Hey guys!
Today is finally here! We've launched Megascans!
Plus Megascans Studio and Megascans Bridge for Windows and OSX, a standalone toolset for customizing, blending and exporting scans. From all of us at Quixel, a HUGE thank you for your support. As you guys know, we've got deep roots here at Polycount - this is a fantastic community that's helped push our toolset to the next level. That said, we present you with...
Today is finally here! We've launched Megascans!
Plus Megascans Studio and Megascans Bridge for Windows and OSX, a standalone toolset for customizing, blending and exporting scans. From all of us at Quixel, a HUGE thank you for your support. As you guys know, we've got deep roots here at Polycount - this is a fantastic community that's helped push our toolset to the next level. That said, we present you with...
https://www.youtube.com/watch?v=BnSekpFSLnU
Replies
Megascans general feedback
Website:
Megascans Bridge:
Megascans Studio:
- what is this thing?
- what are the plans for this thing?
- it would take some serious innovation to pry my fanboy fingers from substance, I am looking forward to said innovation
edit: little bit too bluntThat's what i have for now. Thanks Quixel for all your hard work on Megascans! I hope to see more assets in the near future
I was really struggling with whether I should buy a pack, or just buy the points. Because right now it's not clear how the packs work.
A few notes :
http://i.imgur.com/WQSeY80.jpg
yeah well I mean as far as I can tell its just for compositing megascans textures, blending them together and stuff? I mean the water features etc are cool. I'm just curious where they are going with Megascans Studio, seeing as they already have the quixel suite and stuff.
my initial impression is the studio is fun to play around with, but still obviously a beta. I apologize my original wording was kind of blunt, I wasn't implying there is a lack of vision, i'm just honestly curious to hear the vision. Right now its fun to play around with but i dont see any reason to stop using my current tools i guess.
Megascans is a game changer, maybe i'm not seeing it with the studio. Is it supposed to be a utility for the sole purpose of compositing megascans, or is this some new texturing thing that will solve all my texturing needs? It's a beta so i'm not trying to complain, i'm just giving some honest feedback.
I agree SD is only as good as the machine you're working on.
15360x15360 JPEG (55 MB): https://dl.dropboxusercontent.com/u/30348037/mud_15k.jpg
- It would be nice to view what you have purchased on the website in a library rather than just a List.
- Ability to filter out already purchased results when searching.
- Have an option to toggle page view rather than forever scrolling page. It makes it overwhelming to find exactly what you want.
- Have a notification confirming that you might have missed 'X' when downloading a package. (it wasn't very apparent that I missed the Ztools for 3d models etc).
- Have an option to download both Real time and Offline rendering.
Other than that, Love your tool Quixel your hard work has paid off. Hopefully this feedback helps.(With thumbnails etc so you can visually view what we have already purchased with the points)
1. Library seems to have an overabundance of nature materials (I guess Jungle Book influenced things a little here
2. Freelancer price is great, but Indie price is not amazing. I'd argue it's pretty bad actually. First of all, it's very expensive for tiny studios (I am the only artist at my studio so the up to 10 artists thing seems like overkill) and second, a meager 10 point increase for 4+ times the price seems like a pretty small bump.
3. 2x2 and 4x4 (which are coming from what I gather) costing more points is a little... weird? Why does covering more area result in more points? I just want 4k or 2k maps I can actually use in games, not sure why the real world attributes of the materials affect the price so dramatically.
4. Which brings me back to the pricing issue. Considering that the actual useful materials are those that are 2x2 and upwards, means that 125$ per month, I'll be able to get less than 30 materials per months. Which is not that great value wise.
Jungle, Rainforest, Desert & Wasteland Biomes
Nordic Biome, Australian Lava Ecosystem, 3D Trees & Foliage
Fabric, Brick, Tile & Wood Surface Scans
Metal, Plastic & Rubber Surface Scans
I can't give any ETA on the other waves, but we're steadily working toward processing all of the scans we have and preparing them for the public.
I have used my Megascan trial in order to try to beat our current pipeline with some Megascan textures.
And the result was disapointing. [No worries the service is good but i will explain why i decide to stick with my old pipeline].
So Why Megascan was disapointing ?
