3D Kitchen interior - reachin' for the stars here

Become a member of the CGSociety

Connect, Share, and Learn with our Large Growing CG Art Community. It's Free!

THREAD CLOSED
 
Thread Tools Search this Thread Display Modes
  06 June 2012
Back to the drawing board with lighting. With all lights off and only the HDRI's Final Gather emitting, I get a 20 minute render (finally!) and predictable results:



Too dark, but much better than the previous FG/skylight rendering which was obnoxiously too bright. And a 20 minute render is also much more predictable and useful, but that obviously means the can lights and portal lights are causing the rendering overhead too, in previous renders.

*Edit: Note that I forgot to set the can light's bulb surface to hidden as well, which means they were also emitting FG light, if only a little. I'll set their geometry on the CanLights layer for future elimination in diagnosis.

With FG and the Can Lights on, no photon GI or caustics I get a 1h28m render:



At this point I modeled the flooring, as the texture on a plane method alone isn't "cutting the mustard", as they say. With the flooring and backsplash in place all that's left will be a few props, such as a flower vase and some cutlery and dishes, on the modeling end of things.

With the can lights and the skylight portals on, FG only still, I get a 1h15m render:

*Edit: The can lights geometry was on, but the lights themselves (disc area lights) were not, here.



I'll do one today to test photon GI and caustics again, but is this latest one any better than previous iterations? Any progress?
__________________
Commodore 64 @ 1MHz
64KB RAM
1541 Floppy Drive


"Like stone we battle the wind... Beat down and strangle the rains..."

Last edited by InfernalDarkness : 06 June 2012 at 05:53 PM.
 
  06 June 2012
Quick udpate, no interior lights, Final Gather and photon GI/caustics. 47 minutes rendertime (progress!):



It's still way too dark again, and the blotchies are no fun. Next I'll tweak the photon settings themselves and try to control the brightness there, instead of using higher multipliers on the portal lights and physicalSky which caused the FG blowout earlier. GI+FG is still darker than just FG, so tuning the photon exposure should help with this.

Ideas and thoughts? Beating a dead horse here?
__________________
Commodore 64 @ 1MHz
64KB RAM
1541 Floppy Drive


"Like stone we battle the wind... Beat down and strangle the rains..."
 
  06 June 2012
It seems to me you won't even need gi and photons in this image as it is, what are you hoping to achieve on adding them? (Remember that combining gi and fg alters trace depth.)

To me the bump on the floor seems a bit exagerrated (at least comparing to the sort of floors most common where I live). (Also compare your own posted reference image.)

This search gives some reference images to compare and think about: http://www.bing.com/images/search?q...FORM=BIFD#x38y0 . If you look at them, which ones do you prefer? Why?

If you want to improve the image I think your time is better spent adding some plants and kitchen accessories as previously suggested. Also look at the reference image you have posted yourself, it has plants, the knife set, the plates, things in the cupboards.

edit: At least where I live modern kitchens are generally lighter in color, and I'd be surprised to see that somewhat dark and dull greenish color on the wallpaper. I realise that this might be very different in other countries of course.

Last edited by tobbew : 06 June 2012 at 10:59 PM.
 
  06 June 2012
Hmm, I haven't posted any reference images myself. I do have the photos from the as-built job but we didn't end up doing the remodel, thus this project was safe for me to use as a personal project. The photos could help vaguely with the lighting though, but since everything is different in my design, they are not an exact reference.

I do agree the bump on the floor needs to be much lower, maybe 1/4 of what it's at now!

The goal here is to approach realism with lighting and shaders, while unifying my workflow so that in future scenes I don't have so much trouble, and can cut down on test-renders. With photon GI and caustics, you can achieve a much more vivid spread of color and indirect bounce lighting. Also, Final Gather isn't a final solution really in mental ray, although in some scenes you can get away with using just FG. Caustics play a part in everything we see (although they aren't actually caustics, they're just called that for some reason!) in reality.

As you can see, my lighting still leaves much to be desired. FG alone won't cut it, nor will just photon GI. Another important factor is that many of my interior scenes have glass lighting fixtures and panels, cupboard doors, glass shower doors, etc. To get most glass to look accurate, some form of caustics is usually in order, especially when it comes to glass lighting fixtures.

