The Official aweSurface Test Track

Sven Dullah

So, sadly, after testing two hair models, I get these horizontal black lines and can't get rid of them. I've tried base resolution and SubD, tried turning off bump translucency highlights, basically everything that can be turned off but it makes no difference. Progressive mode looks a little better but the artefacts are still visible, although different. Also noticed that changing the opacity optimization parameters has an impact on how they look, but won't fix the problem. This first test was made with DS 4.7, so I will do some more testing in 4.9 tomorrow;) Except for the black lines it really looks promising.

wowie

Yes. That's expected. The current build relies on 3delight 12 and doesn't work properly with 3delight 11 (which is the version used by DS 4.7). I already have made a build that works with DS 4.7. Need to solve some other bugs too though.

Edit: Already worked out a workaround for point/spot/distant light support. I'll upload a new version tonight/tomorrow.

Sven Dullah

wowie said:

Yes. That's expected. The current build relies on 3delight 12 and doesn't work properly with 3delight 11 (which is the version used by DS 4.7). I already have made a build that works with DS 4.7. Need to solve some other bugs too though.

Edit: Already worked out a workaround for point/spot/distant light support. I'll upload a new version tonight/tomorrow.

Wooow this is good news=) Looking forward...

Sven Dullah

Copying wowie's initial AWE Hair shader tips from the laboratory thread:

Some notes:

• If you're seeing faceting, you will need to convert the prop you're using so it uses subdivision.
• The default samples of 128 will likely produce noise. Use either 512 or 1024 samples to get cleaner renders.
• Although you can use specular textures and diffuse/color textures, the shader is built not to use such textures. If you do want to use hair color maps, you will need to set Melanin and Red Melanin to 0 so that you'll get the color entirely from the map. You can still mix in a bit of melanin/red melanin if you want to though.
• The default melanin/red melanin value should produce a blonde hair look. If you want less redness, simply dial down the red melanin value, giving you either very light blonde with low melanin or dark brown with high melanin.
• If you want additional color controls, you can vary the Hair Color or enable Color Correction to play with gamma and saturation.
• You can also tweak the absorption levels of the root and tip independently of each other.
• AWE Hair uses the Marschner hair model. The first specular corresponds to the R lobe, while the second is the TRT lobe and Translucency is the TT lobe.
• The Translucency Strength also indirectly controls the opacity of the hair. For very complex (multiple layers) hair, enabling this will add to render times. For non focused characters, I recommend dialing this down to 0 to save render times.
• Higher melanin colors or darker hair color will render a bit faster than low melanin / light hair colors.
• If you want almost black hair, dial down hair color strength to at most 25%. A value of 0% will be unrealistically dark.
• The Hair Ramp controls allows you to essentially change the root to tip properties. You can add an offset, make the curve closer to the root or tip, or invert the direction if necessary.
• The Highlight will add some variation to the hair. The controls will likely change though. In essence, it allows you to override the melanin color for specific parts. Unlike the root to tip variation (along the v direction), the highlight control takes effect in the u direction.

wowie

Uploaded the new build.

Some notable changes:

• Backported the raytracing code so the shader works with 3delight 11 / DS 4.7. This should fix the artifacts of stripes/black blocks when rendering in DS 4.7.
• Better support for point/spot/distant lights. Previously, using distant/point/spot light will cause brightness clipping, visible in renders as black spots.
• Raised internal sample cap to 8192 samples. Users will bump into this cap when the shader is set to 1024 samples or higher in the surface tab.
• Tweaked raytracing weights. This should lead to much less noise in areas where there's not much direct light.
• Tweaked the ratio of R and TRT to the final output.

I've also looked into working with Garibaldi hair and consequently, DAZ hair. Although the shader will render, it wouldn't produce what you would expect. So for now, AWE Hair won't support those two.

Sven Dullah

Tks wowie! I finally got to my primary DS rig with 4.9, and made this test. Quite a relief to get rid of the artefacts:) Didn't know about the new build, so will have to install that for the next render. I find the new shader to be very easy to use thus far. And it seems far easier to get good clean results than with aweSurface, no need to tweak and testrender and tweak again:) Very nice!! Also nice to be able to just delete the diffuse maps!

