The Official aweSurface Test Track

Sven Dullah

wowie said:

That's inline to what I'm talking about before. Though my render times don't go up that much.

I did the test with my ancient laptop in 4.7, should try on my main DS rig with a simpler sphere, as you suggested, to see where the sweetspot is;)

wowie said:

If you apply the path traced shadow catcher to the plane, you'll see it will catch shadows. You can cut render times by making a simpler environment sphere. I think a sphere with 12 The AWE Environment Sphere prop is basically a 20 meters sphere primitive with 19 segments and 20 sides, with a -1000 unit offset in y axis. Using a simpler sphere with less segments should allow for less render times, but still avoid excessive distortion.

You still need to use a HDRI with a strong sunlight to cast shadows. If the sun isn't strong enough, then you'll just have flat ambient lighting.

 

wowie

Try this.

1. Make either a geoshell of the environment sphere or load a second copy.
2. Apply the area light shader to one. Make this sphere slightly larger with uniform scaling. I used 110%.
3. Disable Visibility - Occlusion & Indirect Light on the main sphere. We want to get reflection of this sphere, but not diffuse light.
4. Make sure the second sphere with the area light shader applied is set to only emit diffuse light.

Render times won't change, but you will get proper reflection of the main environment sphere as before.

OK. Worked on the shadow catcher as a break from doing the others. This is not doing any environment map lookups, only path tracing from the environment sphere with ambient.

Looks promising with the HDRI's I have. It looks good and useable. I'd already fix the AO like problem where the sphere collides with the plane.

You'd still need to use an area light if you want long shadows.

Sven Dullah

Tks wowie, will have a go, sounds like a working concept=)

Renders look nice

Just saw your edit, nice:)

Sven Dullah

From one thing to another LOL, working on a space cowboy:) WIP;)

 

Sven Dullah

wowie said:

Try this.

1. Make either a geoshell of the environment sphere or load a second copy.
2. Apply the area light shader to one. Make this sphere slightly larger with uniform scaling. I used 110%.
3. Disable Visibility - Occlusion & Indirect Light on the main sphere. We want to get reflection of this sphere, but not diffuse light.
4. Make sure the second sphere with the area light shader applied is set to only emit diffuse light.

Render times won't change, but you will get proper reflection of the main environment sphere as before.

Ok so I tried your tip on this set, made a geoshell with only diffuse lighting, bumped up all the samples and quality settings, hit render and went to sleep. When I woke up I found this:

I though hmm shadow terminator problems, testrendered with various lighting setups but couldn't get rid of the artefacts until I turned off displacement on the ground. So it dawned on me it's the good ol' DoF bug Waay too long focal distance  Adjusted that and did a spotrender and it worked. So will have to redo the test. But yeah your tip seems to work, wowie:)

While I was at it, I reverted to my original light setup, the HDRI using the normal environment shader + a light disc with the arealight shader, placed in the hotspot. Testrendered, the lightdisc naturally cast reflections in the water even if made invisible to the camera. Tested rendering with diffuse light only, now it produced a black reflection in the water, which was kind of unexpected=) Sipping some coffee and thinking how to solve the problem I found the solution, using light categories and trace groups. First time I've used this feature, worked like a charm:) So I selected the env. sphere and typed in hdri in light categories, trace group and trace group membership. Selected the area lighdisc and typed in area. Selected the waterplane and in light categories typed in hdri, area. In trace group and trace group membership I typed in hdri. Selected all the other stuff in the scene, typed in hdri, area in all three. Voila, no reflection from the lighdisc in the water. Damned I'm good

Hmm btw, I noticed that using an opacitymap on an arealight emitter will not have an impact on the reflections it produces.

 

Sven Dullah

Here is a render using the arealight and light categories:

wowie

Sven Dullah said:

 First time I've used this feature, worked like a charm:) So I selected the env. sphere and typed in hdri in light categories, trace group and trace group membership. Selected the area lighdisc and typed in area. Selected the waterplane and in light categories typed in hdri, area. In trace group and trace group membership I typed in hdri. Selected all the other stuff in the scene, typed in hdri, area in all three. Voila, no reflection from the lighdisc in the water.

