Rendering Critique

[alobartn]

Excellent explanation, Christina. We follow similar routines. One of several things you do differently (and better) that I would do well to apply is to place fewer lights and add the missing hot spots in Photoshop. It cuts down on Raytrace time significantly. I had eight lights turned on, two with shadows. The render took 30 hours! While it's a nice touch of realism, the trade off in time isn't worth it when I could have just dodged in a similar effect under the cabinets and elsewhere.

Your choice of materials and their settings are wonderful. I love what you put on the stools and the window seat. It matches the photo much better, as do the cabinets and steel materials. I look forward to the final product. Hopefully the mag will accept a second submission!

[cgirerd]

The stool cushions are the same fabric as when I got the file, but I changed it to matte finish - I think - I might have made some other small adjustment, I can't remember now and the raytrace is running so I can't check, but the fabric didn't change. The window seat cushion fabric I found on a fabric website for "mid-century" fabrics, then I made it seamless in TextureStudio. (They had some really cool "atomic" designs, deco and even Frank Lloyd Wright inspired ones. I think it was this site... www.decodame.com/fabrics/index.html)

The raytrace I started last night at about 9pm, is still going... since I upped it to Med. Rad with the High AA, so it's been 10 hours, and it says (currently - though that indication is rarely reliable) that it has about 8 more hours. I'm still running it at the larger size 3134x2134 because it is a multiple of my screen size. I will shrink/crop as necessary to get the 2700x1800 in photoshop so I can get the benefit of the better resolution in raytrace. If I had just put 2700x1800 in my raytrace dbx, I don't know where the program might have cropped or stretched or what, and this takes to long to not know what I'll be gettin at the end! Hopefully I'll have something to post late today.

Christina

[alobartn]

Wow, that radiosity really makes a difference. Thank your lucky stars you had the foresight to get a dual core machine, so you can work while "it" works.

I agree, the last thing you want after you've waited for eternity for the thing to finish is a surprise. On the other hand, you don't want to wait longer than necessary by Raytracing an image larger than needed. You can stretch/drag the Open GL render window to an easy-to-figure ratio, pixel height by pixel width. When you click and hold on the sides or bottom or a corner of the render window, there's a little information on the bar that reveals its size. Then you'll know that if you assign the Raytrace size to be a multiple of the window size, nothing will be cropped.

For example, if 2700 x 1800 is your target size and your screen maximum is 1567 x 1067, you can divide the target by two which would make it 1350 x 900, which fits on your screen.

Since you can't see the Raytrace progress when you're rendering larger than your screen, this is a good way to know exactly where the borders will be before you start, and you'll render at maximum efficiency.

[George Godwin]

Christina & Adam, - I am looking forward to seeing the final results. In the meantime, here is a little Photoshop trick that I have used over the years. See the image below.

Put some white lines and an oval on a Layer in your image program and adjust Gaussian Blur and Opacity of the Layer to suit. Coloured lines can be used. These can help with can lights under cabinets etc. Also, a hot spot on a glass light fixture can be made the same way with a small circle blurred and opacity adjusted. This approach can be used to create lighting highlights where needed as well. (Adam, this method for highlights and rays is easier to control compared to using the Dodge tool in my experience.) It can give some highlights on your countertop for example and the highlights can be stretched or contracted with the Scale adjustment to suit.

This is an obvious approach and is very easy to do. You may be familiar with this tip but it may be new to others following this thread.

[Larry Kumpost]

For example, if 2700 x 1800 is your target size and your screen maximum is 1567 x 1067, you can divide the target by two which would make it 1350 x 900, which fits on your screen.

I may be missing something here but I would NOT do this. As I understand it Chief is screen dependent with the renderings so if you render an image that is half the size you want to print you will have half of the detail (pixels). That is why they added the ability to render images that are larger than screen size.

[alobartn]

You're not missing anything. I just didn't explain it well. Christina's target size is 2700 x 1800. Her maximum render window is 1567 x 1067, if I understood correctly. She didn't want to tell the Raytrace to render at 2700 x 1800 without knowing what POV Ray would crop, so she just told it to double her render window.

Doing this makes the Raytrace larger than necessary, thereby requiring more time. I simply explained that there's no guesswork once the render window ratio is correct. I don't think it matters that the render window is as large as possible for the Raytrace resolution to be maximized. At least I can't see a difference when Raytracing from a small render window.

[Larry Kumpost]

Adam:

If what you are saying is:

If your window size is in proportion to what you want for the final image you should get the same view boundary then I believe that is correct. There is, however, something to be said for having a little extra. It is much faster to do it once at a slightly larger size than it would be if you just miss and need to start over.

[George Godwin]

Bigger is better since you can do so much more with a lot of pixels if you need to adjust some of them. As Donald Trump would say "Go Big or Go Home".

[alobartn]

I agree completely unless you already know exactly what your boundaries are and don't want to spend extra hours waiting for needless render areas.

[cgirerd]

I could have adjusted my window as mentioned and doubled or 1.5x the size as needed to get the exact 2700x1800, but I had already done the double size for lowradHiAA and since it took under 7 hours, I didn't think there would be such a huge leap of time involved upping it to med. radiosity. Obviously I was wrong! It's now been running 15 hours and says 9 hours left... I guess this is why I normally don't run at medium - my normal clients would never even notice the difference. A magazine spread is different though - you want really good detail. I hope the result is worth the change! I also hope the small changes I made work out okay! And as you said Adam, good thing I have dual core. I'm sure it's slowed down also because at least since I started working this morning, I have been running up to 4 other programs in addition to the raytrace... Maybe I'll hit your 30 hour time after all!

Christina

[alobartn]

What line is it on? Seems that with this rendering there's such little information on the bottom quarter, it starts to run fast once it gets there.

[cgirerd]

It's on 1365 of 2034 and it's 2:50 pm, it says 69% and 8:49 left. You're right - it usually does go faster at the end.

Christina

[cgirerd]

It's finally finished! It finished sometime while I was at my daughter's Open House at school, between 6:30 to 8. So overall it ran around 22 to 23 hours, for the Med Radiosity, High AA and size of 3134x2034. Although I resized it to the magazine requirement of 2700x1800. I've uploaded the whole picture (although it is med. resolution and small to fit the 100kb limit...) and a few details to see better.

Christina

[George Godwin]

Great job, Christina.

[Larry Kumpost]

Christina:

Looks great!

My guess is you probably tweaked a few settings from the Low Radiosity image you did and that would account for some improvement even at the Low Radiosity setting. The question for me is if you think the extra time, for even the Medium Radiosity setting, is justified in most cases? Do you get/see 15-16 hours worth of improvement?

Just curious.