The Official aweSurface Test Track

1293032343566

Comments

  • Sven DullahSven Dullah Posts: 7,621
    wowie said:

    Looks nice, you have a link?

    https://www.daz3d.com/manga-student-for-genesis-2-female-s

    Just in case, the pencil skirt is from this set

    https://www.daz3d.com/sexy-librarian-for-genesis-2-female-s

    Tks, wishlisted;)

    wowie said:

    Yup I get fireflies on old scenes, confirmed;)

    OK. That's actually expected behaviour, which is why I haven't posted on the freebie thread yet. I pretty much revised the output to 12 f-stops. I didn't get any fireflies on my end, but there's also some updated on the area lights and environment shaders. I've changed the f-stops limits to 10 and uploaded the update. Let me know if there's still fireflies.

    Will do!

  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    Hmm I still get the odd firefly both in progressive and non progressive mode, but it's much better. Only tested that latest woman scene thus far, with bump, displacement and SS enabled both on skin and hair I got exactly one firefly in her right eye and two on the hair:) That's non progressive, with progressive I got a few more.

    image

    Zola awe firefly test non progressive.png
    1800 x 1913 - 5M
    Post edited by Sven Dullah on
  • Mustakettu85Mustakettu85 Posts: 2,933
    wowie said:

     The three of us must have roughly the same tastes then. The sweater in that test render is also from Oskarsson.

    Manga Student :) I guessed before I saw the link you posted. I don't have it, though - Duffle Dream comes with a very similar bottom, and then there was the Real Feel Cozy Turtleneck... :) Real Feels aren't Oskarsson's, but the whole series is quite nice.

  • Mustakettu85Mustakettu85 Posts: 2,933

    Hmm I still get the odd firefly both in progressive and non progressive mode, but it's much better. Only tested that latest woman scene thus far, with bump, displacement and SS enabled both on skin and hair I got exactly one firefly in her right eye and two on the hair:) That's non progressive, with progressive I got a few more.

    It's interesting to compare the first non-postworked render and this one - have you done anything to scene colour temperature, lights or any skin setting? The framing looks a bit different, so I guess you did do something else as well, right?

  • Sven DullahSven Dullah Posts: 7,621

    Hmm I still get the odd firefly both in progressive and non progressive mode, but it's much better. Only tested that latest woman scene thus far, with bump, displacement and SS enabled both on skin and hair I got exactly one firefly in her right eye and two on the hair:) That's non progressive, with progressive I got a few more.

    It's interesting to compare the first non-postworked render and this one - have you done anything to scene colour temperature, lights or any skin setting? The framing looks a bit different, so I guess you did do something else as well, right?

    Yes I did fiddle a bit with SS settings, the first version was a bit reddish, tried to find more of an olive tone. Also raised light intensity for the main light some, and zoomed in a bit, just to come a bit closer, not an artistic choice=)

  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    ...and a progressive render, made the environment lighting a bit colder, turned off SS on the hair, made specular color for the skin a tad darker...zoomed back out a bit=)) Her forarms and hand palms are too glossy, I use specular strength maps here, so need to adjust them I think.

    image

    Regarding the fireflies, not sure if the one in her right eye actually is a highlight in the tear? Tried to zoom in to have a look but it's impossible to say for sure. The hair though has two or three of them.

    Zola awe firefly test progressive.png
    1800 x 1913 - 4M
    Post edited by Sven Dullah on
  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    Testing some shiney stuff...a few observations: You really need to be careful with specular levels, highlights tend to blow out surfaces totally, both metals and dialectic. I first set the windows the way I'm used to, and they rendered solid white. After a bit of fiddling I ended up disabling specular and only using reflection. I also wanted to test the new adaptive sampling with transmission, so I used both opacity and transmission on the windows, and enabled adaptive sampling on the interior. Result: doubles the rendertime, more or less. But no fireflies:) Well, this simple scene with HDRI lighting and Kettu's radiumcatcher took 45 min at 10x10 ps non progressive, after disabling adaptive sampling. The windows are the culprit.

    image

    And same scene with progressive on, 22 min, interesting:)

    image

    Only thing that looks different is the window glass, which of course was expected. Would be very interesting to know exactly how progressive mode simplifies the rendering of transmission, refraction etc.