Yes. Light linking and trace sets is indispensible in production. Especially when combined with AOV/masks.

Kettu is looking into adding AOV to the render script. So keep your fingers crossed.

Sven Dullah

wowie said:

Sven Dullah said:

 First time I've used this feature, worked like a charm:) So I selected the env. sphere and typed in hdri in light categories, trace group and trace group membership. Selected the area lighdisc and typed in area. Selected the waterplane and in light categories typed in hdri, area. In trace group and trace group membership I typed in hdri. Selected all the other stuff in the scene, typed in hdri, area in all three. Voila, no reflection from the lighdisc in the water.

Yes. Light linking and trace sets is indispensible in production. Especially when combined with AOV/masks.

Kettu is looking into adding AOV to the render script. So keep your fingers crossed.

Sven Dullah

Converted  https://www.daz3d.com/lake-village to awe and tested using the double HDRI trick on it. Well.. that method produces more distinct shadows for sure, this particular HDRI gives a strong blue tint to the whole scene, so warmed it up a bit in post. Still, I think I like more the results I get with using one env. sphere with indirect light enabled, and adding a PT arealight for direct sunlight;) And now that I know how to use trace sets I think I can get it to work even better.

khorneV2

Sven Dullah said:

wowie said:

Try this.

1. Make either a geoshell of the environment sphere or load a second copy.
2. Apply the area light shader to one. Make this sphere slightly larger with uniform scaling. I used 110%.
3. Disable Visibility - Occlusion & Indirect Light on the main sphere. We want to get reflection of this sphere, but not diffuse light.
4. Make sure the second sphere with the area light shader applied is set to only emit diffuse light.

Render times won't change, but you will get proper reflection of the main environment sphere as before.

Ok so I tried your tip on this set, made a geoshell with only diffuse lighting, bumped up all the samples and quality settings, hit render and went to sleep. When I woke up I found this:

I though hmm shadow terminator problems, testrendered with various lighting setups but couldn't get rid of the artefacts until I turned off displacement on the ground. So it dawned on me it's the good ol' DoF bug Waay too long focal distance  Adjusted that and did a spotrender and it worked. So will have to redo the test. But yeah your tip seems to work, wowie:)

While I was at it, I reverted to my original light setup, the HDRI using the normal environment shader + a light disc with the arealight shader, placed in the hotspot. Testrendered, the lightdisc naturally cast reflections in the water even if made invisible to the camera. Tested rendering with diffuse light only, now it produced a black reflection in the water, which was kind of unexpected=) Sipping some coffee and thinking how to solve the problem I found the solution, using light categories and trace groups. First time I've used this feature, worked like a charm:) So I selected the env. sphere and typed in hdri in light categories, trace group and trace group membership. Selected the area lighdisc and typed in area. Selected the waterplane and in light categories typed in hdri, area. In trace group and trace group membership I typed in hdri. Selected all the other stuff in the scene, typed in hdri, area in all three. Voila, no reflection from the lighdisc in the water. Damned I'm good

Hmm btw, I noticed that using an opacitymap on an arealight emitter will not have an impact on the reflections it produces.

 

Sven Dullah

khorneV2 said:

Sven Dullah said:

wowie said:

Try this.

1. Make either a geoshell of the environment sphere or load a second copy.
2. Apply the area light shader to one. Make this sphere slightly larger with uniform scaling. I used 110%.
3. Disable Visibility - Occlusion & Indirect Light on the main sphere. We want to get reflection of this sphere, but not diffuse light.
4. Make sure the second sphere with the area light shader applied is set to only emit diffuse light.

Render times won't change, but you will get proper reflection of the main environment sphere as before.

