- This topic has 8 replies, 8 voices, and was last updated 2 years, 1 month ago by
.
-
Posts
-
I try to really getting better at realistic renders. However its hard to find the perfect Ratio between Workload and results.
I always go for Gen9 (Hd) and 4k Resolution.
8k would be a little bit higher Q but takes 2-3longer at least for me...I watched a whole lot of Videos, but maybe someone here has a really nice and round setting?:)
theres no ''one setting'' that will fit all your scenes and needs. It can depends on the lightning and how far or close up your characters are. Not to mention that ''looking good'' is something subjective and will vary from person to person
Here are some of my settings.
Auto Headlamp: Never
Render Quality: Off ( No reason to have this on)
Max Time: 97200 to 259000 (doesn't matter so long as its a long time)
Max Samples: 5000 or less.( Ive never needed more than 5000 and you can probably get away with a less depending on whatever it is)
Pixel Filter Radius: 1.0 or less Cant remember ever using less than .9
Pixel filter: Usually Mitchell but sometimes Gaussian. No idea if it really matters.
Sometimes Ill try Caustic Sampler but rarely.
Ground Position Mode: Manual
Environment Map: https://polyhaven.com/hdris almost always
Lighting resolution: 1024 - 2048 not sure if it matters
That probably 99% of what I always do in every render
Dont see any reason to use 8k textures unless the camera is 1 foot from the subject.
With AI now being able to remove noise and upscale things effectively I dont think render size matters much anymore.
@gelstat
Always been like that as far as I can remember but I have a bad memoryMy advice for render settings would be to render tons of test images and vary settings one at a time. As said above render settings will vary dramatically depending on scene content; rendering a shot filled with mirrors will take far more iterations to converge than a scene of a single character against a matte fabric backdrop.
The number of iterations you'll need to converge an image will depend on A) the scene amount of scene content, B) the shaders being used, C) your GPU. Even comparing core for core an older 10xx or 20xx series will require more iterations due to older compute cores which are less efficient. Test your scene to see what is good enough. Experiment with the denoise filter to see what works for your scene(s).
Render to the size *you specifically need* for a project. Rendering everything at 8K is probably overkill. For wallpapers rendering to 4K is nice, but most 2K renders can be upscaled adequately to make this unnecessary. Unless you're going to print rendering to 2K is usually enough for portraits and square aspect. Also, take full advantage of canvases and spot rendering to speed up complex scenes. With canvases you can render a depth map pass and do your DoF in post which saves an enormous amount of time (and will give you more realistic bokeh and light drop off than Iray can do).
Experiment! That is the key to learning new techniques and finding what works best for your unique workflow. The more you experiment and begin to grasp what various settings do then the better you'll know what specific questions to ask when you have a particular issue. Just asking, "what render settings?" with no scene (or hardware) info is too open-ended to answer specifically.
Same Adam. Pretty much the DS default settings with a few tweaks. For outdoor renders using an HDRI for lighting, increasing the dome radius can improve the lighting.
Seems to be like most settings. People leave it on default. I messed with it awhile ago. Can't recall any difference.
I wss reading about ways to speed up renders and settings. Supposedly it's better to just render to a file. I lmost never compare times so I've no idea if what I do makes a difference.Depends on the scene.
Most of the time it's 2000 pixels by 1500 pixels or Full HD, if I'm making wallpapers for myself.
Max time - 7200 (2 hours)
Render Quality - 1.00
Rendering CR - .95I'll also render a background to save time separately. Other things depend on the scene (such as environment maps and lighting)
A lot depends on what kind of image you're making, and what kind of hardware you got.
I tend to play around in and with cartoon figures and renders, so for me, the Post Denoiser feature is the big setting that gets my attention and refinements the most.
De-noising is great for making an image smooth very fast, but it blends and 'smears' colors together (which is perfect for cartoon renders which is why I love it, but lousy for photoreal pics.) The default "Post Denoiser Start Iteration" is at the 8th render iteration, but I like to generally bump that slider to the 450th-500th iteration or so, just to get a solid range of colors and imperfections blended into the toonish results, but that's just me.
Now if you want to do photorealistic pictures, you can instead crank that sucker up to 1200, 2400, or higher - at that point it becomes a catch-all to compensate for crap lighting. Either way, you have to adjust it to the lighting - the brighter the picture, the lower the Start Iteration setting, and the darker/dimmer the light, the higher you have to go, so you can collect enough iterations to work with.
(Note - I leave the "Denoise Alpha" setting firmly off, because toon.
Finally, let's talk about hardware. low-end GPUs (like a GeForce 10xx RTX series) needs more aggressive settings (or a far greater tolerance for time settings like "Max time (secs)") than newer, high-end GPUs. Low-end GPUs can only handle simper scenes, so if your GPU only has 4GB of vRAM, you're stuck with scenes smaller than that (minus overhead too!) - otherwise you're going to CPU rendering and it will take forever.
- You must be logged in to reply to this topic.