    F550 Sports Car awe.png
    1920 x 1080 - 4M
    F550 Sports Car awe progressive.png
    1920 x 1080 - 4M
    Post edited by Sven Dullah on
  • wowiewowie Posts: 2,029
    edited August 2019

    Hmm I still get the odd firefly both in progressive and non progressive mode, but it's much better. Only tested that latest woman scene thus far, with bump, displacement and SS enabled both on skin and hair I got exactly one firefly in her right eye and two on the hair:) That's non progressive, with progressive I got a few more.

    No problem. Just need to use lower f-stops,which is the old value I used for the 1.3 build. I will be adding the option to change the clamp via the updated AWE Environment Light (when the update for the commercial pack is ready). Also discovered some small bugs along the way.

    Testing some shiney stuff...a few observations: You really need to be careful with specular levels, highlights tend to blow out surfaces totally, both metals and dialectic. I first set the windows the way I'm used to, and they rendered solid white. After a bit of fiddling I ended up disabling specular and only using reflection.

    Well, you can just dial down the specular exposure. That will bring down the sterngth of specular highlights without affecting reflection strength. Though to be honest, if you're getting specular blowouts, I think you need to change your lights.

    I also wanted to test the new adaptive sampling with transmission, so I used both opacity and transmission on the windows, and enabled adaptive sampling on the interior. Result: doubles the rendertime, more or less. But no fireflies:) Well, this simple scene with HDRI lighting and Kettu's radiumcatcher took 45 min at 10x10 ps non progressive, after disabling adaptive sampling. The windows are the culprit.

    Well, if it's solid glass, you don't really need opacity. Opacity should only be used for mask/cutouts. If you want to use 'old school' glass, simply enable opacity, set opacity strength to 0, then make sure Multiply Specular/Reflection with Opacity is disabled. Don't enable transmission if you use this approach. Diffuse rays will just pass the glass since its fully transparent.

    If you want refracting glass, you can't use opacity. Well, you can, but render times is going to go up.

    Here's what I come up with - 7 min 2.81 secs - 8x8 pixel samples, 512 irradiance samples.

    The ground plane is using my updated shadow catcher. Windshield windows is non refracting glass, so I enabled transmission and enabled thin glass. I did lowered the opacity strength, but keep opacity disabled mostly so it looks 'right' in the viewport.

    7 minutes 2.81 seconds.jpg
    1067 x 600 - 355K
    Post edited by wowie on
  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019
    wowie said:

    Hmm I still get the odd firefly both in progressive and non progressive mode, but it's much better. Only tested that latest woman scene thus far, with bump, displacement and SS enabled both on skin and hair I got exactly one firefly in her right eye and two on the hair:) That's non progressive, with progressive I got a few more.

    No problem. Just need to use lower f-stops,which is the old value I used for the 1.3 build. I will be adding the option to change the clamp via the updated AWE Environment Light (when the update for the commercial pack is ready). Also discovered some small bugs along the way.

    Ok, cool!

    wowie said:

    Testing some shiney stuff...a few observations: You really need to be careful with specular levels, highlights tend to blow out surfaces totally, both metals and dialectic. I first set the windows the way I'm used to, and they rendered solid white. After a bit of fiddling I ended up disabling specular and only using reflection.

    Well, you can just dial down the specular exposure. That will bring down the sterngth of specular highlights without affecting reflection strength. Though to be honest, if you're getting specular blowouts, I think you need to change your lights.

    Ah yeah reducing specular exposure is a good idea. That was with just HDRI lighting, nothing special.

    wowie said:

    I also wanted to test the new adaptive sampling with transmission, so I used both opacity and transmission on the windows, and enabled adaptive sampling on the interior. Result: doubles the rendertime, more or less. But no fireflies:) Well, this simple scene with HDRI lighting and Kettu's radiumcatcher took 45 min at 10x10 ps non progressive, after disabling adaptive sampling. The windows are the culprit.

    Well, if it's solid glass, you don't really need opacity. Opacity should only be used for mask/cutouts. If you want to use 'old school' glass, simply enable opacity, set opacity strength to 0, then make sure Multiply Specular/Reflection with Opacity is disabled. Don't enable transmission if you use this approach. Diffuse rays will just pass the glass since its fully transparent.

