scale figures. They want to avoid offending the audience that would be viewing the renderings by making it impossible to determine ethnic or cultural distinctions. We usually end up using models such as these as simple silhouettes. Would be interesting to see a boxed solution for this particular need with a cost to match, as such models would not need UVs or photographic textures.

[Devin Johnston]

Daniel I like the direction you've taken with the clothing, one of the biggest complaints I have about AXYZ's models is they are a fashion nightmares. Being able to change the outfit of a model makes that model much more valuable as you can use them over and over again in the same scene. I recommend you come up with a simple way to mix and match the clothing on each model, I don't want to spend a lot of time dressing each one of them. As far as the quality of the human goes for me it always comes down to their face, that's why when I use 3d people in my scene there almost always facing away from the camera except when they are far away. Your models look great, they are just as good as AXYZ's models in the face and even better from the neck down. I think if you can find a way to make them look more human in the face you'll have a hit on your hands.

[RyderSK]

I am surprised to see the standard of what looks good to be so low. If people are used in visualization, they better make sense and look good. If "The Sims3" is what you like, be it. What I see now, is hardly suitable for close-ups (maybe bird perspective)

 

Here is how the current tech looks like though:

 

This one is from local guy from my city:

 

More info :http://www.cgfeedback.com/cgfeedback/showthread.php?t=3776&page=49

 

Infinite-Realities:

 

Slide UK (this is realtime!)

 

 

Scanning is almost the most simple of the whole process, good shaders are altogether different.

Regarding scanned vs simulated cloth, that's again a pitfall. You just won't get the microdetail without bringing some additional info from real-scanned source.

 

This is simulated 3D cloth:

 

[Devin Johnston]

Juraj there's a big difference between a digital model you'd use in a movie and one you'd put into Arch Viz imagery. I agree with you that the ideal solution would be to have 100% photo real humans but the technology and hardware just isn't there yet in our industry. Most of the arch viz 3d people need to be relatively low in poly count, the shaders usually don't use SSS at all and there is no cloth sim. This is done to minimize rendering time and to keep the resources needed by the model to a minimum. I know we'll eventually get there but it will take time to develop them, companies like Daniel's are doing their best to push the envelope and I think we're all better off for it.

[RyderSK]

What I posted aren't really VFX double quality yet, so these aren't top level, actually, far from it.

 

I don't think it's argument either, because atleast in moving imagery, you won't notice all those "imperfections" (well..) but in static archviz imagery,

poor 3D people look like salt into eyes.

 

Technology in archviz isn't really underwhelming for use of such quality, polycount and shaders can be optimized to heavy extent and still look incredibly real, the real-time above posted from Slide takes less memory and performance than single Axys "model".

 

If they can't be done right, then why switch to them ? Purely for easyness, despite the horrible results ?

 

Photoshoped people still reign supreme from advertising to archviz. I can't stop looking at the craftmanship of MIR, so tastefully and technically well done:

 

 

I simply believe "pushing the envelope" looks differently.

 

And honestly, I don't see a place for them. As I see it, either you push photoreal work, and these aren't good enough at all, or you don't and these neither fit to illustrative work. All I can imagine is, they fit into the mediocre, neither real, neither illustrative, nor artsy, nor technically well done.

Edited December 16, 2013 by RyderSK

[Scott Schroeder]

No pancake hands and don't put them in stupid ass poses. No waving to the camera, no pointing at some off the camera object, nothing. This guy from axyz looks like he is very annoyed you are putting him in your scene; https://secure.axyz-design.com/shop/cman0007hd2o02p07s

 

I won't even comment on this, but sad to say this is in a paid collection....

 

My biggest and loudest gripe with all 3d people out there, and even some 2d people, is that the companies who make then almost want them to stand out. As an artist, I want the person to blend into the scene. I don't want someone looking at my rendering and go, "Oh you used company x's 3d models." I also generally hate the corpse/dead pan expressions most 3d models have in the face.

 

I'd like to see models that can have their textures easily swapped out. Most of our work is with sports teams, so any human in the scene needs to be in some variation of that team's colors and even have branded shirts to that team. Just normal people, we'd open up our wallets and say, "Take my money!"

 

You want to avoid having the perky nipples and open shirts (5th from the end) where you can clearly see her boobs too. Any female client seeing this may get offended, at least in the all too prude US. http://letswatchtv.com/wp-content/uploads/2013/12/Steph_outfits2.jpg

[Scott Schroeder]

What I posted aren't really VFX double quality yet, so these aren't top level, actually, far from it.

