Making Really Big Scenes? Try this.
areg5
Posts: 617
Big scenes look great, but they can be a pain in the ass to set up, pose, and fit into your VRAM. I make comics, and try to crank out 15-20 pages/day. But some of those scenes are big, and the VRAM can be overwhelmed. Not only that, but if you have many objects in the scene, and characters, they can take forever to pose! Here is what I have learned in making really big scenes. My VRAM is 11 gig, and to those of you with VRAM’s of 6 or 8, that might seem like a lot, but no matter how big you go with VRAM, you will eventually fill it up. Here are some things you can do.
- Optimize the environment. What I do on every scene is make the set just how I like it, including lighting. Then I render, and use GPU Z to see how much VRAM is used before adding characters to it. I like to keep the usage for just the environment to be no bigger than 2-3 gig, or I get limited in the number of characters I can add to it. So if the usage is too high, I run a single pass with Scene Optimizer. I have not seen a drop off in render quality doing one pass, but you may see some issues with more than one.
- Assemble the cast. In my current commission, there are 13 characters so far. I can tell you that I think it likely that just the characters, clothed, would likely overwhelm the VRAM. Of course, that’s not to say that every character needs to be in the scene, but if you are planning a bunch of multi character scenes, you may want to take the cast, with whatever they are wearing, with hair, and run scene optimizer once. Again, I have seen no drop off in quality.
- What if the optimized characters are still too much for the VRAM. My next step is to take everything the characters are wearing, and their hair, and change the resolution to base. If that still is too much a load, I start turning the character resolution to base. This is the point where you might see some issues. Base resolution in a character seems to primarily affect ears, fingers and toes. If I have a character whose ears aren’t showing, or their fingers or toes, I can safely toggle the resolution to base with none of my readers saying “that ear looks funny.”
- Save the characters as subsets. You can do this individually, or if you have a few that are always hanging around, save groups.
- Create a camera for the scene.
Ok, now you’re all set. But let’s say you have a ton of objects in the scene, like landscaping, gardens, piles of books and things. You’ll notice that posing takes forever, and the camera jerks around when you try to rotate and translate it. It’s just a hassle. You can sometimes alleviate this by choosing solid bounding box as a view. Sometimes not, as when you get to fine details like expressions and eye position, it’s once again very slow. Here’s how I get around that:
- Load the scene with at least one character in it, where you want that character to be in the scene. You don’t really need to pose it at this point. Save that character as a subset, which if you followed the steps above, will copy the pose. Again, it really just needs the x,y and z axes to be where you want them.
- Open a second instance of Daz. Load the subset of the character from above.
- In the second instance, load the other characters in the scene, or even in the entire project. All of them. Don’t worry about VRAM in this second instance, you won’t be rendering it.
- Create the scene in the second instance with the characters in that scene. If you are using say 4 characters in this part, pose them how you want. Because the first character is spatially where it has to be, if you pose the characters in the scene around the first one, the relationships stay the same.
- Create a camera in the second instance and put it where you want. In other words, this is where what you will be looking at is exactly what you want the render to look like, only it won’t have an environment or lighting.
- Go back to the first instance and load the characters you want in the scene. I.e. all of the same 4 characters that you set up in instance 2. Don’t worry about posing them. Don’t worry about the camera.
- Go back to the second instance. Save the pose of each character and the camera. Use an easy system. I basically start on the left, save a character pose as “a,” then the next one as “b” and so on. The camera is the last pose I save,
- Go back to the first instance and merge each character and the camera with the appropriate pose. When you are done, if you hover over each instance of daz the preview will look identical, the only difference being that in the second instance there’s no environment.
- Save both instances. Instance 1 I save as “1” in this example. The second instance I would save as “characters.” The second instance is now your working pose instance.
If you optimized everything, the scene should render if it fits into whatever VRAM you have.If it doesn’t, you may need to adjust the resolution further on some of them.But for now, let’s say it renders in your system without a hitch.
Now you go to the next scene.In instance 2, pose the characters.Let’s say in this part 2 more characters appear.Put them where you want them and pose them, and add those characters to the first instance as well. You don’t need to pose them in the first instance. Pose the camera in the second instance.Are some of the characters not in this scene off screen?Save their subsets, either individually or in groups. Then delete them from the first instance, to save VRAM. Don’t delete them in the second instance. That way when you add them back to the first instance, their location and pose will be unchanged. When the scene is how you want it in the second instance, save all of the poses of the characters and camera.Merge the poses in the first instance.Save both instances.The first instance I would now call “2. “ You don’t need to change the name in the second instance.
As you proceed you will notice that the scene is really filling up in the second instance.That’s fine.Those characters are there and ready for further posing, and they will be starting from previous poses which in my opinion adds continuity to a comic.Makes it seem less like characters are popping in out of nowhere.
So that’s about it.Big scenes are a lot of work, but I have found that this method speeds things up considerably.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
Hey quick question somewhat unrelated to your post. Do you recommend the Thermaltake View 71 Tg Full tower? You have it listed in youir system specs. Building a new system over the next few months and was looking at this case. Thoughts? If best we can Pm so as not to derail the thread. Thanks for your insights.