    If you want refracting glass, you can't use opacity. Well, you can, but render times is going to go up.

    I thought you had to have opacity enabled for diffuse rays to pass?

    wowie said:

    Here's what I come up with - 7 min 2.81 secs - 8x8 pixel samples, 512 irradiance samples.

    The ground plane is using my updated shadow catcher. Windshield windows is non refracting glass, so I enabled transmission and enabled thin glass. I did lowered the opacity strength, but keep opacity disabled mostly so it looks 'right' in the viewport.

    Ok tks for the input, need to test some more. Not feeling 100% comfortable with the specular behavior atm. Things are not reacting the way I expect, need to find a new point of reference;) Btw, the glass in your render looks ok, but I can see some noise inside the carwink

    Post edited by Sven Dullah on
  • wowiewowie Posts: 2,029
    edited August 2019
    I thought you had to have opacity enabled for diffuse rays to pass?

    If you want to use the old school way of doing non refractive glass. It's a hack, but a good one. Enable opacity and set opacity strength to zero. Doing so means diffuse and shadow rays will go through. If you want to make sure, also disable Visibility - Occlusion & Indirect Diffuse and Visibility - Shadows.

    Ok tks for the input, need to test some more. Not feeling 100% comfortable with the specular behavior atm. Things are not reacting the way I expect, need to find a new point of reference;) Btw, the glass in your render looks ok, but I can see some noise inside the carwink

    Yes. That was a quick test after all. You can always add an area light if you feel there's too much noise. The old school glass hack produces the least amount of noise for that type of scenarios. If you want some absorption, then you'll have to use transmission with thin glass enabled (with opacity disabled). Make sure you turn off Visibility - Occlusion & Indirect Diffuse and Visibility - Shadows.

    I also enabled anisotropy (100%) and glossy Fresnel (20%) with specular roughness set to 1% for the glass. These settings help 'spread' the highlights and provide extra control for roughness at/near grazing angles. I tend to use these values for roughness (0%, 1%, 2.5%, 5%, 7.5%, 15%, 20%, 35% and 50%) with the default Ashihkmin Shirley BRDF. They provide just enough variation and difference between each settings.

    Left : 100% Dielectric/ 0% Metalness. Right :35% Metalness

    Car paint is technically a dielectric, but sometimes mixing in a bit of metalness works.

    Having a bit of thin film on metal and glass will also add that little bit of extra touch. Don't mind the settings, I was just randomly selecting thin film thickness and IOR values.

    10 minutes 59.24 seconds.jpg
    1067 x 600 - 356K
    11 minutes 19.37 seconds.jpg
    1067 x 600 - 355K
    11 minutes 20.22 seconds.jpg
    1067 x 600 - 365K
    Post edited by wowie on
  • Sven DullahSven Dullah Posts: 7,621

    ...speaking of car renders...what, in your opinion, would give the most realistic result (talking window glass), if you don't worry about rendertimes? Because I want it alllaugh. I want colored glass with colored shadows, and I want diffuse- and shadow rays to go through. So, in my car render above, that's why I ended up with pretty weird settings for the glass. IIRC transmission strength 30%, a bit of absorption with a greenish color, abs. shadows enabled, thin glass on, opacity strength about 0.1, you get the picture:) Using those values you'd better have progressive on=) Well this still renders in a reasonable amount of time, but better not use adaptive sampling for the interior LOL. So what would your settings be like, say, if you wanted to render a babe or dude in the car, beautiful sunrise HDRI that shines through the colored glass and tints his/her face green and the hair casts beautiful shadows onto the seat or whatever? laugh

  • wowiewowie Posts: 2,029
    edited August 2019

    ...speaking of car renders...what, in your opinion, would give the most realistic result (talking window glass), if you don't worry about rendertimes? Because I want it alllaugh. I want colored glass with colored shadows, and I want diffuse- and shadow rays to go through. So, in my car render above, that's why I ended up with pretty weird settings for the glass. IIRC transmission strength 30%, a bit of absorption with a greenish color, abs. shadows enabled, thin glass on, opacity strength about 0.1, you get the picture:) Using those values you'd better have progressive on=) Well this still renders in a reasonable amount of time, but better not use adaptive sampling for the interior LOL. So what would your settings be like, say, if you wanted to render a babe or dude in the car, beautiful sunrise HDRI that shines through the colored glass and tints his/her face green and the hair casts beautiful shadows onto the seat or whatever? laugh

    Window glass is basically non refracting glass.

