WIP/project log interior

[Iain Denby]

Ernest

. We used to be able to deliver in 3 days but the clients nowadays are getting more demanding. Can't impress them in one shot like old days.

 

That's exactly the same thing with my situation. Clients are slowing me down these days.

[Iain Denby]

I have to back up 4DM on every point I'm afraid Ernest (but I was too apprehensive to say anything, because it's been a great thread, and a good education for many of us)

 

With your finished style in particular, you could (should) take many shortcuts during every process, which would be undetectable in your finished visual. It's the likes of the 'Vray boys and girls' who need to be pedantic to get the best results.

 

Perhaps this is why you have to search for something new every time, because you become stale during every job because it goes on and on.

 

Anyway, you're on the home straight.

[Ernest Burden III]

Perhaps this is why you have to search for something new every time, because you become stale during every job because it goes on and on.

 

I enjoy making images, and look at each project as a new opportunity to do something good. I don't want to just pump out the same image everytime, though my work in the last year has begun to look that way. I also made more money last year than any other--maybe there's a connection?

 

Anyway, the main reason I chose THIS project to do a log on was that I had a six-week production schedule. There was time to write all this stuff, prepare images to post, etc. Also, there was time to try out some new techniques and try using Cinema for an interior project.

 

And I took almost two weeks in the middle to work on another project. Enough sounding defensive.

 

I have to deliver this rendering on Monday, and still have a few important things to do. The worst of them is adding some people. I always thought this would be a money-losing aspect of the project. People in a restaurant are never easy. There is simply no easy way to add them. They must sit in chairs and be arranged just so. No photo clips to drop in and run.

 

The rest of the image is in pretty good shape--a few textures to add or improve, a few lights to adjust. The downlights above the tables are not centered exactly over the tables on the designers plan, but need to be, so I must move them.

 

I still have to tweek the radiosity samples. Right now I'm just guessing ala STRAT. Yet, as Dibbers points out, I can get away with a messier image since I fuzz it up anyway.

 

I ran some general images overnight to get an idea of how the lighting was working. A few bits to fix, but pretty much there.

[Iain Denby]

Ernest

 

The first of the 4 images looks great. However, although you said there were a few tweaks, at this stage the other 3 images are too dark. Your style, together with the current lighting scheme, looses too much detail and interest, leaving little to look at. Perhaps it's too much contrast, I'm not sure, but not enough ambient lighting for me.

[Ernest Burden III]

The first of the 4 images looks great...the other 3 images are too dark.

 

Absolutely. I have no intention of leaving them like that. It's just what the software left for me this morning, plus my 'one-button' NPR treatment. I will look at the overall light level. By having it overly dark right now it is a good way to see the effects of the lights. By the way, I have no reason to believe that the real space will not also be dark. Did I miss any lights? There are 100 or more of them. Do they spread more than I show? Maybe. But I think the place will be dark.

[Chad Warner]

I do agree with dibbers that the scene looks dark, but based on what I see of your imagery, it looks like the scene should be dark, so in that case, I wouldn't make the images brighter just for the sake of making them brighter.

[Iain Denby]

I wouldn't make the images brighter just for the sake of making them brighter.

 

No, but you should if it makes a better visual, and if it prevents the viewers thinking 'it's too dark'.

[Ernest Burden III]

No, but you should if it makes a better visual, and if it prevents the viewers thinking 'it's too dark'.

 

My thoughts exactly. It looks too dark. It shouldn't. Its fine if some areas are dark for the purposes of 'popping' another part, but this, so far, just feels dark.

[Ernest Burden III]

I love Cinema's stochastic mode of radiosity. Its always smooth and beautiful, even with low samples, so long as you don't mind some noise, which I don't mind at all. But it just takes too long, and the lightness values change from regular mode.

 

Here is a small illustration of that, a small area in which I was seeing a lot of radiosity artifacting rendered in both regular mode and stochastic. Both are with accuracy of 66% and stocastic samples of 64. The regular one took 32 seconds, the other 32 minutes, or 64X The stuff I rendered overnight took a few hours each at 1000x600 resolution. So could I live with an 80 hour render? Uh...no.

 

However, when I do my noise thing, you can see how those ugly artifacts just become part of the 'artistry'. They are no longer so ugly.

 

(By the way, I was working in CAD while rendering these, so it may go faster with the CPU to itself, but not by much).

[Ernest Burden III]

Cinema4D is really letting me down. This program needs some work.

 

I spent most of Thursday and Friday getting moving on two new projects, so have now to just finalize the lighting on this piece, render it and, I guess, figure out the people issue. But I'm looking at that as an 'add-on'. I want to get the main rendering done without them first, then add people in a seperate 'version'.