Ok so I tried your tip on this set, made a geoshell with only diffuse lighting, bumped up all the samples and quality settings, hit render and went to sleep. When I woke up I found this:

 

I though hmm shadow terminator problems, testrendered with various lighting setups but couldn't get rid of the artefacts until I turned off displacement on the ground. So it dawned on me it's the good ol' DoF bug Waay too long focal distance  Adjusted that and did a spotrender and it worked. So will have to redo the test. But yeah your tip seems to work, wowie:)

While I was at it, I reverted to my original light setup, the HDRI using the normal environment shader + a light disc with the arealight shader, placed in the hotspot. Testrendered, the lightdisc naturally cast reflections in the water even if made invisible to the camera. Tested rendering with diffuse light only, now it produced a black reflection in the water, which was kind of unexpected=) Sipping some coffee and thinking how to solve the problem I found the solution, using light categories and trace groups. First time I've used this feature, worked like a charm:) So I selected the env. sphere and typed in hdri in light categories, trace group and trace group membership. Selected the area lighdisc and typed in area. Selected the waterplane and in light categories typed in hdri, area. In trace group and trace group membership I typed in hdri. Selected all the other stuff in the scene, typed in hdri, area in all three. Voila, no reflection from the lighdisc in the water. Damned I'm good

Hmm btw, I noticed that using an opacitymap on an arealight emitter will not have an impact on the reflections it produces.

 

The learning process can be painful at times=))

Sven Dullah

Does anyone have a tip on how to handle specular/reflections on garibaldi hair, especially brows? I've tried different amounts of roughness and strength, but it seems extremely tricky to get right. Made a character, created the brows and set up the surface inside the GB editor, then converted to awe and adjusted diffuse strength to get a dark brown color. Mid gray specular color, about 3% roughness and maybe 10% specular strength. Looked ok when testrendered with a HDRI. Then added an area light and the brows turned white. Had to set spec strength to 1% with a dark brown color to fix that. What would be the proper roughness/BRDF setting? Or should I just skip specular altogether?

Gone

I find that specular always seems to blow out Garibaldi and SB hair. I usually turn off specular and just use reflection at around 10%.

Sven Dullah

Gone said:

I find that specular always seems to blow out Garibaldi and SB hair. I usually turn off specular and just use reflection at around 10%.

Tks, good idea Probably wiser to try adjusting the salt n' pepper setting and turn off specular.

Gone

Oh yeah! That's the other thing.

When you use AWE shader, it's pretty much a given that you have to turn salting off or, at least, set it to very low values. The colour of the hair will make a big difference in how much salting is visible. Peppering usually isn't as noticable unless the hair is a very light colour.

A long time ago, I seem to recall reading that specular was a quick and dirty replacement for reflection. As Wowie mentioned elsewhere, reflection in the AWE shader is very fast so, as a general rule, I normally turn specular off for pretty much everything and just use reflection.

I find it gives that little bit extra boost to realism when you don't have a faked specular highlight showing on the surface - especially shiney objects.

Mustakettu85

Gone said:

A long time ago, I seem to recall reading that specular was a quick and dirty replacement for reflection.

If you're using aweSurface and path-traced lights, specular is fine. It's an optimised reflection of the lights only. 

If you are using "oldschool" lights, then yes specular is an approximation even with aweSurface (that's just the way "oldschool" lights work, they do not "physically exist" in the scene as geometry; they're basically just a mathematical direction and/or origin).

Sven Dullah

Found yet another older set that I haven't used for anything. I remember trying to testrender it with UE2 in bounce light mode, aborted after 24 h IIRC;) So thought I'd do a proper render with awe. First attempt didn't go to well:

...Had to enable progressive, 8h 34 min: (12x12 pixel samples)

Not done with this yet, I seem to have lost some normal maps along the way, so will have to revisit. Also need to work on the outside.

timeoff

Sven Dullah said:

Found yet another older set that I haven't used for anything. I remember trying to testrender it with UE2 in bounce light mode, aborted after 24 h IIRC;) So thought I'd do a proper render with awe. First attempt didn't go to well:

...Had to enable progressive, 8h 34 min: (12x12 pixel samples)

Not done with this yet, I seem to have lost some normal maps along the way, so will have to revisit. Also need to work on the outside.

Very nice indeed, but the rendertime is a bit of a killer. I'm too keen on instant gratification, I think the longest I have been prepared to wait is about 90 minutes. Perhaps I should be more patient ;)

Sven Dullah

timeoff said:

Sven Dullah said:

Found yet another older set that I haven't used for anything. I remember trying to testrender it with UE2 in bounce light mode, aborted after 24 h IIRC;) So thought I'd do a proper render with awe. First attempt didn't go to well:

 

...Had to enable progressive, 8h 34 min: (12x12 pixel samples)

 

Not done with this yet, I seem to have lost some normal maps along the way, so will have to revisit. Also need to work on the outside.