    512 samples for both.

    Left without Use Face Forward, Right with Use Face Forward.

    So you'll want to not use the old school hack and go with transmission with thin glass enabled. To get colored glass then use a saturated transmission color and dial up transmission scale/absorption. If you want to minimize noise, you need to enable 'Use Face Forward' for the glass. I didn't enable that on my previous renders, which is probably why there's a ton of noise on the interiors.

    Here are shots with 2048 samples. 8x8 pixel samples same as before.

    Left Progressive enabled. Right Progressive disabled.

    If you want to use progressive, stick to the old school hack for window glass (to avoid excessive noise). However, progressive is always going to be noisier. It won't be as noticeable, but the noise will still be there. As you can see, there's barely any noise with non progressive.

    These are done with area lights. Since global illumination uses brute force sampling, it will be noisier with a HDRI lit scene. So, add an area light to your HDRI scene to avoid/minimize noise.

    Glass1.jpg
    364 x 600 - 40K
    Glass2.jpg
    364 x 600 - 48K
    Glass3.JPG
    324 x 450 - 42K
    Glass4.jpg
    364 x 600 - 44K
    Glass5.jpg
    364 x 600 - 36K
    Post edited by wowie on
  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    Ok so I needed to go back to basics to try and sort out the matter of transmission vs opacity:) Simple test scene: one emitter, environment dome with only reflection/refraction, a glass plane, and some primitives.

    First render: glass set to 100% transmission, no diffuse, 100 specular strength 0% roughness, transmission color 0, 255,255, thin glass 100% transmission color, transmission scale 4.00.

    image

    Obviously to get colored shadows you need to have visibility/shadows on, but I learned that also occlusion/indirect light has to be on. Due to the high transmission scale setting the glass is almost opaque.

    So lets dial in 50% opacity:

     

    image

    and to get more reflection let's set metalness to 50:

    image

    And finally, setting opacity to 0 means no transmission shadows:

    image

    Not sure what my point is with this quick and dirty testlaugh. Maybe that to get the right amount of color, reflections and transparency you need to use both transmission and opacity, shadows and occlusion/indirect light need to be enabled. But then again, if speed is important, skip transmission altogether and do it the oldschool way.

    Still has to test the same with HDRI light, I have a feeling it will behave quite differentlyfrown

     

    transmission 100.png
    800 x 494 - 642K
    transmission 100 opacity 50.png
    800 x 494 - 639K
    transmission 100 opacity 50 metallicy 50.png
    800 x 494 - 636K
    transmission 100 opacity 0 metallicy 50.png
    800 x 494 - 634K
    Post edited by Sven Dullah on
  • Sven DullahSven Dullah Posts: 7,621
    wowie said:

    ...speaking of car renders...what, in your opinion, would give the most realistic result (talking window glass), if you don't worry about rendertimes? Because I want it alllaugh. I want colored glass with colored shadows, and I want diffuse- and shadow rays to go through. So, in my car render above, that's why I ended up with pretty weird settings for the glass. IIRC transmission strength 30%, a bit of absorption with a greenish color, abs. shadows enabled, thin glass on, opacity strength about 0.1, you get the picture:) Using those values you'd better have progressive on=) Well this still renders in a reasonable amount of time, but better not use adaptive sampling for the interior LOL. So what would your settings be like, say, if you wanted to render a babe or dude in the car, beautiful sunrise HDRI that shines through the colored glass and tints his/her face green and the hair casts beautiful shadows onto the seat or whatever? laugh

    Window glass is basically non refracting glass.

     

    So you'll want to not use the old school hack and go with transmission with thin glass enabled. To get colored glass then use a saturated transmission color and dial up transmission scale/absorption. If you want to minimize noise, you need to enable 'Use Face Forward' for the glass. I didn't enable that on my previous renders, which is probably why there's a ton of noise on the interiors.

