Best real world setup in order to match physical materials?

[thebeals]

Hey everyone,

I am really interested in hearing how all of you create your materials to match a physical sample. Specifically, how do you set up your real world scene to have something to visualize in order to replicate in 3d?

Do you look at it under a lamp at your desk and move the sample around?
Do you put it in a calibrated light box and take a photo?
Do you take photos of it under various known lighting conditions that are replicated in 3D in order to try and match?

I am really curious and struggling myself to find an approach that I can consistently use. Thanks!!

[snivlem]

If it looks good enough, it's good enough in my opinion. Most of the time it seems that you have to just eyeball it. I work in Arch Viz, so materials are critical but there are always cases just to enhance or correct it in Photoshop because if it is behaving correctly, architects always seem to want me to tweak it!

The only way to get it physical is to scan it with Chaos Groups new service in my opinion or unless you know the iOR of whatever you are working on and all the other elements of the material.

[thebeals]

Yea, that's always been my mentality as well. Unfortunately, in my current position it needs to be as accurate as possible because decisions will be made based on the look of the material and if it's not accurate in as many situations as possible, then that's not ok. :/ We are going to be using scan data and are looking into different options, but even with scan data we are realizing that they're not capturing all the detail correctly and that manual intervention is necessary. (We haven't looked into Chaos Group yet, but that is on the road map to check out soon). We are seeing a lot of angle dependent effects (such as darkening in satin) are difficult for solutions to capture accurately. We've even thought of creating our own custom rig to capture the brdf, but the development time seems pretty intense. The question is, if we have to go in and adjust things, what's the best way to do that in the best environment for consistency.

I don't know the technicals of material creating. I know how to create realistic materials by adjusting stuff until its right, but I'm pretty hazy on what brdf is needed when, how to create one, how to model a satin shader correctly without having to guess with falloff ramps etc... (Any tips greatly appreciated btw )

My thought is somewhere along the lines of:

have a light box time scenario that has different lighting possibilities
have a complete match to each one in Maya/Vray with the same model as in the photograph
start with an overall lighting and tweak until its close, then turn off all lights but one from different angles to get a better idea of translucency, asperity at the edges, fresnel etc...

but this could be complete overkill and destroy productivity and may not even be needed.

Thanks for the input, I greatly value it, and any other thoughts I would love to hear!

[Art48]

You could always capture an HDRI (with the FULL range ^^) of your office interior with artificial light, or somewhere outside - in short somewhere where you actually can be or are while working.
That way you have captured the lighting conditions and can reproduce them in 3D and match your materials from there.

[Pixelcon]

Sorta like this? Watch til about the 27 mins mark for a brief example.
https://youtu.be/FQMbxzTUuSg?t=22m5s

Good watch.

[alias_marks]

I also tend to do the guess and check method, and have not seen a compelling 'scientific' way to recreate a material with accuracy. I think at some point the decision has to be made to do a real world physical mockup and/or order the material to the site where it will be installed, and view it over a period of time with different lighting conditions. With all the spinners and nobs I think it can be easy enough to 'expect' that we should be able to re-create an accurate image out of the computer, but if a certain level of scruitiny is involved, nothing beats catching the nuance or a real world mockup in my opinon. *at least for now

[Deflaminis]

I would break it down into 2 main chunks:

1) Recreating the physical lighting conditions. Using reference photography, some HDRI's of your exact lighting setup, and some coffee and patience. Matching angles, intensities, and temperature being the main goal there, and trying to get reflections just so. Maybe start with a chrome ball, so you know the materials are more easily replicated, especially if scanned data is involved this might be a good option. If you want to recreate photographic scnenarios, maybe get HDRIs of your actual reflectors, diffusers, etc. Note that even these wouldn't be "perfect" as is, because the original light sources are not planar. The closest you could come is to map them onto geometry that matches the reflectors/sources. (And mesh lights are kinda slow/noisy, but with 3.4 maybe that doesn't matter much anymore.) That is, if a photographic result is the intended output.

2) Then it's time to nail the materials. Due to the nature of BDR functions, it's a tough one. (Whether mental ray, vray, or what have you.) We can't replicate the n and K funtions of light transmission exactly (given the nature of how max and maya deal with the outputs) but we can come pretty close by splitting up the r g an b channels separately when defining our reflection falloff curves. (For example, color changes in metals are more feasible here.)

Grant Warwick has some good tutorials on getting this method started, and the good folks at Siger Studio put out a free plugin that can read your n/k values from refractiveindex.info. Here is the link: https://www.sigerstudio.eu/shop/plug...omplexfresnel/

Not sure if a maya version exists, but it's not too much work to replicate.

As you mentioned you may have materials that look right at one angle, but weird at other angles. Typically due to the lighting, limitations in fresnel complexity, or other such things. Experience has shown me that in these cases blending two different materials together based on viewing angle does work quite well, though it tends to slow down the rendering. The alternative is of course to boogie down with the new GGX shaders, but it's still hard to get them exactly right from all angles. This is compounded by looking at photos, where photographers use a number of tricks to get highlights in specific places. Some materials just look bad in some conditions, even in real life. (Example: http://m.cheersparty.com/pages/image...ches%20010.jpg)

Last but not least the scanned materials are pretty sweet, but from what I understand they lack the ability to really be changed much. So that's good and bad. Good because you always know that your lighting is the problem if there is one, but also hard to tweak for "artistic" purposes. However in your case this may not be much of a deal breaker, as lighting is generally easier to setup (especially for junior artists) than material work. And often less frustrating for them.

Finally you might also consider using "pre-made" stages (once you have your mats) which are well setup and lit so hopefully you can get a nice plug and play situation.

All of this just my 2 cents, interested in seeing what Vlado or the usual suspects have to add. (Or, if I'm off, please set me straight.)

EDIT: So many typos, tried to type it all between test renders....