Very nice indeed, but the rendertime is a bit of a killer. I'm too keen on instant gratification, I think the longest I have been prepared to wait is about 90 minutes. Perhaps I should be more patient ;)

I hear you=) Well some sets just want to render for hours and days, no matter what you try. And I often end up spending more time trying to find the culprit than what it takes to actually render it out:) Hmm, with this light setup I could probably cut an hour or something by tweaking some settings, maybe I'll try a different approach for the next render. Would be interesting to see an IRay rendertime comparison...

kyoto kid

...now this is where a fast high core count CPU and lots of memory are useful.  

Wish Daz would implement networked rendering like Carrara does. I'd have 20 threads and 56 GB of memory to throw at it through a gigabit router.

Sven Dullah

There's the 3Delight cloud rendering, but looks like DS is not yet supported:(

https://www.digitalengineering247.com/article/3delight-cloud-and-3delightnsi-debut-as-new-rendering-tools/

https://www.3delight.com/signup-landing-page

kyoto kid

...yeah. 2¢ per 24 cores per minute is not that bad.  That's 8¢ for 96 CPU cores which for even an hour would be under 5$, fairly reasonable charge for a final render considering getting the same core count on your own system would cost 1,000's.

wowie

Sven Dullah said:

...Had to enable progressive, 8h 34 min: (12x12 pixel samples)

Not done with this yet, I seem to have lost some normal maps along the way, so will have to revisit. Also need to work on the outside.

Nice render, but could probably be faster (even without progressive). From my testing, these kind of shots can be done in around 30 minutes at 1280x720 at 8x8 pixel samples. Some points to consider:

• Unless you're using DOF, there's really no need to go higher than 8x8 pixel samples. Using 10x10 or 12x12 pixel samples will raise render times by about 2x.
• Higher resolution will take more time. Rule of thumb seems to be 4 times the number of pixels will be 3x amount of time.
• Noise is due to irradiance samples, not pixel samples. You can either raise the number of samples or just use more light/emitters in your scene. I recommend sticking to max 2048 with the current AWE Surface 1.2 build.
• Simple, large emitters is better/faster than many small ones. Complex ones always means longer render times.
• If you don't mind a bit of cheating, turn off diffuse/GI for black surfaces and rely solely on specular/reflection.
• Use single lobe Ashihkmin Shirley specular/reflection whenever possible. Before using the second base lobe, try using Glossy Fresnel instead.

Btw, the new build is just a bit faster and don't necessarily need as much samples as before. I'd say quality wise, what was 2048 samples before is now 1024 samples. With my usual test scenes, using 1024 samples instead of 2048 samples means 2/3 the render time.

I did notice that DS 4.7 is still slight faster than 4.11 (using 4.11.0.366), all other things being equal. Using the standalone renderer (12.0.59) is still faster than both.

Of course, hardware comes into play as well. I'm using the Intel Core i7 4770K, which is quite old now. Looking forward to the new 3rd gen Ryzen/Threadripper. Like PC World says "Damn, this CPU is fast." Something like the Ryzen 3700 should give you 1/2 the render times. Though that Ryzen 3900 is tempting, I'm waiting for the 3950 to launch before getting one.

Sven Dullah

wowie said:

Sven Dullah said:

...Had to enable progressive, 8h 34 min: (12x12 pixel samples)

 

Not done with this yet, I seem to have lost some normal maps along the way, so will have to revisit. Also need to work on the outside.

Nice render, but could probably be faster (even without progressive). From my testing, these kind of shots can be done in around 30 minutes at 1280x720 at 8x8 pixel samples. Some points to consider:

• Unless you're using DOF, there's really no need to go higher than 8x8 pixel samples. Using 10x10 or 12x12 pixel samples will raise render times by about 2x.
• Higher resolution will take more time. Rule of thumb seems to be 4 times the number of pixels will be 3x amount of time.
• Noise is due to irradiance samples, not pixel samples. You can either raise the number of samples or just use more light/emitters in your scene. I recommend sticking to max 2048 with the current AWE Surface 1.2 build.
• Simple, large emitters is better/faster than many small ones. Complex ones always means longer render times.
• If you don't mind a bit of cheating, turn off diffuse/GI for black surfaces and rely solely on specular/reflection.
• Use single lobe Ashihkmin Shirley specular/reflection whenever possible. Before using the second base lobe, try using Glossy Fresnel instead.

