I don't get it. I killed all the Area Lights, and rendered with GI to let the lum surface panels of the lights do their thing. It was too dark, so I increased the lum from 300% to an extreme of 1000%, just to see a difference, and no change in the render at all. Am I really that long forgotten since reading my lighting books...
Am I missing something, or am I indeed just stupid, LOL.
EDIT: I think I may have some surfaces named wrong...
EDIT EDIT: Never-mind, I forgot I had a map set for lum to darken the edges of the lights. It was set to Normal blending mode thus modulating the lum amount as is.
Haha, yeah with LCS you find that you need to make luminous surfaces even brighter, in a weird way. Most of my luminous panels are over 700% bright. If you want to modulate it just use the textures in the node editor and use a maths>scalar>multiply and set it to say 7 or 8, and yeah the video covers that.
I'm working on making it better, stronger, faster (can you hear the 6MDM sound effects?) right now, per the guys' tips.
So, I killed all the Area Lights and am working just with lum surfaces and RAD. way faster already, but that's only with low setting GI tests. We'll see what the final refined higher settings bring. Either way it'll be a ga-gillion times faster than 17 hours!
tobain, what I am reading up on is the insights you guys were giving ST-One in her space-station thread over at F3D last January. I get it all in concept, just kinda fuzzy on practical application still in regard to 3D apps.
So you have to darken all your colors, and up all your light to basically wreck the image to then apply the 2.2 gamma to make it look all pretty and normal and better then.
Not clear on those expressions Dave shared, or the how-to's as to the darkening of color. I am on an Intel Mac so Input Spy seems a no-go for me in your one node example. I'm rusty on my node use too. I know how to use them in LW, just fell out out of practice, as my E was pre-node LW and I didn't really re-node all the surfaces.
That was my first read through, so it was a lot to ingest and make sense of in regard to application in 3D apps. I get the whole gamma correction stuff as I was a pre-press digital color analyst for years, developing color profiles and all that crap for years back in the dawn and infancy of digital color management. It's just recalling that info back up, and learning how to apply it in 3D, per the apps.
You can still use the DB&W simple colour corrector to correct the colours/colour textures and feed the output to colour.
You have it more or less. but the one thing to remember is that Lights and Luminous polygons are not equivalent: With an area light, how big you make it will affect the softness of the shadow, but little else. With a luminous poly, the larger it is, the more energy it is putting into the system. You are also never going to get hard shadows from the current radiosity engine, and luminous polygons (well duh, if they are huge hehe) and NEVER make teeny tiny super bright lights, as that'll almost always never work. If you must have small source lights, combine a 'filament' luminous polygon, with a real light, such as an IES light. Large areas of luminous polys is the only practical way to go, for radiosity use!
Also remember with luminous polygons, any values don't really have any real-world meanings: What is '400%' in terms of candelas or watts? You often need to go by eye to get good results, but I recommend you keep radiosity set to 100%, to make the settings more balanced, with other surface settings. So yeah you sometimes have to go up to 1000% - That said try not going over 1000 as it can cause blotchiness issues, in my experience.
So it is! they keep messing on with the link format and structure, I notice, annoying, when you have any hard-links to anything Fixed! and yes that avatar amuses me too, Samurai made it for me, as well as abusing my signature, along with half the staff! Not much new in there, as I haven't done much worth posting recently, lots of laying out of deck plans, and routing of corridors, which are still somewhat featureless, so a bit boring
That's all covered in the video and the blurs in that image are done with the DP reflections node set to 20%. Otherwise I have the exact same settings as is shown in the vid. Sorry this sounds totally like I'm pimping my article LOL The only settings which changed in that image was slightly tone down the occlusion shading on the white material, so it's less strong, and swap out regular reflection blur shading for DP ones.
Also, I'm not quite clear, he (Matt Gorner) says LW renders as linear, and that his first image has no Gamma applied (thus it's all dark). But where or how the the Gamma not get applied?
I mean, as I was doing it, I render something and it's not that dark (looks OK decent), so is Gamma being applied somewhere to lighten it at some point?
OK, then say we go into stripping Gamma from colors and images, sure dark now, but how do you know how much to light it to compenstate to arrive at the end Gamma making it look good? Trail and error tests, I suspect...?
And the lights, he says in the vid, that you need less light (which makes sense), and yet you were speaking about having to crank your lum surfaces up to work...
Is that just to compensate for them not being actual lights then, I assume/figure...?