 

So I did the lighting, which amounted to raising the levels on a few lights and adding an area light outside to simulate some last rays of sun coming in. The problem with that is it ends up being cery much light the crutch of adding 'ambient' light to a space--even, lifeless brightening. So I carefully set a falloff so the outside light was stronger by the windowwall than the interior. In the end, though, it caused dark patches because it was shadowing badly, so I had to turn off the shadows. There, I did it. I have been reduced to using 'infill' lighting, which I really, really, do not want to do. But without it, areas of the picture render black. If a part is simply dark, I can bring it up in Photoshop. But black pixels have nothing to bring up, no color.

 

So I've locked down the view (and not one I particulary like, though not as bad as the client had suggested but not as good as what the designer had told me from the start would be used) and was on to rendering the thing.

 

Now, I admit I have used a lot of lights. A lot. Somewhere between 100 and 150 I would guess, as required by the design I'm rendering. But I have been careful with setting falloffs and using small shadowmaps where possible, etc. Too many lights? No, software not up to the task of architectural vis more like it. Maybe next year...

 

Rendering the final involves several passes. One id for materials so I have a mask to use in Photoshop for object-by-object tweeking, that map has all textures removed except alpha channels. Another image is made with three or four lights without shadows to provide a basic edge-defining picture from which I produce linework.

 

Radiosity is the next issue, which I'll get to in the next post. First was to just render a final without radiosity, which you see attached. I am working to a final size of 3000x1500 I usually dont go over about 2800, but since the client wanted 'more panoramic' I have a lower height so a wider width. I find too much detail takes away from the look of my finals, and of course takes longer to render.

 

It actually took two hours to render the picture WITHOUT radiosity. I was eating dinner and watching a movie with my wife, but COME ON. Two hours? That's with only moderate AA except for high AA forced on the wall metal rods. Not that it eliminated the moire pattern they created. I suppose it would have been worse without forcing 8x8/16x16 AA. I'm not sure how to deal with that. I guess I'll find out later today.

 

 

Cinema says the file has about 347,500 polygons (without any 3D figures).

[Ernest Burden III]

I would love to post a radiosity image. The trouble is, I don't have one.

 

Here's the workflow idea--render the radiosity to 1/2 size (so in this case 1500x750) using the multi-pass output. That makes just the radiosity a seperate layer that you put on top of the main render and set to either 'screen' or 'linear dodge' mode and it simple adds the bounced light to the existing render. So I will enlarge the image to the 3000x1500 size. It doesn't matter if the layer gets a bit fuzzy from the enlarge, the detail is carried by the base image, the rad layer justs adds lighting.

 

But I do not HAVE a radiosity render. Twice I have had it rendering on another PC (via NetRender) for the better part of a day to simply have it disappear. No error message, no hint of crashing, it just stopped, closed the NR client and left no result that I can find.

 

Then I ran it on my main machine with the full C4D program and it was doing fine all night, but was nowhere close to being done this morning, even though I have followed STRAT's suggestions for settings. Try again, I guess.

[William Alexander]

But I do not HAVE a radiosity render. Twice I have had it rendering on another PC (via NetRender) for the better part of a day to simply have it disappear. No error message, no hint of crashing, it just stopped, closed the NR client and left no result that I can find.

 

Then I ran it on my main machine with the full C4D program and it was doing fine all night, but was nowhere close to being done this morning, even though I have followed STRAT's suggestions for settings. Try again, I guess.

 

 

what about rendering out regions, and comping the pcs into a full image? That may reduce some of load that's making C4D blink out. It's possible to do regions on different machines. But that may depend upon the radiosity clcs not being the problem.

[Ernest Burden III]

what about rendering out regions, and comping the pcs into a full image?

 

That is possible. I hadn't bothered because I had it running on a machine I wasn't using, so I didn't care if it took longer. I do care if it never renders, however. Currently it is 1/3 - 1/2 through on this machine, at about 1 hour so far. We shall see.

 

I do plan on handing this thing in tomorrow, so it would be really nice if it would render the radiosity.

 

It is a bit of freedom to be able to seperate out the radiosity and proceed with other things while it runs. You want to believe that an image is some magic 'whole' but in fact it can be broken into reflections, refractions, diffuse, direct illumination and radiosity. The render engine probably processes these seperately, then combines them for a final. Getting it split out in a Photoshop file gives control of those elements back to the artist. I like that part.

[Alex York]

I feel your pain! My newest render (in my thread) took 11 hours to render last night. I woke up this morning to witness the last bucket in the final stages. Gotta love that luck

 

Passes do, indeed, rock. Without them I'd be useless. It's just so handy being able to tweak something knowing you're not ruining something else.