So here I used the Free Spirit hair for G3F. It has normal maps, so I tested that at the same time. Probably not recommended but it seemed to work:) Non progressive render using both HDRI and 3 area lights. On a sidenote, we still have some firefly issues with aweSurface. The hairshader seems to be ok, but here I got some of them on the skin, FYI. The environment sphere is pretty much at default settings, no boosted exposure.

Looking forward to testing the new build

Raw render:

And with some color correction and fixed the fireflies:

wowie

Sven Dullah said:

 I find the new shader to be very easy to use thus far. And it seems far easier to get good clean results than with aweSurface, no need to tweak and testrender and tweak again:)

Yep. Took longer than expected, but with time and countless trials, everything seems to be falling into place. The special 3delight hair brdf is faster than using specular, but it's not exactly easy to work with.

Sven Dullah said:

Very nice!! Also nice to be able to just delete the diffuse maps!

Always hated the baked in highlights. I haven't managed to get better hair details though. Not really a problem with slightly higher detailed hair props, but the lack of details can stick out like a sore thumb on simpler hair props.

Sven Dullah said:

On a sidenote, we still have some firefly issues with aweSurface. The hairshader seems to be ok, but here I got some of them on the skin, FYI. The environment sphere is pretty much at default settings, no boosted exposure.

Most likely to using the current public build of AWE Environment and AWE AreaPT.

Sven Dullah

Tks wowie, the hair shader now works in 4.7. Decided to test a simpler hair (North Hair for G1/G2M), along with some M4 morphs transfered to G1 with GenX. I can see some noise, used 512 samples, so 1024 seems to be the minimum for a clean result. But not bad IMO;) Have a couple of questions for you wowie at this point:

So both specular lobes use the 3DL hair BRDF I assume? Is there a difference between the two, or do they just allow for two sets of specular maps/control maps? Do they both react the same way to the roughness settings? If I want to limit the translucency to just the hair tips, can I use some sort of maps for that? I see there is no such option for translucency, but the highlights channel has a slot, so would that work? Also for the ramp option (which I haven't tested yet) I could use some guidelines;) And finally, in this render I had the opacity optimization at 75, will that produce more noise or is there any other tradeoff using it? Uhm...one more thing...AO distance, is that some sort of shadow bias?

Non progressive render, hair samples 512, Irradiance/SS samples 512/256, 3 area light set up:

 

Sven Dullah

Another quick test with a G1 simple hair prop(Alexios hair). One arealight for direct light and HDRI for ambient/fill light. 1024 samples is definitely better than 512=) Although the part of the hair lit only by HDRI is a bit noisy.

wowie

Sven Dullah said:

I can see some noise, used 512 samples, so 1024 seems to be the minimum for a clean result.

Pretty much. There is still some noise in dark areas at 1024 samples, but it's something I can live with. Eliminating it completely is possible, but at the cost of increased render times.

Sven Dullah said:

So both specular lobes use the 3DL hair BRDF I assume?

The 3delight hair BRDF basically lumps together three specular lobes of the Marschner model into one shadeop.

Sven Dullah said:

Is there a difference between the two, or do they just allow for two sets of specular maps/control maps?

As I noted, the first specular is the R lobe (monochromatic main reflection), second specular is the TRT lobe (colored reflection) and third specular is the TT lobe (translucency). You can control them separately if you want, either via their respective strength settings (with maps) or via the Specular/Global Illumination Exposure overrides. Specular Exposure affects the first specular lobe, while GI Exposure affects the second specular lobe.

Sven Dullah said:

Do they both react the same way to the roughness settings?

The shadeop actually has separate inputs for each lobe. However, as newer Marschner model shows, there's a definite relationship between each of them. This is why in most newer Marschner hair shaders, the roughness is linked (by a certain ratio) and presented as one input to the user.

Most of the rendered look will come from playing with Roughness Longitudinal, which is the roughness that goes along the root to tip axis. Technically, Roughness Azimuthal 'wraps' around the hair cylinder. It doesn't affect the look as much though it may be more noticeable with proper strand based hair.

As with other proper Marschner based hair model, you can move the highlights (to a certain extent) along the root to tip axis by playing with the Root to Tip Shift.

Sven Dullah said:

If I want to limit the translucency to just the hair tips, can I use some sort of maps for that?

With the preview build, you can probably achieve something similar by playing with Absorption Tip. Most hair shades (except for very, very black hair) always have translucency. The amount is controlled by absorption, just like transmission in glass (ie darker color has high absorption and vice versa). I think this is what those opacity masks are trying to do, so I also take into account the opacity level for translucency.