 

I don't think it's argument either, because atleast in moving imagery, you won't notice all those "imperfections" (well..) but in static archviz imagery,

poor 3D people look like salt into eyes.

 

Technology in archviz isn't really underwhelming for use of such quality, polycount and shaders can be optimized to heavy extent and still look incredibly real, the real-time above posted from Slide takes less memory and performance than single Axys "model".

 

If they can't be done right, then why switch to them ? Purely for easyness, despite the horrible results ?

 

Photoshoped people still reign supreme from advertising to archviz. I can't stop looking at the craftmanship of MIR, so tastefully and technically well done:

 

 

I simply believe "pushing the envelope" looks differently.

 

And honestly, I don't see a place for them. As I see it, either you push photoreal work, and these aren't good enough at all, or you don't and these neither fit to illustrative work. All I can imagine is, they fit into the mediocre, neither real, neither illustrative, nor artsy, nor technically well done.

 

I have to agree with Juraj's points in both his posts. Video games and film are miles ahead of arch viz in character quality. Heck, all video games do most of their stuff in real time and are targeted to an average hardware audience. Why are we so stuck in the stone age with all of our processing power?

 

I'd rather see an empty room done to perfection that great rendering ruined by awful 3d people. Why do so many places feel the need to stick clearly sunlit people indoors? Or have the person with the sun coming from the left, put into a scene where all of the light and shadows are exactly opposite, and the even match the person's shadow to the scene? What are we, on Tatooine with 2 suns?

[Devin Johnston]

You can waste time with your friends when your chores are done

 

I think adding people is usually a last thought for most artists and it takes a lot of work to make them look good. I hate to say it but that's why I use 3d people, I don't have the time or motivation anymore to spend hours tweaking 2d people.

[heni30]

It's amazing how sensitive our eye is to body and facial gestures. You realize this when you try to fix something in Photoshop and even the most minute adjustments make a huge difference.

 

It's going to be hard to get that right in affordable 3D.

Edited December 16, 2013 by heni30

[Tommy L]

Someone mentioned polycount....

Polycount is practically irrelevent now, vertex level animated proxy meshes can be in the billions with VRay. Animated trees is a good example.

[Nic H]

they look to be OK for background people but they still look terrifying.....and will need to be pointed away from the camera.

the clothing looks more normal than other collections which is a good start.

 

my main gripe is that they aren't rigged or usable in animation. thats where the gap in the market is. there is nothing out there at the moment. rocketbox and axyz are hideous.

[Crazy Homeless Guy]

Your collection needs to be higher fidelity and they need to be rigged as Nic points out. I really like the idea of the better quality clothing but the mesh resolution appears to be low, and the quality of your shaders can be improved.

 

These people suffer the same problem... they are posed and not rigged.

http://humanalloy.com/3d-people/

Edited December 16, 2013 by Crazy Homeless Guy

[Tommy L]

I think to use words like 'hideous' and 'terrifying' is overstating the case a little. I also know what it takes to scan/digitize people and its quite a Herculean task, not a point 'n shoot exercise at all. If it was simple there would be good solutions all over the place....

 

I think its probably one of the best of the available products. Not perfect, but with the relentless march of technology Im sure Ionomo will be well placed to bring better solutions to market and are one to watch. Subscribed.

 

Edit: It should be noted that the human brain has a whole seperate sector for facial recognition. Its way easier to fake anything other than the human form. Thats why you cant find your other sock but you can pick out a face in a crowd. You know you're getting close to a solution when people say its creepy rather than tell you specifically whats wrong with it.

Edited December 16, 2013 by Tommy L

[RyderSK]

You know you're getting close to a solution when people say its creepy rather than tell you specifically whats wrong with it.

 

I don't think that's the case here though. Some of the stuff I posted above might be creepy because of uncanny valley but here I could start naming and it would be very long list. It's not overstating, to me they're grotesque as they are, although...yes, not horrible like the Axyz.

I am perfectly fine if it's just the right quality for lot of people, but let's not pretend they are good in any real objective measure for any sort of CGI. They're not.

 

I can't even comment on the human alloy, those are just weirdly deformed to start with. If I would continue, I would mention non-existant shaders, poor silhuette, baked-in lighting and reflection making them not fit into any environment properly,etc.

 

What I take of all these scanned models sudenly appearing is that it looks like new lucrative market but most of those who offer this don't truly have the capacity to create them properly. They all feel rushed and amateurish right now.