Btw, the new build is just a bit faster and don't necessarily need as much samples as before. I'd say quality wise, what was 2048 samples before is now 1024 samples. With my usual test scenes, using 1024 samples instead of 2048 samples means 2/3 the render time.

I did notice that DS 4.7 is still slight faster than 4.11 (using 4.11.0.366), all other things being equal. Using the standalone renderer (12.0.59) is still faster than both.

Of course, hardware comes into play as well. I'm using the Intel Core i7 4770K, which is quite old now. Looking forward to the new 3rd gen Ryzen/Threadripper. Like PC World says "Damn, this CPU is fast." Something like the Ryzen 3700 should give you 1/2 the render times. Though that Ryzen 3900 is tempting, I'm waiting for the 3950 to launch before getting one.

Tks wowie! I was using DoF with a high F/stop to simulate the pixelfilter (kind of), as was using progressive mode. And for an artistic render you probably would want to use a stronger DoF. Yeah I used 2048 Irradiance samples for the ceiling and walls, and 4 large 1 poly emitters + the HDRI. I'll try to optimize it a bit more, will be interesting:)

I have the DS 4.9 on my main comp. Pretty happy with it, so not likely to update any time soon. Can do dForce simulations in 4.10 beta. GB hair works well for my strand based hair needs, so don't want to mess things up and possibly break something:) And frankly, the newest builds obviously need to be sorted out. Can't see any 3DL related fixes either;)

Still no luck finding the stand alone renderer. So if you happen to stumble on it, let me know:)

Mustakettu85

Sven, if all else fails, you could probably ask on the 3Delight forums. They dug out a super ancient version for someone once.

Another idea for getting some extra "oomph" without much (if any) rendertime investment is using thin film. Where's that link... here: https://blog.maxwellrender.com/tips/pushing-the-limits-of-realism-of-materials/

Sven Dullah

Mustakettu85 said:

Sven, if all else fails, you could probably ask on the 3Delight forums. They dug out a super ancient version for someone once.

Worth a try, tks!

Mustakettu85 said:

Another idea for getting some extra "oomph" without much (if any) rendertime investment is using thin film. Where's that link... here: https://blog.maxwellrender.com/tips/pushing-the-limits-of-realism-of-materials/

Tks, nice article! Hmm I guess one needs to experiment some more with thin film spec. rougness then, and roughness maps. Will look into that;)

Sven Dullah

...a character I'm working on, the hair may have to go though... With this 4 arealight setup 256 SS- and 1024 Irradiance samples seem to be sufficient. 8x8 pixelsamples, 19 min. Still need to make some GB brows for her;) Used the AshikhminShirley double specular lobe approach for the skin and hair.

wowie

Well, this took some time figuring out. But now it works.

AWE Surface now has proper adaptive sampling. With this, you don't need to manually raise samples to get cleaner renders. From testing, I'm seeing comparable quality with 1024 samples and adaptive sampling compared to renders with 4096/8192 samples, at around half the render times.

Tested using this set - https://www.daz3d.com/apartment-kitchen

Sven Dullah

wowie said:

Well, this took some time figuring out. But now it works.

AWE Surface now has proper adaptive sampling. With this, you don't need to manually raise samples to get cleaner renders. From testing, I'm seeing comparable quality with 1024 samples and adaptive sampling compared to renders with 4096/8192 samples, at around half the render times.

Tested using this set - https://www.daz3d.com/apartment-kitchen

I'll be damned, that's simply beautiful! Sounds like the next build will be 2.1 rather than 1.3

 

wowie

Sven Dullah said:

I'll be damned, that's simply beautiful! Sounds like the next build will be 2.1 rather than 1.3

With it enabled, the shader will internally try to use very high samples most of the time. So, it'll feel like it's slower.

For convenience sake, I'll add a toggle button so you can disable it when fiddling with materials in IPR.