While I will add some props at some point, getting the shaders and lighting to work is a much more intensive project so far.

Quote: (Remember that combining gi and fg alters trace depth.)


I'm not sure what you mean by this though? If the photon GI traces are set to 6/6/12 and the FG traces are set to 2/2/4, this is still independent from the main raytracing limits as far as I'm aware. In what way is trace depth altered?
__________________
Commodore 64 @ 1MHz
64KB RAM
1541 Floppy Drive


"Like stone we battle the wind... Beat down and strangle the rains..."
 
  06 June 2012
I appologize. I skimmed throught the thread quickly fast. I was referring to the one posted by Kev. Sorry for that.

If I remember things right when you are combining FG and GI it only uses one step of FG and then uses the GI for the rest. (You might be able to prove me wrong on this one as well, this is as I recall having read things previously. I was considering trying to look up a source before writing that, but I'm still too tired to try to find one. It might have been a suggestion as for how to combine them, and it might be dated.) I was referring to the trace depth on the FG, not on the general ray tracing settings.

I'd agree on using caustics when needed, but I don't expect you'll need GI in this image. At least not currently. If you are mainly looking at lighting and shaders, are you using, or have you considered trying, the wombat_archlight shader? Orusing ies profiles in the normal mr shader (if you aren't already, but if so I'd consider changing to more interesting ies profiles)?
 
  06 June 2012
I guess my confusion in your assertion that FG will be enough is that it doesn't look good yet. Do you feel that this last image with FG-only is realistic, or perhaps photorealistic, from a lighting standpoint?



I've never read that the photon GI solution cancels out subsequent FG depths before, but that would still put us at 1/1/2 for Final Gather. Perhaps this is why it seems lighter in the FG-only images though? I will certainly test this and get back to you! Should be a pretty simple test.

Hope I didn't seem too argumentative here, just not sure why I wouldn't want to use photon GI in conjunction with FG when:

1. It renders faster
2. More vivid color and control (hopefully!)
3. FG alone isn't photorealistic here
__________________
Commodore 64 @ 1MHz
64KB RAM
1541 Floppy Drive


"Like stone we battle the wind... Beat down and strangle the rains..."
 
  06 June 2012
FG should be enough to give you photorealistic lighting, but perhaps what's throwing you off is the shaders you are using now. Have you tried setting up lighting with just grey shading on everything? (except the windows of course)
Might help make it easier to see what's doing what, and especially help to see if your FG settings are high enough. It's hard to tell with the last render, but something about it just looks off particularly in the white panels on the bench in the middle, and I can't tell if it's the FG interpolations set too high or it's the shader, but I would expect to see more "detail" in the indirect bounces. I would not put the interpolations above 5.
 
  06 June 2012
I agree with your thoughts on the shaders, thanks for the input! Looking harder at the cabinetry, it really doesn't seem reflective enough. We should see some details of the opposing wall in the island's cabinetry on the right there, and more detail. Should see SOMETHING besides just white and some edges! Gonna give it a whirl in the morning.

Also, I think the doorway that this last angle is shooting from needs a Portal light as well perhaps. Not as powerful as the window Portals, but there would be some light coming through that doorway as it connects to the rest of the house, and I'm not simulating that currently.

My FG settings are:
Quality: 100
Point Density: 1
Interpolation: 25 (in an attempt to shave off some rendertime)
Filter: 0

I'll also try lowering the interpolation to 5 and see if that helps. Thanks for the tip, Pap87!


*Edit: Last night's render, Portal lights and interior Can lights, GI, caustics, and Final Gather. Rendertime 1h23m:

__________________
Commodore 64 @ 1MHz
64KB RAM
1541 Floppy Drive


"Like stone we battle the wind... Beat down and strangle the rains..."

Last edited by InfernalDarkness : 06 June 2012 at 05:12 PM.
 