Edited December 17, 2013 by RyderSK

[harryhirsch]

I am with Juraj on this one. Maybe his argumentations sounds a bit harsh to some people but basicly he is right. Most of these 3D people look like dolls and they are stiff. Yes, I have seen the ""IONOMO" gallery and the 3D people are getting better, no question. But its still visible that they are not humans. Simply ask your clients if they like 3D people or 2D psd. We do our render work "inhouse" and we hire arch viz companies if the deadlines too tight. When I present the project to our clients I want to discuss the project. I do not want that the clients get distracted or think about the people in the rendering. Yes, 3D peopel are ok for birdviews, but any close up and I ask the arch viz company to replace them with 2D.

[Tommy L]

I think these guys are not far from the mark. I also think that when you take a closer look at the execution of the people in Human Alloy they are not that bad. Compositionally poor, like they are not generic/vanilla enough, but if you take the face and hair away they are practically photoreal. Look at one of the turntables from the back. The layman would not be able to spot that its not a photo, especially just at a glance.

Once these things get rigged and animated convincingly I think its close. I know Im in the minority with that opinion, but I really think its close and the general opinion here is wrong. Especially when the true comparison is not between photography and a 3d model, its between the cost/logistics of a photo/video shoot versus the convenience/cost of one time purchase 3d assets.

 

Edit: Im also not someone thats ever used 3d people. I shoot greenscreen or photoshop stock. My experience has been CREATING 3d people using the Artec MHT scanner. 3d reverse engineering of the UVWs is a bitch.

Edited December 17, 2013 by Tommy L

[Stephen Thomas]

Some of the recent work I've seen using the newer axyz models has been pretty convincing.

 

 

And getting the correct lighting on a 3d model vs having to spend time getting a 2d asset to sit well in a scene has to be considered too. I think there is a place for both. Having said that I think the axyz examples I posted above have the edge over the OPs collection, facially at least.

[RyderSK]

Yes, those are decently good, atleast in geometry.

 

What is better compared to original models by OP: Scanned clothes contain low and high frequency detail that's impossible to simulate properly, it looks natural. They have natural poses and hints of good facial expression.

Negatives: Lazy simpleton shaders, no advanced hair meshes. Artistically no point in using them there ?

 

I already dislike how they are used...I pressume next year will be full of out-of-place people placed for the sake of novelty.

 

It's gonna look like this :- D Again, take it with ease ;- )

 

[himanshuchoudhary]

Great discussion guys!

I must say, the ideal situation would be when we could make custom people cutouts on a green screen depending on the camera angles and positions. But then again, not everyone has the time to do this. It all depends o nhow tight your deadlines are.

[Scott Schroeder]

When I think of greenscreen people, I tend to think of how Pixelflakes (http://pixelflakes.com/) does their imagery. You can tell they always shoot from the same height across all of their projects so their 2d photoshop assets come in with little to no fiddling to make them fit. To really make 2d assets work across your projects, you need to maintain some consistency in your camera views.

 

3d people are almost there. Yet to get that high quality that some of us would like, the price point wouldn't be competitive. The axyz examples posted in the previous page are pretty good, and the people fit that sort of scene. Though the guy standing facing away, his sagging pants just didn't scan well. It kind of looks like he pooped his pants and he's looking out the window thinking, "What do I do now?"

[numerobis]

Some of the recent work I've seen using the newer axyz models has been pretty convincing.

 

And getting the correct lighting on a 3d model vs having to spend time getting a 2d asset to sit well in a scene has to be considered too. I think there is a place for both. Having said that I think the axyz examples I posted above have the edge over the OPs collection, facially at least.

 

I thought the same - the latest scanned axyz models are a clear improvement. Most of the humanalloy models look also quite good to me - except their staring eyes.

[jordanp]

If you're looking for passable, rigged 3D models of people, why not try a place like Turbosquid?

 

http://www.turbosquid.com/3d-models/3d-model-axyz-rigged-characters/412696

 

http://www.turbosquid.com/3d-models/3d-axyz-2-rigged-characters-model/335760

 

http://www.turbosquid.com/FullPreview/Index.cfm/ID/526189

 

They're pretty close, and game-ready assets (like much of what Turbosquid and other model warehouses offer) will use polygons and textures more efficiently than designed-for-CG models, allowing for quicker render times - not noticeable in images with a single individual, but probably beneficial for large crowds.

 

Also, hi, first post!