-Because we create 16m*16m textures, and if we want to achieve such big texture we need at least 8m*8m inside one texture. [Cuurently i can find 1m texture, or 2m ... really not enough]
-The megascans textures are good quality no doubt. But currently it can't beat my workflow because of size limitations of the scans. [Shooting lot of photos and photomerging them is still easier].
-I need also some function into 'Studio' that will help me to create masks [not just based on height and blending between two materials]. --> i need more classical mask ala photoshop [procedural or handpainted]. Currently i edited everything inside substance because i needed more control.
-We need a function that will blend some piece of scan if we change the target size. Currently the same piece is repeated ... This is the most disapointing point. We need more shoot of the same surface.
-I like to see the material as a ball but i prefer being able to watch the textures themselve in 2d before buying them. [currently it's in perspective --> Pretty but usless from a pure technical point]. And there some shadow sometime on those renders ...
-We need a real naming convention with number as well because currently all concrete are called concrete construction ... Just adding 01, 02, 03 could have helped me to find faster the textures for tweaking.
My conclusion is that --> Current Megascans can be used for small piece of floors , or decals but not for large tilling surface. [seen from plane, or flying cars haha].
The future is bright. Lot of work have been done, and lot still need to be done.
Thanks quixel, now i want you to make all my points false with next release.
Massimo
Management will just laugh their heads off at the Studio Lite pricing. Please help me give you my money
Thanks!!
now I'm really curious about those other planned features :awesome:
For anyone interested, I dragged a couple of the free assets into max and rendered them out with Octane.
I went ahead with a membership, and am working on a bigger scene now. I'll throw it up when I'm finished. This image is just the free rock, one of the free muds, and a whole bunch of depth-of-field for the hell of it.
This guy makes a good comment about 2m:30s in. These elements are often used as secondary elements and not even the focus, so why waste valuable time on the stuff that isn't even the focus of your image. It's almost like making a level inside a game. you set dress, and put assets together from a library and less about actually making the individual assets.
I see Megascans as a hardware store. Carpenters didn't become obsolete or unoriginal because Lowe's or Home Depot are able to sell nails, screws, boards, etc. It meant that more people have access to create and build things at an affordable price.
Of course, just having access to the source materials (or in my example, the raw materials) necessary to create something doesn't ensure that it'll look good. You should see how terrible the new door frame looks that I installed in my bedroom. I had to cover up a bunch of the mistakes with drywall filler. If I were experienced in that sort of thing, I wouldn't have made silly mistakes that led to a lot of rework and touchups.
I'd make the same argument for developing an environment with Megascans. The service doesn't guarantee that everything looks the same. I've already seen huge variations in our scans from work posted on Twitter, Facebook, Artstation, etc.
@Maximum-Dev - 30 days after your subscription - on your billing date.
are these really spheres or smoothed cubes... uvs or some kind of triplanar projection.?
@oglu - http://www.filedropper.com/preview
This is the 3DO project we used to render out the material orbs.
is it broken.?
I think what @oglu and myself are looking for is a simple way to preview one of your material balls, in realtime 3d, with the displaced mud applied to it - either in one of your own viewports or in Toolbag. I'd happily attempt to recreate it in Toolbag by exporting the mud textures from Bridge, but there is no way for us to know the way your sphere is UVed and no way to guess the displacement values that we are supposed to input in Toolbag in order to recreate the exact same displacement (I would assume a standard midpoint and 1x multiplier, but we can't be sure). Basically we are looking for an accurate point of reference to evaluate a few assets, in order to reproduce the look in other environments without any guesswork.
In short : we need either a .xml Quixel project complete with all textures that we can load from the Photoshop Quixel Suite UI ; or even better, a .tbscene with relative paths to provided textures (and neutral lighting, and no color correction, and a provided .obj of the sphere) that we can open in Toolbag in one click. Icing on the cake would be a couple more scenes for other renderers : Vray/Mentalray for Max and Maya, Octane, Keyshot, Blender Cycles, Modo. Again similarly to the way Ten24 provides handy references Modo and Blender scenes complete with a free head scan and shader setup.
http://www.3dscanstore.com/index.php?route=information/information&information_id=16
Thanks !
even better would be an offline render scene...