Though actually, I was thinking about adding a parameter to directly further vary translucency / gradient opacity along the root to tip.

Sven Dullah said:

Also for the ramp option (which I haven't tested yet) I could use some guidelines;)

The ramp option basically controls the ratio between the melanin/hair color/absorption at the root and the tip. The offset basically shifts the starting point of the root, while the curve will decrease or increase the transition between root and tip values. Well it should anyway, it looks like it's not working correctly at the moment.

One note, the Absorption Root and Absorption Tip don't require you to enable the 'Hair Ramp' switch.

Sven Dullah said:

And finally, in this render I had the opacity optimization at 75, will that produce more noise or is there any other tradeoff using it?

The Opacity optimization code is more conservative in this shader. What the camera 'sees' is always the raw/original opacity texture value. The opacity optimizations only goes into effect after certain number of ray bounces. It shouldn't affect/introduce additional noise. One way to check is do an IPR, set opacity optimization at 100 and play with the opacity filter value. You should barely see any changes in the render.

I will also use the same code for the next AWE Surface build.

Sven Dullah said:

Uhm...one more thing...AO distance, is that some sort of shadow bias?

It's a dev build debug mode. It's not going to be available in the final build. There is actually a reflection bias for the hair shader, though I don't think I've exposed it on the preview build.

Hmm, I'm rather curious why are you still seeing that level of noise on the hair. Maybe the bump is too strong? Or are you using normal/displacement maps?

Sven Dullah

Tks a lot, this was very helpful. Yes on North hair  I used bump and displacement. On Alexios hair only bump. Will try to reduce the strength, may have bumped it up too high;)

wowie

A test with some fibermesh stuff.

Depending on how the UV is laid out, the ramp stuff can either work or not. But for fur-like stuff, you don't usually get that much variation between root to tip anyway.

Also did a test with and without translucency. The difference is small, but it's there. Enabling translucency makes the color a tad brighter (as it should).

Sven Dullah

wowie said:

A test with some fibermesh stuff.

 

Depending on how the UV is laid out, the ramp stuff can either work or not. But for fur-like stuff, you don't usually get that much variation between root to tip anyway.

 

Also did a test with and without translucency. The difference is small, but it's there. Enabling translucency makes the color a tad brighter (as it should).

Aah that's very cool looking! Next thing to test:)

You were right about the bump strength, I reduced it and get much cleaner result Will post a new testrender shortly;)

Sven Dullah

North hair with less bump/displacement (progressive testrender), 1024 samples:

wowie

I'm still tweaking the shader, so keep that in mind. It's very likely the final shader will be very different than the preview.

Just playing around with North Hair while tweaking the shader.

Sven Dullah

wowie said:

I'm still tweaking the shader, so keep that in mind. It's very likely the final shader will be very different than the preview.

I understand:) Very greatful to have a chance to play with it!

Just playing around with North Hair while tweaking the shader.

Very nice highlights you got there!

wowie

Sven Dullah said:

Very nice highlights you got there!

Thank the 3delight developers for making this possible.

Oh yeah, one other thing. I didn't create any 'Hair Dye' option because you can just simply play around with the Hair Color value to get a dyed hair look, which you can combine with color/strength maps.

I did cheat a bit in this one by raising Specular Exposure to 1, so the primary spec is more visible.

Sven Dullah

wowie said:

Sven Dullah said:

Very nice highlights you got there!

Thank the 3delight developers for making this possible.

Oh yeah, one other thing. I didn't create any 'Hair Dye' option because you can just simply play around with the Hair Color value to get a dyed hair look, which you can combine with color/strength maps.

I did cheat a bit in this one by raising Specular Exposure to 1, so the primary spec is more visible.

Hahaha that cracked me up

Sven Dullah

Was about to do some female stuff but got sidestepped by RawArt, ended up getting Roark for M4, transferred the morphs to Genesis, slapped Evans Hair on him with AWE Hairshader, and hit render...

The hair looks clean but the skin got the odd firefly again with HDRI lighting, env sphere with default settings, well, easy enough to fix in GIMP.

wowie

This is for those who are curious about using Garibaldi / DAZ strand hair with AWE Hair.

