stallion opened this issue on Jan 29, 2019 ยท 1258 posts
Blackhearted posted Mon, 11 February 2019 at 6:14 AM
llynara posted at 6:13AM Mon, 11 February 2019 - #4345805
Haven't had a chance to catch up on this thread, but here's a test I did, with postwork. Still struggling with grain in Superfly. La Femme is gorgeous and worth the effort!
Very nice.
You can get around the grain/slow rendertime by reducing the amount of unnecessary samples in your render.
Keep in mind I am not a technical person, so if anyone wants to correct me or add info by all means, I'm all ears. I achieve results through sheer stubbornness and patience, so my way may not be the best way but this is how I rendered all of my promos. This has no transparency, transmission, etc. SSS and emissives seem to be controlled by the diffuse samples so you will get those. Eyelashes and refract shader eyes will render black. I render my scene like that then bump up settings and re-render the eyes/lashes with high transparent/transmission bounces and caustics turned on using Area Render to draw a small box around the eyes and the immediately adjacent area they will affect with bounced light/shadow.
This brings me to the hair. Even a midrange PC has a lot of cores/processing power available these days - not to mention modern GPUs. However people have been spoiled by making hair for biased (ie: toy) render engines like Firefly and we've reached a point where any sort of optimization has gone out the window. We're stripping figures down to ~20k poly base meshes and then loading ~300k poly transmapped hair with a half dozen 4k maps and thousands of individual transparent strips stacked 20+ layers deep. This is not ideal. After 20 years of development there ideally should be a working strand-based hair engine in Poser, but failing that there should be some attention paid to optimizing hair for a good balance of realism and rendering speed.
In the meantime you can get around this by rendering the scene with the settings above and then going back and Area-Rendering hair with about 6 max transparency bounces (if necessary going back and spot re-rendering any spots that may still remain dark). This is still much much faster than simply turning up all render settings and taking 10 hours to render your entire scene.
But people have been spoiled by using biased rendering engines for years. Remember the mass protests when Poser added Firefly and D|S added Iray, while people were still learning to use them. The Poser 4 rendering engine was an absolute POS, and the only reason it was fast was because it was wildly inaccurate, and both the P4 renderer and Firefly further sped up by 'cheating' due to Poser's small scale: items in the scene were so microscopically tiny that the renderers were discarding a lot of scene information and thus rendering faster. For years I used to squeeze better renders out of them by scaling everything in my scene up to 1000%, with a significant boost in shadow quality but also in rendertime. But quality renders take time -- a high quality ~2000x1500-ish render always took me between 2-10 hours whether it was in Firefly, Iray or Superfly. Good results take time, there is no instant gratification when it comes to indirect lighting.
I'll typically do a few dozen small, grainy preview renders while I tweak lights and pose, use area render to check areas like the eye focus/highlights, feet/hands properly in contact, etc, and then just leave the final render going while I go do something else, watch a movie or sleep. In the final render if something is amiss like a tiny spot of pokethrough I didn't notice in the preview I can fix it and spot-render that little area in a minute.