Animated TV advert problems

[hilby]

Hi everyone! First time posting so I'll be quick, basically we're making a few nationwide animated TV adverts in the UK for a window company called Coral Windows, we decided to use Octane due to it's blazing fast results and instant feedback. We have a few questions though that hopefully the community can help with

1. The Client wants a christmas advert doing but also a new years advert. the new years advert is basically the same as the xmas one minus the christmas lights and decorations. Is there any way to render the christmas lights/wire on a seperate layer so that we can remove them when new year comes around? The lights are emissive materials but do not cast illumination, so we're using camera effects to achieve the glow. We are also using a daylight system to add some 'moon' ambience.

2. Any tips on how to speed up render times? In all fairness, apart from the street lights the scene is very basic and highly optimized in terms of poly count, apart from the snow which was achieved with an 8K disaplacement map. We've only been using Octane for the past month with no previous experience, it's very easy to pick up, although we don't have a full understanding of all the options. The lighting seems to be quicker in some instances after playing around with the sampling rate, what does this mean exactly?



Here's a render we did earlier, it took about 3 to 4 minutes for 2000 samples at 1920x1080 on dual Titan x. what we want to know is why it would take this long when we're using the most basic render settings. For example we're rendering using the direct light kernel, in ambient occlusion mode, with as low a depth as we could get away with on diffuse, glossy and specular. We've also used as few glossy materials as possible and no specular materials which should have sped it up as well.

Any help would be very appreciated

[Moonhowl]

Hi,

my guess is that the emissive materials on the lights are causing your render times to be relatively high. I have read somewhere before that using the same emissive material on say, 5 different objects, will treat those 5 objects as 5 "separate" emissive materials thus shooting up your render times, i.e. you will have 5 separate sampling calculations for those 5 individual objects, even if you are using the same material for all of them.

But on the other hand, if you have those 5 objects attached into a single object (say you have 5 spheres that are attached into one object in 3ds max), you will notice that you will have to crank up the emissive material strength in order to get the same light intensity compared to those 5 spheres being individual objects - only "one big light" would be sampled in that case though rather than 5 individual ones.

Though since you said that your emissive material is not casting any light, I'm not sure how that would affect it, but I'm guessing it might as well.

Second thing to consider is the overall size of the scene. At least with the path tracing kernel (not sure about AO since I never use it), if you have for example a camera in a room in a house, and you have the whole house set up for rendering (lights in other rooms), octane will be calculating light sampling for the WHOLE HOUSE and all the lights in every room, even though you are just seeing that one room that has 5 lights out of the 20 throughout the house. Once I figured that out I had a scene I was working on drop from 7 hours to half an hour after getting rid of the geometry and the lights that the camera did not see.

Apart from optimizing the kernel settings, these were the main "speed killers" for me when I was learning Octane so hopefully this helps.

[Goldisart]

Will challenge. The light in the other room does not respond to a miscalculation if its rays do not fall into the camera ... use GI clamp and optimise the amount of Diffuse depth ///

[Moonhowl]

goldistar well, that's what happens during my tests lol. Quite literally. I had a house set up. Took 7 hr to render. I delete everything else apart from the room I'm doing, it's 45 minutes.

I move to another camera on another floor. Takes exactly the same time as the camera in the room below. I deleted everything else again and again it takes under one hour.

I'm doing stills not animation, so there is no geometry reloading or anything.

That's the only explanation I have, but I'm open to a better one, either way it seems to make a huge difference this way.

[coilbook]

we have 12 780TIs and rendering bright scene 1080p in daylight with just one sun at 1500 samples takes us 2-3 minutes per frame . About 60% of all render time is eaten up by trees and grass

I think diffuse depth does not work when you use AO

[coilbook]

Hi,


Second thing to consider is the overall size of the scene. At least with the path tracing kernel (not sure about AO since I never use it), if you have for example a camera in a room in a house, and you have the whole house set up for rendering (lights in other rooms), octane will be calculating light sampling for the WHOLE HOUSE and all the lights in every room, even though you are just seeing that one room that has 5 lights out of the 20 throughout the house. Once I figured that out I had a scene I was working on drop from 7 hours to half an hour after getting rid of the geometry and the lights that the camera did not see.

Apart from optimizing the kernel settings, these were the main "speed killers" for me when I was learning Octane so hopefully this helps.

thanks for the tip about lights that are not shown but still using the resources. I used to think only what camera sees that's what gets calculated because I get big slowdowns if 40% of the screen is grass in the camera view. and if i just render sky it takes seconds.

[coilbook]

we have 12 780TIs and rendering bright scene 1080p in daylight with just one sun at 1500 samples takes us 2-3 minutes per frame . About 60% of all render time is eaten up by trees and grass. Try to disable alphashadows

I think diffuse depth does not work when you use AO

[Moonhowl]

thanks for the tip about lights that are not shown but still using the resources. I used to think only what camera sees that's what gets calculated because I get big slowdowns if 40% of the screen is grass in the camera view. and if i just render sky it takes seconds.

My guess is Octane has to deal with a lot of struggling rays if you have lights that even don't affect your scene directly, or are even far enough that they don't really contribute to your shot much or at all. I don't know exactly why this is happening but based on the house example I mentioned somewhere above it made a massive difference for me. And GI Clamp set to 1 didn't help past a certain point either, while deleting everything else apart from the room I needed for the shot, well, the render time literally went down by a factor of 5.

Grass and similar intricate geometry will always slow down render times massively, regardless of the render engine used. In such cases it's as much a matter of drawing out the multitudes of tiny shadows and computing anti-aliasing for all of that geometry as it is the actual light samples required. Raising the "AA" in octane (the "Filter" value in kernel settings) might help get a good result for such geometry quicker, at the expense of sharpness of the image.