Skylighting

[robk]

What should I do to get more skylighting into the attached scene thru the windows.
1. Increase Ambient lighting.
2. Boost backdrop lighting.
3. Boost global Illumination lighting.

The sunlight itself does not seem to bounce much and effect other surfaces either. Is there a setting for surfaces to control how much light is passed on/reflected to other surfaces. I have also posted some other questions in the Works in progress thread if you would care to look.

When using lightscape the radiosity solution always seemed a lot brighter than in lightwave. I wonder why that was? There does not appear to be a high level of colour bleed (passed on light) in the Lightwave solutions.
Any help would be appreciated.

I think I will also replace the luminous rectangular uplight panels in the bulkhead with wide spread spotligts to get more light on the ceiling.

I have also attached some actual photos from the project for reference.

[ingo]

Looking at the photos you need to boost the backdrop lighting, and/or the raydiosity intensity. The windows in the photos are completly white, but in your rendering you can see the blue sky.

[robk]

Maybe when I am setting up the backdrop in my mind I know the sky is blue so I show it that way even though the photos show it washed out. I suppose if I would balance the lighting in my photo so the outside sky would be blue then the interior would appear much darker in reality. The human mind compensates for the lighting while the camera tends to show it like it is. The day the photos were taken it was a cloudless sky.

[vpii]

Are your windows double sided? If they are you maybe are not letting enough light through. If not the case do like ingo said.

[robk]

The windows are not double sided as in one polygon assigned to be double sided but are an enclosed element, like a cube which happens to be 10mm wide, so there are two faces to make up a window 10mm apart from each other with the normals pointing out from the faces.

[robk]

I also used the interior glass settings from the lightwave glass presets panel on them. I don't know if that preset makes the panes double sided of not.

[Captain Obvious]

I'd turn up the backdrop lighting quite a bit, maybe increase the radiosity strength a tad, and, if at all possible, increase the number of bounces. Increasing the number of bounces will make the scene appear brighter, and the light more filling. Finally, play around with the gamma settings (or levels in Photoshop).

Oh, and how are you rendering it? FPrime? Standard Lightwave? If it's LW's native, is it monte carlo radiosity? Backdrop? Interpolated?

[robk]

I am trying to render in fprime (Monte Carlo 2 bounces at the moment) Biggest problem is I am running out of time. Project needs to be complete for next Monday and I need three shots. Two interior and One Exterior. I'm afraid if I assign too many bounces I will not get them done. I only have two computers to handle the load at the moment. I am looking at setting up a renderfarm but although cash from previous projects has just arrived I don't think i could get them all ordered and set up befor the deadline. Also the idea of not knowing how long they will take with Lightwaves native renderer scares me a bit.
Thanks for the suggestions I will try them out.

[nthused]

Also, I've found that using a gradient texture for your backdrop is MUCH brighter than the standard backdrop setting in LW. Wirh the Textured backdrop you can up your value to 200% or more if needed also.

[Captain Obvious]

Do a render in Lightwave's native renderer without radiosity and just one pass of AA. Using interpolated radiosity with tolerance set to one and rays per evaluation to something low, you can often have 3-4 bounces before it starts taking much longer than a render without radiosity. Of course, I haven't tried rendering anything of the complexity of your scene, so I don't know... But try it! It's not really perfect, but it gives really fast results. The main secret behind interpolated radiosity is motion blur, and lots of it. That makes predicting how long the final render will take a bit easier, since you can do 1-pass test renders.

[otacon]

If i were you i would forget about raising the number of bounces and instead just place some lights inside to lighten it up. When its a big area and the windows are not that big, raising the bounces from 2 to 4 isnt going to do much.

You say the ceiling is too dark, then put some spotlights pointing up toward the ceiling with no shadows. I would put an area light in that big cylinder on the ceiling, it would work better than that luminous poly you have up there now. Any other areas that need brightened, i would use some point lights with linear falloff. You dont have to rely completely on radiosity, you can save tons of time by adding some lights to fake the bounces.

[hulagirl7]

If i were you i would forget about raising the number of bounces and instead just place some lights inside to lighten it up. When its a big area and the windows are not that big, raising the bounces from 2 to 4 isnt going to do much.

You say the ceiling is too dark, then put some spotlights pointing up toward the ceiling with no shadows. I would put an area light in that big cylinder on the ceiling, it would work better than that luminous poly you have up there now. Any other areas that need brightened, i would use some point lights with linear falloff. You dont have to rely completely on radiosity, you can save tons of time by adding some lights to fake the bounces.


Absolutely! Fake it! Make your own bounces. Using falloff and point lights
you can do some convincing stuffs. That's all LW is doing, but using polys as
lights. We LWers only had radiosity fairly recently!
Otacon is dead on with this 1.

[ingo]

Another thing i see in your picture is the glass elevator. It looks like there are a few glass paneels to look through, that means a lot of rendertime too. Try to make the inner glass without reflections and very transparent. And if you render in full 96 Bit , for example as hdr, you can later brighten up the rendering in image viewer, worked well for me several times.

[Captain Obvious]

We LWers only had radiosity fairly recently!
Didn't LW get radiosity in version 6?