  06 June 2012
I tried in a simple test scene and comparing fg only 5/5/8 and then adding gi 5/5/5 and trying with fg 5/5/8 and 5/5/1 does not make any visible difference on my render. (The scene did not have a lot of refractive surfaces so I probably would not have seen any difference between 1/1/1 and 1/1/2.)

edit: Just adding here instead of adding more posts to the thread which would be unneccessary:
It did not have a lot of refractive surfaces, but reflective with a depth higher than five. I do see a difference using fg at the different settings without gi, not with. I don't claim it's scientific (not sure how you mean that should be performed? but I guess at least using a difference filter instead of eyeballing). Feel free to let me know your conclusion if you'd decide to do a test in the future.

Last edited by tobbew : 06 June 2012 at 08:53 PM.
 
  06 June 2012
@Tobbew: And were your results different, Tobbew? That doesn't seem a terribly scientific test scenario. If you don't have depths of 5 in your geometry/shaders, then it wouldn't work?

@Pap87: I did play with the shaders, some tweaks and optimizations, and the lighting a little too. I tried lowering the interpolation to 5 as you suggested, as most of my previous renders were at 25, and here's the results:


(click pic for full size)

Interpolation at 50 works much nicer in this case. In a final render you generally want the point density at 1, but for the sake of speed I'm at 50% for now.

One more full render today and we'll see how it's looking with the adjusted shaders and lighting!
__________________
Commodore 64 @ 1MHz
64KB RAM
1541 Floppy Drive


"Like stone we battle the wind... Beat down and strangle the rains..."
 
  06 June 2012
Is this one any better? I'm trying to incorporate all your suggestions...


(click pic for full-size)

Rendered in 121 minutes on 5 of 6 cores, FG and photon GI+caustics. I don't like the darkness below the central wall cabinet much. Seems like something's wrong over there, still.

Next I'm going to add props like everyone suggested. Some plants, a bonsai tree in a small dish on the countertop, and some china in the cupboards with the glass fronts too. Maybe a vase with some flowers on the island?
__________________
Commodore 64 @ 1MHz
64KB RAM
1541 Floppy Drive


"Like stone we battle the wind... Beat down and strangle the rains..."
 
  07 July 2012
Suggestions Warning, long post ahead XD

Hi, first let me tell you I have been following your advices in mental ray for a long time now in CGtalk, you are one of the artist that more support gives when speaking of Mental Ray. For that I’m really indebted to you.

There is definitely gap about realism in Mental Ray when it comes to lighting. It is really hard to achieve.

I didn’t have the time to go all the post (I do read quite a bit). But I suppose you are using physical sky. I do read you are using some HDRI too.

Something I don´t like about lighting interiors in Mental Ray is that there is a “feeling” in the image that the light isn’t illuminating in the intensity and falloff that should.

I have found helpful in an interior render with a bright window as a main source light to use one o various area lights with portal lights outside the windows, and try first to create the overall lighting with that.

I have found that the portal light gives a falloff of the light more “realistic”, especially in the shadows.

For me, I try not to use interior lights on (like the celling bulb) when trying to simulate an interior lighting in daylight. At least for me, I almost never have the lights bulb on in my house in full day. And the daylight should almost kill all effect that lights give.

I try to make some observations in your image. I feel there is a mismatch between the intensity of light coming in from the window and the intensity and falloff of the light in the walls.

As in image 01:



Like the overall illumination in A vs B. Where A value is something like 42-50. The falloff I talk is that relationship between the brighter points in the wall vs the darkest point in the same wall, in this area A, goes from 42 in value to 50 in value.

I feel this part is too dark.

In image 02



Look at the value from that part of the wall is almost 20% brighter than the spot A in your image. There is an interesting thing in value as well. The overall value is 78 in brightness, but the highest point in that complete wall is 80 in value. This means that the gradation goes from 78 to 80 in that big area. Compared to the little space you have under the window where the gradation changes 8% in value vs 2% in value in a larger space area.

Obvious I talk about a zone in witch for the nature of the light that comes is expected to have a soft change in lighting gradation.

But when it comes to light 1% in change of the brightness is a big effect in the scene. So a drastically change as for 10% in the value of brightness in a short space calls for a dramatic change of light.

I hope this makes sense to you. But I feel this “falloff” of light; this behavior is what we need to feel something to be “realistic”.

I try to make the same point in image 03.



