Quote - In case you are curious how crazy I am, I rendered a 15,000 pixel wide version of the "building the ftl1" image I made and it took just a little over 4 days.

However, I'm going to have it printed 5 feet wide and I'm going to hang it in my office and a lot of people are going to see it in a very big way.  That four days of render time will pay off.

That image was very much worth it, so I wouldn't call you crazy at all.

But in my case I simply cannot afford such render times, just becuase I need my computer during the day to do my work. Unless I can time it to coinside with an away-without-computer vacation, it's not really an option :). So, it's pretty much a once-a-year kind of deal... At most I can leave it to render overnight, but that means I need to keep my renders under 12 hours for the absolute maximum render time.

In this case I am fine with the final render taking those 12 hours, it is a very complex scene, but I need to keep it under this limit, which means that I need to build this project economizing on the render time as much as possible from the very beginning, making sure that any components I add will not slow me down too much. So now I am going to look into the model to try and figure out where the render-slow complexity is coming from.

Straight out of the box the scene rendered on my machine in 15 min, so quite comparable to your results, which I was greatly surprised by.

Then I started shifting it closer to what I am aiming for: One of the first changes is adding the render passes. I need them in my workflow for post, so I never render without them. I added all the usual passes I render and re-rendered your scene with nothing else changed: 20 min. Still quite good.

Now I replaced your model with the classroom I was using before and immediately the render time shot up to over an hour (attached file - 1 hour 12 min render). And that is without the replicated lights or any other requisite items - just the bare classroom. So I am guessing the issue is in the model and, possibly, my shaders. The shaders had little influence on render times as I was changing them, but I intend to experiment more with them. I will post my results as I get them.

In the mean while I rendered a fake bounce scene (So my spot for the window shadows, plus a huge amount of low intensity lights all over the place to fake bounce - about 75 of them. I didn't alter their color yet, so the whole thing is still gray, but I plan to change that eventually to have a more believable color bounce.) with AO only for indirect light. My file (not very small this time, but not full resolution either yet) rendered in 33 minutes. So big improvement over 4 hours, but the quality is clearly not the same :(...

P.S. this should also give you an idea of how populated I want my final scene to be: each of these chairs should have a fully clothed person sitting in it, plus there would be one more fully clothed person by the board.

Kixum, your results here are beyond amazing. Either your system is very far ahead of mine (possible, since this thing is about 5 years old now), or you are doing something that I need to learn how to do :).

Could you share your scene file with me somehow? So I could run a render and compare the times I get with the same settings on the same scene?

(If you don't want to share it publicly, maybe you could e-mail it to me, I am antara.nn on Gmail)

The quality of the light in your scene is exactly what I am trying to get, but I am not getting anywhere near it within the same render time scopes. I was able to fake it though, using Ambient Occlusion only for Indirect lighting and then faking the bounce.

I will post the result when the big image renders.

Andrew, my scene will have more stuff in it... Your image lookg good, but I don't see the kind of shadows lurking in the corners natural bounce that I am looking for...

Stanley, I know that people have fooled Carrara into doing good quality lighting in less time, so I am hoping to see if anyone has good at least some optimization ideas. Maybe there is a setting somewhere which is not strictly needed and can be skipped without much quality drop, but with speed improvement. For example, turning "light through transparency" off really does shave off a lot of render time, so if a scene can be set up without it, it works marvelously. There is also "Full Raytracing", which gives some quality improvement, but it is so insignificant that it's practically never worth the render speed price. So I am looking for other options I can save render time on.

Kixum here is getting pretty awesome results, so I'd like to know how long it takes and what the settings are.

Wow! This looks very impressive (And great quick modelling, by the way! Bravo!)

What were your render times for this image? (And what are your system specs? so I'd know how to approximate that for my own.)

My main problem is really in the render times. I want a classsroom full of furniture, props and human figures, fully clothed. Even if I use proxy figures for the ones near the blackboard, it's still a LOT of stuff for the light to bounce off and slow down calculations.

Is the archived forum gone for good?

This was a freebie by a Carrara artist, whose freebies were absolutely amazing. But I am terrible with names and I no longer remember his forum name. This is horrible, because I can't find the cashed versions of the archived forum threads anywhere. I think the nickname had a "5" in it, but as I said, I am terrible with names. I went back to the zip files he provided, but there aren't any readme files or notes, just the objects :(.

This is all wrong, because now I won't know how to give proper credit! :(

If you remember the person, please help me out!

It's got about 70 lights at various settings, and It's still not as natural as I'd like it to be...

Is there any way to use the Carrara calculated bounce, but reduce the quality/component settings to speed it up? Or to use a simplified proxy environment to calculate the light map, maybe? Would any of these tricks work?

(The glow plane looks black because it was actually set to 100% tranparency, but having the white glow still gave me the lighting I needed even when the panel itself didn't render at all.)

HDRI for background, but no Sky Light or Indirect Light is activated, so it has no effect on the lighting.

The animated snippets look fantastic. I hope I get to see the full short. I've heard about the project before, and I think their use of ToonPro is wonderful.

Eric, thank you again for your plug-ins. I hope and pray that you will continue to update them! (And maybe add a couple more? )

If you are just learning the program, it might be good to learn the core, but Fenric's plug-ins are great time-savers once you know what you are doing. I use his pose helpers for things like mirror poses, reset poses, reset morphs, change visibility of figure elements, etc. all the time. I do know how to do it in Carrara without them, it's just much faster to do it for figures using Fenric's utilities. So, once you get comfortable and find yourself getting annoyed at having to go an manually set visibility property* on all the 16 components of each hand, it will be time to go to Fenric for help. :)

Best of luck!

*With figures, if you want to set a sub-element of the mesh (for example, only the hand) invisible it's not enough to uncheck the "Visible" property under the name (which is there for every element of Carrara), you must also find the poser-imported property of that subgroup (sits right above morphs) and uncheck it there too. Or apply a poser pose which sets invisibility to that part, but you don't always have the right ones ready.

You might also look into Fenric's utilities. I find his pose helpers very useful for Genration 4 figures (like M4, V4,kids4, etc.)

The tree duplicator is part of this package: http://www.daz3d.com/posing-pack-3

Attached Link: Forest Wonders example

Actually, given this information, I would probably go with the free Ivy Generator + Carrara's own plant editor instead.

I've had pretty good results before when I used the plant generator to create leaves and branches and then surface multiplied those onto a modeled trunk or onto Ivy Generator-produced geometry. You can get very good diversity this way, and the bonus is that you are not paying the price in render time for the alpha maps, nor are you limited in how the resulted conglomerate is rotated, allowing for easy randomized replication for a complex landscape. 

I've used this method in the attached image (see link), and replaced the transmapped branches/leaves with Carrara-plant-editor generated ones. (I left a few behind: some Ivy and grass that came with the Stonemason's set are still there and transmapped, but there are only a few of them left, and if I had to push the image further, I'd replace them too)