The shader will work and render. The only problem is that Garibaldi will override the hair color values of the shader. I haven't checked DAZ strand hair yet, so I don't know if the behaviour is different. My advice right now is just to set the color purely within Garibaldi and setting Melanin/Red Melanin to 0 in the shader. If you want black/darker hair, the hair color strength slider works.

I still need to fix the Hair Ramp and some other stuff. Those wouldn't work as expected with Garibaldi though.

Sven Dullah

wowie said:

This is for those who are curious about using Garibaldi / DAZ strand hair with AWE Hair.

The shader will work and render. The only problem is that Garibaldi will override the hair color values of the shader. I haven't checked DAZ strand hair yet, so I don't know if the behaviour is different. My advice right now is just to set the color purely within Garibaldi and setting Melanin/Red Melanin to 0 in the shader. If you want black/darker hair, the hair color strength slider works.

I still need to fix the Hair Ramp and some other stuff. Those wouldn't work as expected with Garibaldi though.

Sven Dullah

Testing some SS settings...512 adaptive/256 SS samples

wowie

Maybe people can help me here. I saw that 'other' thread and wondered about some things.

Do people know they can turn off global illumination AND reflection, but still use path traced area lights? In this direct lighting only scenario, I'd say AWE Surface is still pretty fast. Did some test renders and with the same settings (pixel samples, irradiance samples etc), render times are about 3/4 of what it was with global illumination and reflection enabled (from 12 min to 9 min @ roughly half K res). I find the difference in quality heavily outweight the render times difference though.

I mean you can still dumb it down further, by disabling GI/reflection and path traced area light and just work with point/spot/distant/ambient occlusion light.

Sven Dullah

wowie said:

Maybe people can help me here. I saw that 'other' thread and wondered about some things.

Do people know they can turn off global illumination AND reflection, but still use path traced area lights? In this direct lighting only scenario, I'd say AWE Surface is still pretty fast. Did some test renders and with the same settings (pixel samples, irradiance samples etc), render times are about 3/4 of what it was with global illumination and reflection enabled (from 12 min to 9 min @ roughly half K res). I find the difference in quality heavily outweight the render times difference though.

I mean you can still dumb it down further, by disabling GI/reflection and path traced area light and just work with point/spot/distant/ambient occlusion light.

Yeah I read your comment, and in fact the whole thread from the start to see what it was all about:) The only useful thing IMO was an official DAZ statement that they have no plans of dropping 3DL. The rest of it was the usual stuff...preconceptions, opinions, whatever.

Frankly, I have no idea what people know or don't know. I know what I know, and if I don't know, I ask you or Kettu. (Or someone will jump in to help, which is always a nice thing.) I just guess most users want and need presets to be able to function. IRay has presets, so IRay it is. 3DL presets are (more or less) a thing of the past, and if they are included with a new product, they are likely to be useless anyway. PAs don't know 3DL and can't use it. I've seen some terrible stuff lately, to be honest.

Sadly it looks like even 3DL users are not aware of the possibilities with aweSurface, pathtracing etc. Or they won't bother because converting stuff is timeconsuming and sometimes difficult. They have their reasons, I'm sure...

Yeah, rant over, back to your question. I haven't actually even tried dumbing down aweSurface. As you said yourself, not worth it, you gain so little by turning off those features IMHO. If I need really fast rendering for animations I go with DS default and even leave out AO = about 30s/frame. Since I'm just a hobbyist, the quality is good enough for my needs, although I would LOVE to use aweSurface, maybe there will be cloud options in the future...

wowie

Sven Dullah said:

Sadly it looks like even 3DL users are not aware of the possibilities with aweSurface, pathtracing etc. Or they won't bother because converting stuff is timeconsuming and sometimes difficult. They have their reasons, I'm sure...

Yeah, rant over, back to your question. I haven't actually even tried dumbing down aweSurface. As you said yourself, not worth it, you gain so little by turning off those features IMHO. If I need really fast rendering for animations I go with DS default and even leave out AO = about 30s/frame. Since I'm just a hobbyist, the quality is good enough for my needs, although I would LOVE to use aweSurface, maybe there will be cloud options in the future...

A few days ago, Paul Bussey of Digital Art Live contacted me about doing some webinar/tutorial stuff. Would having a video tutorial help? Making conversions isn't difficult, it's just tedious. There are tips and tricks you can do to manage that (ie don't try to apply the preset to all materials in the scene in one go etc). They're mostly due to DS script engine being single threaded (as is the animation playback).