Something interesting is, if you imaging the image in gray scale, this applies to middles grey, bright gray and dark gray. I have found that if your contrast ratio between the darkest dark and the brightest dark is the same, or almost the same in your entire image the sensation of realism is stronger. The same applies for the darkest middle gray vs the brightest middle gray and the brighter high or almost white gray vs the darkest high, or bright gray.

So, to make my point here is image 4.



Sometimes I fell for every image; we should go for a contrast ratio and stick to it. Every light conditions have different contrast ratios and fallout ratios and those are comes from the effect we want to have. Is a dramatic scene, is an architectural beautiful visualization render, is a realistic photo take render, is a horror scene , so, be aware of the kind of contrast and fallout ratio for me is one of the more important steps in my workflow. Is subtle? Yes in a great majority of cases, but sometimes is quite strong. It depends of the image and what you want to achieve.

In image 4 I show the changes in value in some of the image. You can see how little and “soft” the gradation goes.

Other important thing is the value as it is. Walls almost always, are in the values or + 70%, even when in shadow when are close to a light source (window, bulb). When they are in complete reflection light (no direct light source close to them) can go from 25% to 60%, more probably values around 50% in value.

This is not rule, just my observations but in my work happen to be a great reference for me.

In your image, at least the last one. The highest values or the brighter high or almost white gray is from around 79%. This makes your image low in contrast and you are losing almost 20% of light completely. The overall value goes for a middle gray from 30% to 40% in almost the complete image.

You can see it here in image 5.



I think, maybe a way to go, is trying to get the general fallout of light, working the contrast ratio of your image.

I hate to say this, but mental ray plays hard to get when it comes to this kind of gradation using FG or GI or anything.

Hope it helps, this is my workflow of analysis when lighting maybe it will help you a bit.

Edit: I read all again I see you are using portal lights. Maybe using it as a light card as well in place of area lights can make a little change in gradation? just an idea.

Last edited by IsidroFC : 07 July 2012 at 03:57 PM. Reason: Clarification.
 
  08 August 2012
Thank you, IsidroFC, for the excellent post! Not sure right off how I'll incorporate it into this scene, but it's surely good data and I appreciate it.

Finally I had some time to get back to this project, and am learning a lot from you folks. I've added some plants and flowers, but don't like the corner flower pot much in this one:



I think it's time for a decent backdrop, and perhaps the glass exterior door needs some detailing. Any further suggestions would be wonderful.
__________________
Commodore 64 @ 1MHz
64KB RAM
1541 Floppy Drive


"Like stone we battle the wind... Beat down and strangle the rains..."
 
  02 February 2013
Great creativity. Mine also looks like this but not so much fancy, remodel from San diego General Contractor.
________________
pacificrestoration.com
 
  02 February 2013
I'm only familiar with mentalray in 3dsmax, so some of these points may not be of any use.

Have you disabled FG/GI contribution and possibly even shadow casting on the glass in your windows? Can save considerable amounts of rendering time.

If GI is activated, it will handle secondary bounces - FG bounce-values will be ignored. GI will only emit photons from lightsources and will not emit photons from the sky environment. Using portal lights will fix this, but also means that you will have to add portal lights to pretty much every opening to ensure realistic lighting. Omitting portal lights will result in indirect light gaining warmer color tones because it will be based on sun photons and not on photons emitted from the sky. Definitely something to take into consideration if you are aiming for photorealism. On the other hand, having a decent amount of portal lights will probably lead to noticable rendering time increases. On top of that portals will ignore objects outdoors (if you ever plan on adding trees etc. as outdoor geometry). There is an option to enable sampling for that, but that will increase rendering times even further.

I never felt as though pure FG renderings with high enough sample settings look unrealistic. Specially when comparing results from iray to mentalray. FG may seem slower at first (slower secondary bounces), but I've found that using GI often ended up leading to quite a few drawbacks (outdoor geometry not being sampled, problems with portal lights + FG calculations etc.), which ended in rendering times being very similar and often even higher than a FG-only approach. That will of course vary depending on the kind of scene you are working on.

Nice to see things coming along! Will definitely keep following this
 
Thread Closed share thread



Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
CGSociety
Society of Digital Artists
www.cgsociety.org

Powered by vBulletin
Copyright ©2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump
Miscellaneous

All times are GMT. The time now is 05:14 AM.


Powered by vBulletin
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.