Current Residence: Kingston, Ontario
Favourite Band/group: Ace Of Angels
Operating System: Win 10 Pro & SolydX
Music player of choice: Poweramp for Android
Personal Quote: Whatever
Disclaimer (of sorts)
This was created with the purpose to help you put together large scenes in Daz Studio and render them in Iray. It is not a stand alone method but simply a guide, to be used in conjunction with other known methods to getting the most out of Daz Studio and Iray. This is not a quick and easy solution and its one that you have to figure out from scene to scene as needed. Hopefully the information shared here will be of use to those who read it and thus, can put it into practice.
All the products created by the different published artists and vendors across the different store fronts are created to look there best, as the end user only want the best. However, with the limited amount of video ram (VRAM) on a video card, it is difficult to fit more then a few high end items in a scene and have it render in Iray.
So why is that? What is the issue? It is the textures! Out of everything in your scene, the textures on the model will take up the most about of the vram. What we don’t often realize is just how many textures are used on each model to get the final effect. There is the diffuse map, spec map, translucency, bump, displacement, normal, opacity and more. Also, the size of the texture used is a factor as well. A 1000x1000 image map is ¼ the size of a 2000x2000 image map and therefore uses only a quarter of the vram compared to the larger map. Another texture issue is how many surfaces on the model use that texture. For every surface on the model that uses a texture, its gets loaded in to the vram just for that surface. So you kind of get the picture as to why the vram usage can jump up so high and even be impossible to render with any more then just a few items in the scene, especially if you have an older video card.
So what is the solution? It may not be obvious cause its rather simple and to be honest, one that did not occur to me until recently. The solution is to remove textures that you don’t need! Or replace them with smaller tiled textures that can do the same or similar job. All those textures are there to make the product looks its best when rendered up close. However, beyond a certain distance, their effect is not noticeable. The bump and spec maps may be needed to render a character up close but move the character back a few more feet and you start to loose the effects of both. Also, if you use depth of field when rendering or a z-depth mask in post to get the same effect, you are going to blur out even more details.
Now this is where it can get time consuming and where we can make mistakes. Deciding on what texture maps to keep and which ones we can get rid of or replace requires us to go over each model surface and see whats being used, where and what happens when you take it out. Most of the time when I removed a map related to translucency, spec, gloss and top coat, I had to decrease the strength of each effect to a minimum or just shut it off. But it depended on how far the model was from the camera as to what I did.
This image was rendered in Iray - mattymanx.deviantart.com/art/T…
Lets break down the scene. There are 18 Genesis 2 Female characters, all fully clothed and with hair, plus the scene itself. Between all of them, the textures only used 2.98GB of vram. The scene only has textures on surfaces that are visible to the camera. The carpet and two pillars on the back wall use textures from shader sets. All other textures were removed and a solid colour was used instead. The scene was lit using only the mesh lights in the scene.
The nine girls across the back top section of the scene have no textures at all and neither do their clothes. In fact, the only image maps used are the translucency maps for the hair. That was a whole 300mb of vram for the nine of them combined.
As for the nine girls on the stairs, the two facing away from the camera have no textures on them either. Their clothes do and so does the hair. The three girls on the left of the coming down and the three on the right sitting, each have a basic setup of skin textures. While the clothes and the hair are the same as the other two just mentioned. The girl in the center has the most textures used but even still, hers are still reduced to a minimum.
So what did I take out? The girl in the front center is the only one with eye textures and only the diffuse map is used on the iris. The other eye surfaces are just a simple single colour. No other girl in the scene is looking in the direction of the camera and you cannot see their eyes at all. So instead of using any maps at all for the other 17 girls, they all just had basic colours used for those surfaces as well. Same goes for their teeth, tongues, gums and inner mouth. Also, all the maps for the nipples, toenails and fingernails were removed. Toenails and fingernails were replaced by glossy car paint shaders and the nipples are just a solid colour. Just a basic colour is used.
As for their clothes, the original textures were replaced by textures from different shader sets. I did not use the same pair of nylons that came with the school girl clothes but instead replaced it with a different set that only has one surface. That way the map used is only loaded once per pair in the scene and only on the nine on the stairs. Cell phones are all texture-less as well.
I admit I could have gone farther and reduce it even more but I did not. It only took 2 hours for it to reach 22800+ samples. That’s with quality on and set to 7 with convergence at set to 99%. It was originally set to quality 3 and only took 46 min to render but it had too much grain in the depth of field area for my liking. Time wise, it is what I would expect for a scene lit only with mesh lights. I have not tested this with CPU only rendering so I do not know how it will help if you use CPU only.