Sven Dullah

wowie said:

Sven Dullah said:

Sadly it looks like even 3DL users are not aware of the possibilities with aweSurface, pathtracing etc. Or they won't bother because converting stuff is timeconsuming and sometimes difficult. They have their reasons, I'm sure...

Yeah, rant over, back to your question. I haven't actually even tried dumbing down aweSurface. As you said yourself, not worth it, you gain so little by turning off those features IMHO. If I need really fast rendering for animations I go with DS default and even leave out AO = about 30s/frame. Since I'm just a hobbyist, the quality is good enough for my needs, although I would LOVE to use aweSurface, maybe there will be cloud options in the future...

A few days ago, Paul Bussey of Digital Art Live contacted me about doing some webinar/tutorial stuff. Would having a video tutorial help?

Some users seem to like videos, some like pdf:s, hard to say...couldn't hurt though;)

wowie said:

Making conversions isn't difficult, it's just tedious. There are tips and tricks you can do to manage that (ie don't try to apply the preset to all materials in the scene in one go etc). They're mostly due to DS script engine being single threaded (as is the animation playback).

Well I agree, it's tedious, but for new users it can be truly overwhelming. I just today tried to help a user that wanted to use an IRay hair with 3DL. The conversion script wouldn't work because the hair used the OOT hairblending custom shader. So the transmaps didn't convert. Those kind of things can be gamebreakers for newcomers.

Generally: Since DAZ doesn't provide documentation on how to best use 3DL, someone (like you) could maybe make a step by step tutorial on how to use gamma settings, turn on raytraced shadows, the BASIC stuff:) Darn, I get UPSET when I think about how hard they made it, it's 2019 and still raytracing is turned off by default......not to mention that the pathtracer should have been made available to end users by DAZ... YEARS ago!

Ok I'm calm now, carry on people

 

 

Sven Dullah

So, wowie, what shadow samples did you use on your fibermesh facial hair? It looks absolutely clean, I tried it myself, but 512 samples and progressive rendering was obviously a bad combination;) But yeah, the AWE Hair shader is actually working on fibermesh, yay=) Will try GB hair next!

wowie

Sven Dullah said:

Well I agree, it's tedious, but for new users it can be truly overwhelming. I just today tried to help a user that wanted to use an IRay hair with 3DL. The conversion script wouldn't work because the hair used the OOT hairblending custom shader. So the transmaps didn't convert. Those kind of things can be gamebreakers for newcomers.

Generally: Since DAZ doesn't provide documentation on how to best use 3DL, someone (like you) could maybe make a step by step tutorial on how to use gamma settings, turn on raytraced shadows, the BASIC stuff:)

Yeah, I was thinking along the same things. Start with an intro vid of setting things up DAZ Studio/3delight, then maybe a tutorial on the major settings people need to know. After that, go deeper into things like sampling etc. The new sampling override from the AWE Environment light should help a lot. No more setting samples for individual surfaces.

Sven Dullah said:

So, wowie, what shadow samples did you use on your fibermesh facial hair? It looks absolutely clean, I tried it myself, but 512 samples and progressive rendering was obviously a bad combination;) But yeah, the AWE Hair shader is actually working on fibermesh, yay=) Will try GB hair next!

Hair samples are set to 128 on the shader, but I enabled override on the AWE Environment light so, the shader will use the override value (1024 samples). Generally, you don't need opacity masks or have opacity enabled for fibermesh hair.

Another side note about AWE Hair. Since it's actually using specular rays (as it should), so there's no render times difference between using the standard 3delight renderer (with progressive enabled) and using the scripted renderer. Outside of the typical noise of progressive rendering, the shading is also identical. It makes it easy to work with in IPR, especially with 4.7 and 3delight 11 non dynamically sized render bucket. Technically, DS 4.8/3delight 12 can a bit faster, but the renderer overall is actually a bit slower than 4.7/3delight 11.

Rather unfortunately, you can't adjust settings on the fly for Garibaldi hair or LAMH hair in IPR. I believe the same is true for DAZ strand based hair as well.

Sven Dullah

Re rendered with non progressive 2048 samples- think it looks pretty cool...

Sven Dullah

OOT Sara Hair, progressive, diffusemaps removed...