     

     

    Tks forgot about thatyes

  • wowiewowie Posts: 2,029
    edited August 2019

    My test. 22 minutes 41.4 seconds. 8x8 pixel samples. 512 irradiance samples.

    Most likely because there's hair behind the glass. If I wanted to cheat, I'd probably use the geometry shell hack and use trace groups for the window on the driver's side. Or the easiest way, just roll down the window. laugh

    Edit:

    Here's the roll-the-window-down version. 2048 irradiance samples. 39 minutes 24.14 seconds

    22 minutes 41.4 seconds.jpg
    1067 x 600 - 395K
    39 minutes 24.14 seconds.jpg
    1067 x 600 - 386K
    Post edited by wowie on
  • Sven DullahSven Dullah Posts: 7,621
    wowie said:

    My test. 22 minutes 41.4 seconds. 8x8 pixel samples. 512 irradiance samples.

    Most likely because there's hair behind the glass. If I wanted to cheat, I'd probably use the geometry shell hack and use trace groups for the window on the driver's side. Or the easiest way, just roll down the window. laugh

    laughyes

  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    Decided to update one of my characters to work with the new SS and specular...no fireflies here. The suit is converted from IRay and has normal maps, but they seem to work ok,I used about 45% normal strength. Not sure if I overdid the bump strength on the skin, sometimes you end up looking at something you did and you just don't knowsurprise. Here's the raw render and a slightly postworked version, progressive 8x8 ps, skin SS samples 128, Irradiance samples 512.

    image

     

    image

     

    ZEANA AWE.png
    1800 x 2250 - 3M
    ZEANApp AWE.png
    1800 x 2250 - 3M
    Post edited by Sven Dullah on
  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    Progressive 8x8 ps, 17 min.

    image

    ZEANA sunset AWE.png
    1920 x 1080 - 2M
    Post edited by Sven Dullah on
  • Sven DullahSven Dullah Posts: 7,621

    Just a thought...in general, I feel that with the new build specular exposure is set too high compared to reflections...especially when using HDRI light. Even metals tend to look overexposed. Will do some renders and adjust spec. exposure per surface to see if that feels better;)

  • Mustakettu85Mustakettu85 Posts: 2,933

    Here's the raw render and a slightly postworked version...

    In the postworked version, she's definitely using a Natasha Denona highlighter =D

    Yes I did fiddle a bit with SS settings, the first version was a bit reddish, tried to find more of an olive tone

    Worked quite well IMO. Finetuned the scatter balance, right?

    Her forarms and hand palms are too glossy, I use specular strength maps here

    Have you tried switching to roughness maps? Spec strength alone won't help much with thinner parts like arms/hands/fingers.

    Another problem may be that often palm textures are too lo-res, hence bump maps are blurry from the get-go, and they are way way too weak often. While a palm IRL, it has this very specific relief.

  • Sven DullahSven Dullah Posts: 7,621

    Here's the raw render and a slightly postworked version...

    In the postworked version, she's definitely using a Natasha Denona highlighter =D

    Ok had to do a google searchblush. See what you mean=)

    Yes I did fiddle a bit with SS settings, the first version was a bit reddish, tried to find more of an olive tone

    Worked quite well IMO. Finetuned the scatter balance, right?

    Yes, and used the deep surface to blend in a bit of cyan.

    Her forarms and hand palms are too glossy, I use specular strength maps here

    Have you tried switching to roughness maps? Spec strength alone won't help much with thinner parts like arms/hands/fingers.

    Another problem may be that often palm textures are too lo-res, hence bump maps are blurry from the get-go, and they are way way too weak often. While a palm IRL, it has this very specific relief.

    I used a roughness map only on her face here. The specular maps are quite nice for once, not bumpy, if you know what I mean. I just need to darken parts of the limbs a bit. And yes also need to make some roughness maps. She could also use some nice displacement for the palm bits.