That's all covered in the video and the blurs in that image are done with the DP reflections node set to 20%. Otherwise I have the exact same settings as is shown in the vid. Sorry this sounds totally like I'm pimping my article LOL The only settings which changed in that image was slightly tone down the occlusion shading on the white material, so it's less strong, and swap out regular reflection blur shading for DP ones.
Hey NP prob on pimpin' your vid. I can't wait for it to get here already! Easier for you too, as you got all the info we're chattin' about down in there, so no need to repeat yourself here. I get that, bud.
I'll hold off any more Qs until I actually see the vid. I'm sure the viewing will set me straight. I'm not too far off now. I'm a decent quick study. Esp if I have a vid I can just keep rewinding until it sinks in, LOL
And the lights, he says in the vid, that you need less light (which makes sense), and yet you were speaking about having to crank your lum surfaces up to work...
tehbian says random stuff sometimes, you should need less light, the way things blend whites overpower blacks more when you use linear gamma, so you'll find light is brighter automatically
Actually no, it's to do with the balance between luminous surfaces and Fresnel based reflections, as well as the light they are emitting too. Hence why you do need some more powerful luminous objects to show up in reflections properly
linear gamma reflections are brighter so you shouldn't need to make surfaces brighter to reflect in them, fresnel is just fresnel and unrelated to gamma
Not in my experience. In my chrome sets i had to nearly double the luminosity of the glowing panels, to make the renders look correct. Otherwise I got horrible washed out looking grey reflections. FEIL. Just my experience, yours may differ.
I'm thinkin', gettin' down this whole LCS per the 3D process, like most everything in 3D, requires lots of test renders to get the lay of the linear land, so to speak.
Just these RAD test renders, take longer by default! Grrrrrr... Yet, it is what it is, eh, it is what it is. Bright side: Still way-beats 17 hours.
I've been using a proper linear workflow in 3D and compositing for a little over a year. It's the correct way to do things. I'd never go back to the old way. What's funny, is that if you don't know about linear workflow, you don't even know you're doing it wrong. I did it wrong for years before I learned the right way to do things.
I suspect he means 'proper' in the sense he's doing it correctly, as there's a lot of duff info out there, with regards to LCS, which is highly inaccurate., because a lot of people get half of it, or don't get why you are doing what you are doing
Here is a linear workflow tutorial I did for AeTuts back in Dec. I don't know if it will help you with Photoshop, but it might shed some kind of light on the situation. Enjoy.
This is starting to get into the 'not wanting to repeat myself' land
The information in that tutorial is good information, including how to convert colours and textures in LW. The only thing he does wrong there is suggest you use the 'image editor' to transform textures. The problem with that is the image editor does not operate in floating point colour depth, whereas the Node editor does. So it's best to avoid doing colour conversions anywhere in LW except the Node editor. (edit, don't forget the render buffers are internally full floating point, so any colour transforms done in there are with 32bpp precision)
If you create a Gamma 1.0 image within LW, then you should save it out as an EXR image, as this is natively a Float format. (and should be Gamma 1.0) If you then open this in AE/PS they will (by default) apply a display gamma for you (unless you set one of the custom tonemapping options as he demonstrates). AE is especially strong in the area of tonemapping, but it's quite hard to get your head round!
If you apply a gamma 2.2 to the LW image by using one of the several image and or pixel filters available, (FpGamma being the native one) then the image will display correctly within the LW viewport and if you chose to save the image out as a LDR (8/16bpp image) then it will be fully correct at this stage. If you want to continue working in 32bpp/float mode, then you should either not apply the Gamma adjustment in LW, or apply it within PS/AE using a exposure>gamma 0.4545 layer adjustment to it. That's my favourite method, in this case, because If you use the Simple Colour Corrector pixel filter, then the render happens in Gamma colour space, so it looks 'right' as it's rendering, and it also means that adaptive sampling AA works correctly. If you want a huge technical explanation for that, feel free to ask, but it's not easy to understand
I believe that LWHC (the version of classic LW which ships as part of the CORE suite) has native colour workflow tools, but since I don't have that yet (someday, when I can afford it!) takes some of the hassle out of it for you
LW is quite capable of working in a Linear Gamma workflow, it's just there's a number of gotchas' and caveats you need to get before you can work in it, and again, sorry about the broken record bit, but I do cover a lot of those in the vid, but I've typed in most of them in here now LOL
If you want to share more feel free about the "adaptive sampling AA works correctly," eh. If not, not prob. You should have a file, where you can just copy and paste all this info into threads that you find the user wanting to know about LCS. Save your fingers a work-out.