[Ernest Burden III]

Unbelievable. Even with carefully pared down settings, this radiosity render took 10 1/2 hours. And that's at 1/2 size.

 

I'm goint to test 1/4 size next week. I'ld bet it will work just as well. It would be one way to reduce render times. However, you don't see any radiosity in the map for stuff that's beyond glass. I wonder if that's a product of the difuse depth, or a product of seperating the refraction pas. Maybe if its not seperated, you will get 'through the glass' alomg with the radiosity. I'll try that, because I want it together for this smaller-size pass idea to work properly.

 

So even lighting gets you boring radiosity. Isn't even lighting the point of the lighting design? No hotspots, no dark corners?

 

Here's the radiosity render in full and just the rad. pass:

[jackb602]

However, you don't see any radiosity in the map for stuff that's beyond glass. I wonder if that's a product of the difuse depth, or a product of seperating the refraction pas. Maybe if its not seperated, you will get 'through the glass' alomg with the radiosity. I'll try that, because I want it together for this smaller-size pass idea to work properly.

 

Yeah, the refraction pass effectively "claims" any GI that occurs beyond the glass, so for your purposes you may just want to turn off the glass for the GI pass. By the way, the renderings are looking beautiful.

 

Jack

[Alex York]

granted that's a very complex scene, but the quality of the radiosity in those could be a lot better. shame, because the scene is gorgeous.

 

what are your system specs + software packages? for 10.5hrs you should be getting absolutely smooth results with very detailed and focused occluded areas.

 

have you tried Ambient Occlusion and Final Gathering as an alternative to radiosity? That combination can produce stunning results in decent time.

[Ernest Burden III]

granted that's a very complex scene, but the quality of the radiosity in those could be a lot better. shame, because the scene is gorgeous.

 

have you tried Ambient Occlusion and Final Gathering as an alternative to radiosity? That combination can produce stunning results in decent time.

 

In Max, maybe, but I don't use Max. I think we need to see how to get better results out of Cinema...and I do NOT like hearing that the answer is a bunch of faking. Damn it, the results probably look just as good. But we have this great tool, it can produce stunning images, but not in a timerame that makes it usable in production. That's pretty frustrating. Especially since Lightscape made beautiful pictures in reasonable times. But then it got 'bumped off'.

 

I did have one version of the radiosity cooking with all the major glass turned off. But it never rendered.

[Tim Nelson]

Come on, just buy yourself a copy of Viz+Maxwell & be done with it already.

 

J/K. I know you are waiting for the stand alone version.

[Ernest Burden III]

Come on, just buy yourself a copy of Viz+Maxwell & be done with it already.

 

Maxwell works with Viz?

 

Fran has me just about convinced to get the 30-day trial of Max the try Maxwell. I just might.

 

You know, after Autodesk killed Lghtscape they started contacting all of us about how great it would be to buy Max. They mentioned their generous 'sidegrade' discounts. Guess what--not applicable to Lightscape users. Everybody else could get a better price, except those of us that were their own (now abandoned) customers.

 

I'm trying to remember why I don't like Autodesk...

[Tim Nelson]

Maxwell works with Viz?

 

Yep, I just checked their site & it does work with Viz 2005. That's a much cheaper altnernate to Max.

 

I agree there are a lot of reasons to NOT like Autodesk & Discreet, but unfortunately there are a lot of things to like about their products. I just hope that Viz will stay alive so I don't have to shell out the extra cash for Max which is full of features I will never even use. You never can tell with these guys...

[Alex York]

the Maxwell alpha is not free unfortunately. and it's incredibly slow, too, from what I've heard.

 

i would stick with Cinema and keep tweaking your settings.

 

btw there's nothing wrong with "faking" indirect illumination. the only renderer that truely (as much as we can expect for now) doesn't fake anything is actually Maxwell. being able to get the same result in a quarter of the rendertime can't be bad, even if it means a few work-arounds. I don't know if there are any ambient occlusion shaders available for Cinema though, sorry.

[Ernest Burden III]

I don't know if there are any ambient occlusion shaders available for Cinema though, sorry.

 

There are. I've tried it, it's a great technique.

[Ernest Burden III]

I'm going to get beat up by the Brits again.

 

I have ALMOST completed this project. I did the final WITHOUT people. but must put in the peeps.

 

I've already done most of it, but have to rush through to get it DONE and to the client. I'll post the process of trying 3D model people tomorrow, once I'm done.

 

Here's the image:

[Alex York]

that's superb! I was just checking out your gallery again and I really like your style... almost got a hand-drawn quality to it. very original.

 

how are you going to tackle people? RPC?