  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    Re-rendered the F 550 Sports car. Basically selected all surfaces and set specular exposure to -2. After that I could turn off opacity on the glass and use only transmission. And all the metal parts started to make sense too. Also adjusted the car body coat color and thickness + switched to GGX, and that worked fine. Non progressive 8x8 ps render, 14 min. (First version with 10x10 ps was 45 min). It's interesting to compare the renders, quite a difference=)

    image

    F 550 Sportscar 2 awe.png
    1920 x 1080 - 4M
    Post edited by Sven Dullah on
  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    Progressive 8x8 ps, 17 min.

    image

    Fixed this one by setting specular exposure for everything to -2. Even the GB brows worked without being blown out. A bit of fireflies on the reflections on the hood, maybe due to using GGX specular model for the top coat?

    image

     

    Zeana sunset2 awe.png
    1920 x 1080 - 2M
    Post edited by Sven Dullah on
  • wowiewowie Posts: 2,029
    edited August 2019

    Fixed this one by setting specular exposure for everything to -2. Even the GB brows worked without being blown out. A bit of fireflies on the reflections on the hood, maybe due to using GGX specular model for the top coat?

    What exactly do you mean with fireflies? The little bit of noise on the highlights? When I think of fireflies, they are generally very high intensity, often 5 to 10x the value of highlights. Some examples : https://www.blenderguru.com/articles/7-ways-get-rid-fireflies

    Anyway, if you don't use progressive, renders are a lot cleaner. I did managed to find the culprit for specular/reflection noise though. The next build should have much cleaner higlights, even at higher roughness.

    An exposure of -2 (2 f-stops less) means its about 25% of what you set in the settings. If you're doing this, I'm more than certain you're using too high intensity scale/exposure on your lights. Since the renders with just area lights don't have fireflies, it's likely the settings you're using with the HDRI on the environment sphere. I'm thinking of some ideas to avoid way too high intensities when you play with exposure/gain settings.

    There's some big changes to the AWE Environment Light and the environment sphere shader.

    First, there's now a separate exposure/gain/gamma/saturation setting for diffuse. So you don't have to use two environment sphere to have separate specular/camera and diffuse environment light.

    Second, while you can set it on the sphere, you can also be able to control those values and the existing Exposure/Gain/Gamma/Saturation via the AWE Environment Light. Furthermore, you can set sample and clamp value overrides for all my shaders. Using those overrides, you can just set a global scene value for all surfaces without going through the surface pane.

    Env Sphere Shader.jpg
    328 x 638 - 51K
    Env Light.jpg
    329 x 659 - 53K
    Post edited by wowie on
  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019
    wowie said:

    Fixed this one by setting specular exposure for everything to -2. Even the GB brows worked without being blown out. A bit of fireflies on the reflections on the hood, maybe due to using GGX specular model for the top coat?

    What exactly do you mean with fireflies? The little bit of noise on the highlights? When I think of fireflies, they are generally very high intensity, often 5 to 10x the value of highlights. Some examples : https://www.blenderguru.com/articles/7-ways-get-rid-fireflies

    I mean a pixel that is totally bright, but maybe those reflection are more like noise actually. It's just that I've never had this problem with previous builds.

    wowie said:

    Anyway, if you don't use progressive, renders are a lot cleaner. I did managed to find the culprit for specular/reflection noise though. The next build should have much cleaner higlights, even at higher roughness.

    Sounds like hair and skin will benefit from this, great news!

    wowie said:

    An exposure of -2 (2 f-stops less) means its about 25% of what you set in the settings. If you're doing this, I'm more than certain you're using too high intensity scale/exposure on your lights. Since the renders with just area lights don't have fireflies, it's likely the settings you're using with the HDRI on the environment sphere. I'm thinking of some ideas to avoid way too high intensities when you play with exposure/gain settings.

    This is the first render with blown out highlights. I didn't even use the aweEnvironment, only a sphere with the environment shader with defaultsettings, exposure was 0. No other lights.

    For the second render with surface exposure at -2 I was able to set the env. sphere exposure to 1 to brighten it up, without any problems as far as I can see.

    IIRC, for the girl by the car, same thing, no aweEnvironment, but I had the env. sphere exposure at 0.5. Before adjusting the surfaces even metal with no diffuse rendered totally white. The GB brows rendered white so had to turn of specular. With -2 exposure I could turn it back on. Just saying=)

     

    wowie said:

    There's some big changes to the AWE Environment Light and the environment sphere shader.