CORE/LWHC (not even sure what that HC stands for), this has shipped? I thought CORE was still under development. Or are you talkin' about the beta, that's out, correct? I'm bad at keepin' up on this stuff. I was just confused when CORE was announced, all that subscription pricing stuff, and never paid enough attention after that, hoping it would get simpler come the time of actual shipping.
Haha, well the thing is a) I did that video for 3D world which pretty much definitively shows you most of my work-flow for how I light and shade my interiors and b) I haven't exhaustively tested it all out, so I am by no means an expert, I've just went through a lot of doing-it-for-myself pain and many many test renders !
And since you asked. The thing with the adaptive sampling is that the threshold value assumes a difference in brightness between two neighbouring pixels based on a tolerance value. If there's a difference, based on the tolerance value, then LW will render that pixel again, with a different stochastic sampling pattern, and recombine it with the original pixel, to, hopefully arrive at a more smooth interpolation, and smoothed anti-aliasing. The more samples you take of a given pixel, the more likely you will remove 'grain' artefacts, and smooth the edges of geometry and textures, but if you have to sample each pixel several hundred times, then you are going to have horrible render times, especially when things like reflection blur, refraction blur, SSS, soft shadows from area lights, or any complex ray tracing. This turns anti-aliasing from a brute-force method, to an intelligent one.
Lightwave renderer is not gamma-aware, so it sees a 1.0 gamma render is exactly the same as a 2.2 render. It just processes your geometry and lighting. However once you then apply your 2.2 display gamma, it will become obvious that it didn't do extra AA in the 'dark. regions of the render: If you look at the way an image looks if you apply a inverse gamma adjustment to it, (how a 1.0 gamma image looks to the LW renderer frame buffer) the darks become all 'smushed' together (genuine technical term, honest! ) and your mid point drops visibly. It's not exactly the same, but if you look at Photography, a typical measure to help the image have a balanced mid-point is to use a Kodak 18% grey card. This card is roughly analogous to 50% 'brightness' in the image, and it's (almost) the same with a 2.2 Gamma response curve (technically most displays are calibrated using sRGB which is a modified 2.4 gamma response curve, so 18% luminance doesn't correctly convert to 50% luminance, but it's close) This means the luminance scale at the 'dark' half of the image are way more compressed together, unless that value is subsequently gamma modified with the 2.2 response curve. Because LW isn't gamma aware, the closer to 0, or black, a pixel is, the less likely it is going to be able to detect if it's different from it's neighbour.
If you lower the threshold for the difference between neighbours, it will hit mid-light pixels even more, so it will re-sample more of those than it really should be doing (especially given, in the inverse sense, those will be pushed closer together, after gamma correction, so their difference will be visibly smaller, and so you need less AA not more). and you have a rule of diminishing returns with a Gamma curve, so that the closer to you get to 0, the smaller the difference will get in terms of relative luminance. So you have the choice of keeping on lowering the value to get good AA on dark pixels, to the point that you may as well not use adaptive sampling, and just use high AA. because adaptive sampling is running the whole pixel through another full AA pass with each sub-pass it also means much more of the image is being sampled multiple times, completelly redundantly, turning adaptive sampling from a clever system into an exceptionally dumb one!
So it's better if you can gamma modify the frame buffer in-camera (the simple colour corrector pixel filter does this), so the AA is applied to the 'visible' spread of the luminance, in the 2.2 gamma 'display' range. The most optimal use of AA in LW's advanced cameras is to do minimal-possible AA, with a low adaptive sampling on 'edge' pixels, to make the image properly anti-aliased, but LCS renderings can unfortunately massively negatively impact on those renders, and actually result in increased render times, not smaller ones!
Thanks, dude! Combined with my previous knowledge of RGB and LAB color space and color management, and my study of rendering and AA in modo, this all makes perfect sense, eh. Thanks for sharin' all that, eh.
Posts
Am I missing something, or am I indeed just stupid, LOL.
EDIT: I think I may have some surfaces named wrong...
EDIT EDIT: Never-mind, I forgot I had a map set for lum to darken the edges of the lights. It was set to Normal blending mode thus modulating the lum amount as is.
deg
I'm working on making it better, stronger, faster (can you hear the 6MDM sound effects?) right now, per the guys' tips.
So, I killed all the Area Lights and am working just with lum surfaces and RAD. way faster already, but that's only with low setting GI tests. We'll see what the final refined higher settings bring. Either way it'll be a ga-gillion times faster than 17 hours!
tobain, what I am reading up on is the insights you guys were giving ST-One in her space-station thread over at F3D last January. I get it all in concept, just kinda fuzzy on practical application still in regard to 3D apps.