[robk]

this is the result I got after 30 hours of rendering.
otacon I didn't use an area light in the big round skylight, that is sunlight coming through the top of the cylinder.

I am not sure if you can see it in this reduced picture but i also have speckles from the sign/art on the left throughout the scene. The blue panel is luminous but the yellow lights are just point lights, yet it seems they are also causing speckles on objects in the scene.

All this talk about faking it (radiosity that is) bothers me. The reason I am moving from Imagine to lightwave was to do radiosity. If i had to fake it i can Do it way faster in Imagine and my renders will take minutes rather than 30 hours to render.
I used to use Lightscape (and still have it) but since it is discontiued I decided to move to Lightwave. Lightscape produced lovely renderings but was lacking in the texturing department as well as the time it took to make models that would work in the renderer.

I also got another rude awakening when I rendered my outdoor shot. The brick image I mapped on my walls at 45 degrees (in Plan) showed up perfectly in fprime where I proofed it but didn't show up on the side walls when rendered with Lightwave's native renderer. The only reason i didn't render the final in fprime was it crapped out when I tried to do a 3000x1500 image with fprime render, so I did a 4500x2250 with Lightwave's native renderer. (see clips attached)
I have run into other instances where Fprime and Lightwave has produced different results. has anyone else run into such problems?

You will also notice the difference in the skylight glass from fprime to lightwave in the brick pictures. these are the exact same scene as I rendered the lightwave version right after the fprime one crapped out because of lack of memory.
The interior shot is radiosity and the exterior is just raytracing

[robk]

Since I am in a rotten mood this morning I would also like to get this off my chest.
When rendering in Lightwave native renderer you only get a measly 320x200 or 640x480 preview window so when i rendered my 4500x2250 exterior image I only got to see aprox. 7% of the image and that part looked great but of course the part that screwed up (the brick mapping) would hace been evident quite early if you could preview the image will it is rendering. In Imagine you get to view the entire image as it is rendered and can zoom in on it to any level you want so you can check for errors (which should'nt of occured) and stop the render and adjust the problem.
Don't get me wrong I am not really down on lightwave. When I get more experience using it I hope i will be able to avoid most problems. The difference between fprime and lightwave's results is what alarms me. Since i use fprime similar to how I would use imagines's renderer (to zoom in to check if small details are ok) i would assume the results you get in Lightwave's native renderer would be the same. Although I am trying hard to wean myself off of Imagine when I run into problems like this it reminds me that Imagine was a very good program for its time.
I have included a small rendering of the exterior jus for the hell of it

Thanks to eveyone again for suggestions.

[Exception]

I have some answers for you.
You first posted the question why the scene was to dark compared to the photos. It was observed that in the photos the sky was white while it was blue in your render. this indicates a lack of backdrop light strength. while you COULD use a backdrop texture, I prefer to use Picky as a color picker and use the X multiplier to boost the backdrop (find picky on flay.com). Make it a lighter blue, and give it X=3 to 6 for nice effects. Second, I'd try using LW's interpolated radiosity method as a final render. Use the motion blur trick (set motion blur to ON, Cache radiosity to OFF, Motion blur length to 0 if it matters to you, and set the tolerance of the interpolated radiosity really high like 0.7~0.95, you can get multiple bounces really fast this way, which looks good! Let the render finish though, you cannot judge it after one preview pass, quality goes up with more AA passes)

The brick thing seems to me like a texture placement problem. I had those a lot with procedural bricks. Is it a procedural? If so, I bet the wall is exactly in between bricks. Other wise you can photoshop it in easily.

[robk]

The brick thing seems to me like a texture placement problem. I had those a lot with procedural bricks. Is it a procedural? If so, I bet the wall is exactly in between bricks. Other wise you can photoshop it in easily.

The bricks are a large map being repeated (3000x1500) 25 bricks wide by 15 bricks deep. The bricks are mapped on at 45 degrees to the the walls so all I have to do is scale the map width to .707 of it's width in the horizontal direction to get brick mapped on it's correct proportion. I have done this for years in Imagine and it works a long as walls are perpendicular to each other. You have to map straight on for sill bricks though otherwise you might see the grout joints streaking out a 45 degrees on the top of them. If they were procedural I would agree with you. I have had cases where the grout joint occurs right at the wall corners and thats all you see on the side walls. I think I used cubic mapping for the top soldier course and I might try that for my main brick.
My big beef was that they showed up perfectly in Fprime but not in Lightwave's renderer. That doesn't give me much faith in setting up a scene in Fprime and then letting lightwave's native renderer do the final rendering.

I will try your other suggestions tonite. I can't figure out why my scenes take so long to render. I am not sure if it is too much geometry or my lighting setups suck.
I will try your suggestions and some of the suggestions from other people and report back.