    First, there's now a separate exposure/gain/gamma/saturation setting for diffuse. So you don't have to use two environment sphere to have separate specular/camera and diffuse environment light.

    Second, while you can set it on the sphere, you can also be able to control those values and the existing Exposure/Gain/Gamma/Saturation via the AWE Environment Light. Furthermore, you can set sample and clamp value overrides for all my shaders. Using those overrides, you can just set a global scene value for all surfaces without going through the surface pane.

    Great stuff, can't waitsmiley!

    Hmm, I hear what you're saying about exposure, but I feel something is way off. I mean, look at the two renders of the girl and the car. Look at her face, the ear rings, the boots, the metal parts. With my settings I don't think that should happen, really?

    image

    image

    Btw, for the second render I raised the environment sphere exposure from 0.5 to 1, you can see the background is a tad lighter and she has a bit more light on her skin. The reflection of the sun on the hood is probably more due to changes I made (IoR, GGX, coat layer) to the car surfaces, than raising the env. sphere exposure.

    Post edited by Sven Dullah on
  • wowiewowie Posts: 2,029
    edited August 2019

    Sven, it looks like you're trying to use exposure to apply gamma correction to a linear HDRI. I think the HDRI you're using just have very high range. When you increase the exposure (trying to brighten it up, the upper values will be even higher (hence the blowout).

    What you want is actually to use gamma and exposure together. Lowering gamma will increase values in the lower ranges (what you obviously want to do). Values higher than 1 will be higher, but not as high if you use exposure.

    Here's an example:

    Original Value Exposure = 2 Gamma = 2 Gamma = 0.5 Gamma = 0.75 Gamma = 0.625 Gamma = 0.5 and Exposure = 0.5
    0 0 0 0.00 0.00 0.00 0.00
    0.1 0.4 0.01 0.32 0.18 0.24 0.45
    0.2 0.8 0.04 0.45 0.30 0.37 0.63
    0.3 1.2 0.09 0.55 0.41 0.47 0.77
    0.4 1.6 0.16 0.63 0.50 0.56 0.89
    0.5 2 0.25 0.71 0.59 0.65 1.00
    0.6 2.4 0.36 0.77 0.68 0.73 1.10
    0.7 2.8 0.49 0.84 0.77 0.80 1.18
    0.8 3.2 0.64 0.89 0.85 0.87 1.26
    0.9 3.6 0.81 0.95 0.92 0.94 1.34
    1 4 1.00 1.00 1.00 1.00 1.41
    1.1 4.4 1.21 1.05 1.07 1.06 1.48
    1.2 4.8 1.44 1.10 1.15 1.12 1.55
    1.3 5.2 1.69 1.14 1.22 1.18 1.61
    1.4 5.6 1.96 1.18 1.29 1.23 1.67
    1.5 6 2.25 1.22 1.36 1.29 1.73
    1.6 6.4 2.56 1.26 1.42 1.34 1.79
    1.7 6.8 2.89 1.30 1.49 1.39 1.84
    1.8 7.2 3.24 1.34 1.55 1.44 1.90
    1.9 7.6 3.61 1.38 1.62 1.49 1.95
    2 8 4 1.41 1.68 1.54 2.00

     

    Obviously, when you use exposure you'll get blowouts since a value that was 2 before is now 8. But if you use a gamma of 0.5 (close to 1/2.2 which is used to linearize sRGB gamma), then combine it with an exposure of 0.5, you'll get to brighten the lower values while keeping the high values relatively the same. At the very least, you won't push intensities to 4x the original value. In this example, If the sun value were already high, say 16 EV, you'll be increasing that to 18 EV.

    If you look at the second render with the girl, you'll see there's tons of noise on the collar/shoulder part of jacket. If you're trying to get more diffuse lighting, don't change the exposure on the sphere (that affects both reflection/diffuse). You'l want to use the exposure on the environment light (since it has protection for over exposure) or GI Exposure on the material. Obviosly, the actual way around this in real life is just to add lights to the scene.

    Anyway, the new AWE Environment Light and environment sphere shader will try to compensate for this and offer the ability to change exposure for specular/reflection and diffuse separately.