Actually, when I was looking for a new case I wanted another Coolermaster but they were no longer available. At first it took some getting used to, but the more I work with it the more I like it. It is big and user friendly. Now, I'm no youngster, so to me a computer is always a box. All of the guts hidden and whatnot. So it also took some getting used to for all of the lights. With the lighted fans in the front (lights can be turned off), and the video cards and RAM lights. My god, it can just sit there and it's a light show! Very well ventilated. It comes with a vertical mouont for a GPU, which I just started using and it's great for multi card set ups. Really keeps temps down. This weekend I'm putting in 3 more fans, two in the base and one up top. I recommend it.
I create a groups called Sets, Lights, Cameras, and Actors and then in turn create subgroups that are parented to those main groups as appropriate. After that's done it's as easy as turning off and on the eye visibility iconm in the scenes tab so quickly pose the characters, props, sets, lights, cameras, and so on. I hardly ever make a really huge scene but have done it on occasion.
@areg5,
Thanks for taking the time to share this. Some of your tricks I've figured out as well, many more will be used shortly.
I hope others with similar tricks come out of the woodwork.
Pre-rendering spherical backdrops of the large scenes you describe, then loading them as skydomes/backdrops in concert with shadow-catchers and only the nearby props/scene elements, etc. really helps when doing close-up shots in animations. The IBL-master 3DL/Iray syncing product is also useful when melding the various render-engine strengths and effects when rendering layers for post-compositing too.
cheers,
--ms
You're welcome. I do like to use a lot of characters, and it took me quite some time to figure out how not to spend all day posing a complex scene. I haven't heard of the IBL-master 3DL/Iray syncing product. When do you use that, how do you use it, and does it cut time?
I wonder what amount of effect of reducing texture map resolution to the minimum required in the scene(s) has on VRAM. A lot of texture maps are 4096x4096 but don't need to be if the output resolution is only 1920x1080, etc., isn't that correct? There is a setting under Render Settings for setting the maximum textures size. Does lowering this reduce the amount of VRAM required?
Additional ideas I'm curious about but never measured:
-- Using smaller texture maps instead of just lowering the max texture size.
-- Removing displacement maps, or setting the displacement subdivision lower.
-- Using simpler shaders as opposed to the full shader (e.g. Uber Shader)?
-- Using spotlights versus emitters versus environment lighting?
Do you have insight as to how these options impact scenes VRAM usage?
The product is here in the DAZ store (IBL-Master), and I like it as a way to coordinate and syncronize scenes in 3DL/Iray mix/composite separately rendered characters against a generally static background.
... So, it's the technique's paradigm that really matters in my comment - standard stuff - kind of cheating it all with billboarding on steroids...
... And, the real win is when you're animating and rendering many images - the billboarding will save you hours. For a single render, this isn't at all recommended - but for a series of co-located static shots, it may still buy you a win - e.g. a multi-angle fight scene in a traffic intersection in a city, or multiple views of a conversation in a cathedral/room - anywhere multiple shots/views of the same scene are being generated (outer-space scenes too) - stills or animations.
The big idea is to billboard the huge expensive backgrounds and then integrate pose and/or animate the characters in that context - matching lights and using shadow-catchers and other 'tricks' to re-inforce the illusion that the billboards are dimensional (Both 3DL's pwCatch and Iray's HDRI Toolkit help with this - for the unfamiliar, take a look at their product images to get a flavor for the cleverness of the trickery.)
So - if you take static renders of your expensive/expansive cities/spacestations and wrap a new scene with these hi-rez single renders, then integrate your characters... the render-savings are potentially huge. (render at a lower-rez and soften the background with some camera Depth-Of-Field to cheat the clock even more...)
The PA 'Dreamlight' has a long history of tutorials and using this technique and they've recently been offering some HDRI sets that demonstrate this idea very well (I don't have any, but others might have comments), and the idea is hardly new, but very useful when memory/time resources get tight - especially when re-rendering the massive sets you are describing becomes impractical. Magix-101, FlipMode, and Colm Jackson's TerraDome also use these kind of billboard worlds that leverage this alternative to actually creating these mega-worlds as you apparently do so well... :)
I hope this information isn't so obvious as to be a bother. I figure future readers of this helpful thread might be glad for the information as a compliment to your advice and approach - I think the ideas can work well together in many situations.
cheers,
--ms
I'm working on a scene with three of my buildings and characters
---
I know the characters are a couple gs of ram just to load
---
so once I have the big scene nailed I'll render the street with just the buildings
---
the render each building by inself with it's own set of characters
----
and other options would be doing the buildings only
doing the characters only
and then doing the building fronts
---
so place buildings in back.
add characters on top of that
add building fronts saved as tiff with transparent windows on the very top
I'll stick this first one in front of another rendered background of taller buildings
then do the separate buildings with characters etc.
of course making sure the camera is locked in place and the lighting stays the same.
---
In the second one you can see where I've been adding detail to the interior of the house
so I'll do that and then put them all together after each building is updated and rendered with characters etc in it
---
the third one did use 11880 mb on the video card.
And with out any characters in the scene.
---
the fourth one I've used as backdrop behind other partial fronts
---
the fifth one is the firs one with the fourth one behind it.
I have sort of used that technique in the past, when I had less hardware resources. I had no idea they had a product that helps do it. I may look into it. Thanks!
I love it! Looks great!
nifty tips. thanks!
Sure thing!