[Pavlov]

my suggestion for this ambient are:

1 - download Picky and use that as color picker. Set backdrop color as light blue and use a multiplier like 3x or 4x (this mean backdrop casts a stronger light);

2 - use a 200/300 % sunlight (orange/reddish) and a 30-40% ambient (use ambient color to fine tune the atmosphere, making it colder or warmer);

3 - use 2-3 bounces. More than 3 bounces often make LW an unviable solution, especially for high rez. If you need further bounces play with ambient or secondary lights;

4 - Use interpolation; start from things like .4 / 400mm with at least medium AA, then refine it until you get rid of blotches;

5- exclude glasses from radiosity (put them in another layer and deactivate all options you dont really need - you can even cut off them and add reflection in post);

6 - increase radiosity intensity until you are satisfied.

This way you should get optimal results in affordable times.
good luck

Paolo Zambrini

[robk]

Join Date: Mar 2003
Location: Netherlands
Posts: 351 I have some answers for you.
You first posted the question why the scene was to dark compared to the photos. It was observed that in the photos the sky was white while it was blue in your render. this indicates a lack of backdrop light strength. while you COULD use a backdrop texture, I prefer to use Picky as a color picker and use the X multiplier to boost the backdrop (find picky on flay.com). Make it a lighter blue, and give it X=3 to 6 for nice effects. Second, I'd try using LW's interpolated radiosity method as a final render. Use the motion blur trick (set motion blur to ON, Cache radiosity to OFF, Motion blur length to 0 if it matters to you, and set the tolerance of the interpolated radiosity really high like 0.7~0.95, you can get multiple bounces really fast this way, which looks good! Let the render finish though, you cannot judge it after one preview pass, quality goes up with more AA passes)

I tried your method noted above and the results are attached as well as the last Fprime one I had done on Sunday. I only had the interpolated one set to 1 bounce though and did not have the Picky thing happening (I haven't downloaded it yet). The results seem ok except for the blotchiness on the ceiling and the washed out floor. (also some colours on the model have been changed since getting comments from the Architect).
I tried to do a small (300x200) rendering this morning before i went to work with a bounce setting of 6 but it didn't get past the calculating radiosity solution stage.
The interpolated one I rendered last night took about 5 hours for a 1800x1200 image. I can live with times like this. The big problem now is i am rinning out of time. the new completion date has been moved to next monday and i don't know how many things i can try out before then.
Do most Lightwave users tend to use Interpolated over Monte Carlo?
Is the interpolated always faster or easier to control.
Spent most of my time last night fixing colours and other items so I didn't have a lot of time to test the lighting/rendering options. I see there were other suggestion about using interpolated. If someone could give me a list of what all the setting they would suggest i would appreciate it.

Thanks so much again for all the help. There seems soo many settings to deal with it mind boggling. (I still don't have a clue what classic recontruction setting does as well as all the new AA modes)

[cgbloke2004]

we only use monte carlo if we know that we have *plenty* of time to tweak, refine and have oodles of render time.
normally in production we dont, so we do the quick and dirty - that is, use interpolation - usually till the blotchiness goes away and we find a happy medium.
[monte carlo is a 'bit' like turning all options on interpolation to the max - interpolation is like the quick and dirty little brother of monte carlo - you just get faster - but less accurate - results]
we also use a mix and match of both radiosity [to 'boost' colour bounce..] and faking with real lights.
it just speeds up the process - its just doing what you need to get done to get the job out and the client happy.
you can always go back to it later of your own accord and do what you really like with it - within production you dont really have luxuries of time to get it perfect..

hth

[caesar]

Well, why not use Fprime??? You said you got out of memory = HD, isnt?

[robk]

Well, why not use Fprime??? You said you got out of memory = HD, isnt?
__________________

I tried Fprime first but since with the scene loaded the file was bigger than my available ram so it could not save the Image. Worley's group informed me that Fprime uses Lightwaves savers and they need a contiguous section of ram to save an image out and as I want to render at 4500x2250 pixels and Fprime could not do even a 3000x1500 image (without a save error) I decided to use Lightwave's native renderer. I am bumping up may ram to 2 gb so I might try Fprime again after that.

[Hervé]

err... where is LW 64 when we need it.... :D

[robk]

after a frustrating hour of trying to get my bricks to map on I finally got it to work by changing the mipmap setting to Zero. I think it was someone from Worley Labs that suggested I adjust the mipmap setting. At first I adjusted it up and then after set it to zero and voila! the bricks finally appeared. i must read up on the mipmap setting and see what it does.

[Nigel Baker]

Hello there robk,
just by sheer co-incidence I bumped into this for the first time ever, so here is what the LW manuel said...


Mipmaping is similar to what is used in today’s games to avoid graininess of textures in a distance or
at a flat angle. Basically lower- res versions of the texture are generated in realtime and blended in.
This feature is supported in Hardware by most of today’s graphics cards. This feature also works if
Mutitexturing is turned off. Please note however that due to the nature of this filtering- method,
low- res- textures may appear a bit blurry.


All the Best.

[robk]

here are the finals for anyone that is interested

[Nigel Baker]

Hello robk,
Thanks for the insight, it has been interesting to see the developement.
I hope you are proud of your work. I sometimes wish clients would not ask for people within their buildings.

Thanks again and well done.

[caesar]

here are the finals for anyone that is interested

Great building...with, you´ll see the great render LW has!