    Post edited by wowie on
  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    Tks for the input!

    Let's start with the fact I've only downloaded HDRIs with a very high dynamic range since I started using them, typically 16EV or more. I've used them with awe for almost a year now, never before have I had such blowouts. And no I don't want to boost the lower values, on the contrary. Because typically when using HDRIs with awe I get washed out renders (backgrounds). I want to keep the contrast or increase it by increasing the gamma value, typically from the default 1 to 1.5. And 99% of the renders come out with a too dark background, so I typically raise the env. sphere exposure to 0.5 or sometimes 1. That has thus far worked well and given me renders that don't need much adjusting in postwork. But that is with the environment light with a max light intensity (5). Now, because I wanted to test how the new build works with HDRI, for some reason I just added Kettus env. sphere and radiumcatcher without the environment light. And in the first example (the muddy road) everything was at default settings. And everything is too dark BUT blown out. I'm not really getting it. I kind of understand what your'e saying, but one would expect to get maybe a dark but definitely not a dark and blown out render with exposure 0 and gamma 1, right?

    Regarding the second render with the girl and car, maybe a setting of 0.5 exposure was enough to go over the cliff but I doubt it. But I've certainly been wrong before, so I'll concider what you said, and test a few more HDRIs to see how things are working;) And btw that's not noise on her collars, not on the second render. Those areas use metalness and normal maps. I can only see noise on the first version, quite in line with everything shiny being overexposed. The second render is fine as far as I can see. Have a look at the large size portrait render to see the details:) Or am I missing something? Hmm maybe it's noise after all:)

    If the new build is a new deal, I accept that, no problem. And if I can't get the background to look right without having to boost exposure leading to artefacts, I'll just use one enviroment for visibility and one for lighting, until you release the updated environment light.

    Post edited by Sven Dullah on
  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    ...to further try to explain what I want to achieve with HDRI backgrounds, I simply loaded the aweEnvironment and put an HDRI on the sphere. Everything at default. This is how it looks:

    image

    Now, I take it into my image editor and simply click "autolevels". The result:

    image

    This is how it should look.

    With this HDRI, if I set exposure to 1.2 and gamma to 1.2 I get this:

    image

    It may be the wrong thing to do, I don't know, but it's certainly closer to what I want, it also affects everything in the scene and the render needs much less adjusting in postwork. (And with a bit of fiddling it's possible to get pretty damned closed to the autolevels version)

    Autolevels are what they are, not always accurate or even what you want, but they are certainly an indication of something. Pretty much every HDRI I've used behave in the same way, I find it hard to believe they are all bad. It would be great if the new aweEnvironment could handle this;)

    default environment.png
    1280 x 720 - 1M
    default environment autolevels.png
    1280 x 720 - 2M
    default environment exp 1.2 gamma 1.2.png
    1280 x 720 - 1M
    Post edited by Sven Dullah on
  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    Aaaaarrrrrgggghhhh!

    Ok now I'm really confused!laugh I opened the girl and car scene to do some testing. First thing I did was to set the surfaces exposure back from -2 to 0. Then I set the environment exposure and gamma back to default settings (0.5/1.5 to 0/1). Testrendered, no blowouts. Reset environment settings to 0.5/1.5, still no blowouts. Even the GB brows rendered fine. So I thought what the hell and set env. sphere exposure to 2 and rendered. No blowouts. Here's a quick 8x8 progressive render with exposure 2 and gamma 2:

    image

    Sure it's overexposed and the sunreflections on the hood are noisy, but there are no completely white areas, her face has no strange highlights, her boots are ok, the brows are dark. All metal parts are ok. What the hell? All the problems were just a glitch?

    EXP2(GAMMA2.png
    1280 x 720 - 1M
    Post edited by Sven Dullah on
  • Sven DullahSven Dullah Posts: 7,621
    edited August 2019

    Raised the environment sphere exposure to 3 and gamma to 1. Turned everything to a snowlandscape but no blowoutslaugh And the noise is gonesurprise.

     

    image

    I give up, can't reproduce the problems...

    exp3 gamma1.png
    1280 x 720 - 647K
    Post edited by Sven Dullah on
Sign In or Register to comment.