I'll see what I can do to expedite this process. In the meantime - while you wait - please don't hesitate to remind me if there's no update to this thread in a while. I will check in with our team and see if there's been any progress made on this request.
If anything, a lot of this can be contributed by the community with just a few things needed from you guys. For instance I *think* I noticed that Toolbag causes open seams to appear on displaced models ; to double check this (to potentially report it to the Marmoset devs) the only thing I'd really need is a copy of your material sphere model in OBJ in order to precisely compare their results to your screenshots.
I hope this makes sense !
[edit : just got a link to the sphere through the FB support group]
https://www.dropbox.com/s/mok2ixynr3q9hye/Sphere.obj?dl=0
Here is a quick first attempt in Blender/Cycles. The background is glazed_patio.exr from Substance Painter 2 set to 0.5 intensity + one hemi light for the sun + one point light for an added reflection hotspot. The PBR shader used is from https://www.blenderguru.com/tutorials/pbr-shader-tutorial-pt1/
Albedo, displacement and normals seem to be behave as expected. However I had to boost up the gloss as high as possible (by normalizing levels to full black and white in Photoshop) in order for the mud to appear even just a little bit wet. This is without a doubt a problem with the custom shader and not with the Quixel textures themselves though.
http://i.imgur.com/oONCf65.jpg
That aside, the overall image quality is great :
@oglu : This should explain why the displacement seemed confusing and "too good to be true" at first. The sharp cuts/undercuts you are seeing on the cracked mud do not really come from the displacement map itself (indeed, one would need a vector displacement map at very high tessellation to achieve this sort of look). This high level of surface detail is actually the result of the combination of *both* displacement and normals. (You can actually see that the displacement itself doesn't need to be dense at all, looking at my topmost screenshot).
What I am getting at is that in order to get high-detail results in an offline renderer, you will have to make sure that your renderer also supports normalmaps, and not just displacement. This is actually pretty great because that means that only a few subdivision passes (or even no subdivisions at all) are needed for good renders.
- - - - -
Bottom line : I personally don't know much more than that when it comes to rendering with Cycles, so if anyone could chime in to explain how to set things up properly for gloss/reflections, that'd be great ! Eventually it would be awesome to have a direct link to Blender similar to the already existing Megascan export scripts.
I have tried out a bit of this stuff myself as a hobby, and know how to capture albedo with cross polarized photography, normals with directional lighting and gloss/roughness as a difference of polarized and non polarized shots, haven't done it myself but I figure translucency can be done with some sort of lightbox.
However i'm curious how something like displacement/height is captured or achieved through processing.
Now I realize what I'm asking might be part of the secret sauce and not be something you guys want to share, but I'd appreciate any sort of info or be pointed in the direction of articles/papers/websites dealing with this stuff.
For what its worth here is a thread with my experiments so far
http://polycount.com/discussion/167507/alexs-texture-scans
I think Quixel definitely need to release a sample surface Megascans showing a range of material types next to each other - like a 3*3 grid with Mud, rock, mirror, steel, paper, grass, and so on. That would really help for calibration in renderers that are not supported yet !
Regarding Bridge, our programmer for that tool can be reached directly on our Quixel Live Support group - if you could please post there, we'll get you sorted out ASAP.
Unfortunately, the material scans are 1m x 1m, am I missing something or is all the content 1m? Our target texel density on Gears is 4m = 2048 pixels, this means using megascans to creating tiling textures makes it repeat 4 times before we even tile it in Maya, defeating the point of all those unique pixels in our 2k textures. Could you start taking the scans from a farther distance from the surface, most games I've worked on use roughly 4 meters of unique space relative to the real world for a 2k texture.
Also after I exported my textures out of studio they look extremely pixelated, I think something is broken or you need to apply some filtering to the final blended texture, it looks like nearist neighbor filtering was used to scale it down to 2k, I even tried exporting to 4k and it also looked horrible and unusable.
Probably it was best to just write by the corner of each thumbnail what size it is i.e 1x1, 2x2 etc.
As @Maximum-Dev mentioned, we've got 2x2m scans up as well right now and they're identified by the size of the scan thumbnail. We'll be adding tags to differentiate them soon.
Thank you for keeping an eye here.