So you have to darken all your colors, and up all your light to basically wreck the image to then apply the 2.2 gamma to make it look all pretty and normal and better then.
Not clear on those expressions Dave shared, or the how-to's as to the darkening of color. I am on an Intel Mac so Input Spy seems a no-go for me in your one node example. I'm rusty on my node use too. I know how to use them in LW, just fell out out of practice, as my E was pre-node LW and I didn't really re-node all the surfaces.
That was my first read through, so it was a lot to ingest and make sense of in regard to application in 3D apps. I get the whole gamma correction stuff as I was a pre-press digital color analyst for years, developing color profiles and all that crap for years back in the dawn and infancy of digital color management. It's just recalling that info back up, and learning how to apply it in 3D, per the apps.
I'll figure out something. I'll get there, eh.
Thanks, bud.
deg
And thank you,kind sir, for that heads-up, eh. < bows >
deg
You have it more or less. but the one thing to remember is that Lights and Luminous polygons are not equivalent: With an area light, how big you make it will affect the softness of the shadow, but little else. With a luminous poly, the larger it is, the more energy it is putting into the system. You are also never going to get hard shadows from the current radiosity engine, and luminous polygons (well duh, if they are huge hehe) and NEVER make teeny tiny super bright lights, as that'll almost always never work. If you must have small source lights, combine a 'filament' luminous polygon, with a real light, such as an IES light. Large areas of luminous polys is the only practical way to go, for radiosity use!
Also remember with luminous polygons, any values don't really have any real-world meanings: What is '400%' in terms of candelas or watts? You often need to go by eye to get good results, but I recommend you keep radiosity set to 100%, to make the settings more balanced, with other surface settings. So yeah you sometimes have to go up to 1000% - That said try not going over 1000 as it can cause blotchiness issues, in my experience.
BTW, your Pluto Station link, she be broke.
Thanks again, bud.
BTW, your avie makes me chortle esp. as I met Mr. T a few times, as he's from Chicago. Nice guy.
deg
How are you blurring your reflections? Is that a micro-bump, or something else?
If a micro-bump, what procedure and setting (size, etc.) do you use for it?
deg
Tanks,
deg
I mean, as I was doing it, I render something and it's not that dark (looks OK decent), so is Gamma being applied somewhere to lighten it at some point?
OK, then say we go into stripping Gamma from colors and images, sure dark now, but how do you know how much to light it to compenstate to arrive at the end Gamma making it look good? Trail and error tests, I suspect...?
And the lights, he says in the vid, that you need less light (which makes sense), and yet you were speaking about having to crank your lum surfaces up to work...
Is that just to compensate for them not being actual lights then, I assume/figure...?
Thanks, mate!
deg
Hey NP prob on pimpin' your vid. I can't wait for it to get here already! Easier for you too, as you got all the info we're chattin' about down in there, so no need to repeat yourself here. I get that, bud.
I'll hold off any more Qs until I actually see the vid. I'm sure the viewing will set me straight. I'm not too far off now. I'm a decent quick study. Esp if I have a vid I can just keep rewinding until it sinks in, LOL
deg
linear gamma reflections are brighter so you shouldn't need to make surfaces brighter to reflect in them, fresnel is just fresnel and unrelated to gamma
Just these RAD test renders, take longer by default! Grrrrrr... Yet, it is what it is, eh, it is what it is. Bright side: Still way-beats 17 hours.
Thanks, gents!
deg
I hope to get it down and understood shortly. Doing lots of reading now.
Where did you learn it, BTW?
And what do you mean by "proper?"
deg
Oh and while I remember, here's a cool video on the NT forum on the basics of why Linear is working the way it is. http://www.newtek.com/forums/showthread.php?t=102397
Yeah, I have watched that vid a coupla times. I am still unclear on a coupla points as to the actual workflow though.
deg
In what file format? OpenEXR?
And then you apply 2.2 Gamma to it how? In Photoshop? Using what, an Exposure/Gamma adjustment set to 2.2?
deg
http://ae.tutsplus.com/tutorials/workflow/understanding-linear-workflow/
It's all Adobe, pretty easy to interpolate back and forth. I could even just use AE for still work too, come down to it. Bookmarked.
You guys ever see this gent tut(s)? If so, thoughts? Proper of duff?
Color Management and Wide Gamut displays – in a CG Workflow
I am just getting to his Photoshop and Tone Mapping part at the bottom.
deg
The information in that tutorial is good information, including how to convert colours and textures in LW. The only thing he does wrong there is suggest you use the 'image editor' to transform textures. The problem with that is the image editor does not operate in floating point colour depth, whereas the Node editor does. So it's best to avoid doing colour conversions anywhere in LW except the Node editor. (edit, don't forget the render buffers are internally full floating point, so any colour transforms done in there are with 32bpp precision)
If you create a Gamma 1.0 image within LW, then you should save it out as an EXR image, as this is natively a Float format. (and should be Gamma 1.0) If you then open this in AE/PS they will (by default) apply a display gamma for you (unless you set one of the custom tonemapping options as he demonstrates). AE is especially strong in the area of tonemapping, but it's quite hard to get your head round!
If you apply a gamma 2.2 to the LW image by using one of the several image and or pixel filters available, (FpGamma being the native one) then the image will display correctly within the LW viewport and if you chose to save the image out as a LDR (8/16bpp image) then it will be fully correct at this stage. If you want to continue working in 32bpp/float mode, then you should either not apply the Gamma adjustment in LW, or apply it within PS/AE using a exposure>gamma 0.4545 layer adjustment to it. That's my favourite method, in this case, because If you use the Simple Colour Corrector pixel filter, then the render happens in Gamma colour space, so it looks 'right' as it's rendering, and it also means that adaptive sampling AA works correctly. If you want a huge technical explanation for that, feel free to ask, but it's not easy to understand
I believe that LWHC (the version of classic LW which ships as part of the CORE suite) has native colour workflow tools, but since I don't have that yet (someday, when I can afford it!) takes some of the hassle out of it for you
LW is quite capable of working in a Linear Gamma workflow, it's just there's a number of gotchas' and caveats you need to get before you can work in it, and again, sorry about the broken record bit, but I do cover a lot of those in the vid, but I've typed in most of them in here now LOL
If you want to share more feel free about the "adaptive sampling AA works correctly," eh. If not, not prob. You should have a file, where you can just copy and paste all this info into threads that you find the user wanting to know about LCS. Save your fingers a work-out.
CORE/LWHC (not even sure what that HC stands for), this has shipped? I thought CORE was still under development. Or are you talkin' about the beta, that's out, correct? I'm bad at keepin' up on this stuff. I was just confused when CORE was announced, all that subscription pricing stuff, and never paid enough attention after that, hoping it would get simpler come the time of actual shipping.
deg
And since you asked. The thing with the adaptive sampling is that the threshold value assumes a difference in brightness between two neighbouring pixels based on a tolerance value. If there's a difference, based on the tolerance value, then LW will render that pixel again, with a different stochastic sampling pattern, and recombine it with the original pixel, to, hopefully arrive at a more smooth interpolation, and smoothed anti-aliasing. The more samples you take of a given pixel, the more likely you will remove 'grain' artefacts, and smooth the edges of geometry and textures, but if you have to sample each pixel several hundred times, then you are going to have horrible render times, especially when things like reflection blur, refraction blur, SSS, soft shadows from area lights, or any complex ray tracing. This turns anti-aliasing from a brute-force method, to an intelligent one.
Lightwave renderer is not gamma-aware, so it sees a 1.0 gamma render is exactly the same as a 2.2 render. It just processes your geometry and lighting. However once you then apply your 2.2 display gamma, it will become obvious that it didn't do extra AA in the 'dark. regions of the render: If you look at the way an image looks if you apply a inverse gamma adjustment to it, (how a 1.0 gamma image looks to the LW renderer frame buffer) the darks become all 'smushed' together (genuine technical term, honest!
If you lower the threshold for the difference between neighbours, it will hit mid-light pixels even more, so it will re-sample more of those than it really should be doing (especially given, in the inverse sense, those will be pushed closer together, after gamma correction, so their difference will be visibly smaller, and so you need less AA not more). and you have a rule of diminishing returns with a Gamma curve, so that the closer to you get to 0, the smaller the difference will get in terms of relative luminance. So you have the choice of keeping on lowering the value to get good AA on dark pixels, to the point that you may as well not use adaptive sampling, and just use high AA. because adaptive sampling is running the whole pixel through another full AA pass with each sub-pass it also means much more of the image is being sampled multiple times, completelly redundantly, turning adaptive sampling from a clever system into an exceptionally dumb one!
So it's better if you can gamma modify the frame buffer in-camera (the simple colour corrector pixel filter does this), so the AA is applied to the 'visible' spread of the luminance, in the 2.2 gamma 'display' range. The most optimal use of AA in LW's advanced cameras is to do minimal-possible AA, with a low adaptive sampling on 'edge' pixels, to make the image properly anti-aliased, but LCS renderings can unfortunately massively negatively impact on those renders, and actually result in increased render times, not smaller ones!
deg