Hello
Is there any quality limit in the render configuration?
for example : how many Samples can I configure for maximum quality ? what is the limit?
and in the case of the other settings to render ? is there a total limit ?
thanks
Hello, @Beeg! I’m not 100% sure if I understand the question. With the progressive rendering you get with Cycles, you are mostly up against a “law of diminishing returns” situation with render quality. The longer you let it render/the more samples you use, the closer the renderer’s approximation to how light would physically behave in the real world…But there is no “maximum quality” setting. You have to decide for yourself what aspects of the render you “tune” to achieve the quality level you want.
The Adaptive Sampling settings under the Render > Sampling panel allows you to control certain aspects of your render to get the quality you want. The Noise Threshold and Adaptive Sampling features allow you to determine the maximum amount of noise your render has - that allows Cycles to spend samples on areas that have more noise instead of areas without noise so that there’s an overall user-controlled maximum amount of noise.
Take a look at the Sampling section of the manual for more info
https://docs.blender.org/manual/en/latest/render/cycles/render_settings/sampling.html#sampling
1 Like
The limit is more the time you want to allow for the render to refine.
You can set a crazy amount of sample and let blender calculate that image for days.
The trick here is to find a good balance so you get a good quality within an acceptable render time.
There is no general rules, needed samples will change depending on how the scene is lit.
For some scene something like 200 samples will be enough, other may requires much more.
For other settings there might be some limits, it’s worth knowing what you’re doing here, for instance the diffuse and glossy bounces : you can enter crazy values there but that might just make the render slower with very few value added to the render.
My advice is to look for video talking about how to make cycles faster, and look at these settings and see how higher value influence the result.
How should I configure to have the maximum possible sharpness?
thanks
is it appropriate to stop the render while it is being done?
can i use this result as final image?
Is the image of acceptable quality?
or do you have to wait until the blender finishes to have an adequate quality?
thanks
Hum, first , let’s make sure we talk about the same things, are you using Cycles render or Eevee ?
Now to answer your question, yes you can stop a render and use this image.
But in all case there is no real guaranty of “quality” in the render. It’s up to the artist to control the result and adjust settings accordingly if the result lack in quality.
Blender’s defaults are made so you have better chances to get a good result, but if it was easy to be sure of the end result’s quality we’ll have only a “enable good quality” button.
So what defines the quality of a 3D render ?
The most important is getting rid of the noise, if you work with static images you can use the denoiser and even with a few samples you should get good results. In animation it’s a bit more tricky.
There are other parameters that can help render faster but sometimes influence the result.
Try the default settings and enable denoiser , that should give you a good start.
Sharpness it’s a bit different, if you really want a sharp image then it’s more compositing tricks that can help.
Best is to post an image so we are sure we talk about the same things and we can advice you correctly !
Good luck !
Most of time, your limit is “time”.
I don’t use higher samples than 65536, I adjust using noise threshold. Maximum bounces are 1024 but it is better to keep bounces minimum where it doesn’t make visible change to image quality because adding bounces easily add more noise and it is possible that you don’t have enough samples to have noiseless image.
Other limit is memory where you need to fit your assets.
No denoise, using procedural textures, 65536 samples, resolution 21600x14400 works for 36" x 24" poster at 600ppi.
But of course you can have more resolution and samples. Also you can let your computer to render image long time, like a year (there are people who are rendered single image for year).
I really don’t get your question. You always can have supercomputer to calculate one image, for year. You should ask what kind of quality you need. Usually people want their renders to be finished fast as possible, while having their quality requirements.
is there a way to stop the render and reprocess it later?
the noise reducer : is it possible to make the render for only after processing it ?
does the use of layers make the render slower?
thanks
You can post process frame later in composite, but you can’t stop rendering frame and continue later.
In animation you can export your frames one at time and start rendering again from frame you want.
Sure, in composite. Just keep denoising data.
Not necessary. They can speed up rendering and often layers are required when working different scales or need to fit assets in memory.
Layers can make image look more artificial when light interaction between layers is missing so when aiming quality, you don’t want to have too many layers but often you can’t avoid layers.
Hello
Thank you for your help
how can i save cryptomatte’s node?
can it be in .tiff ?
is it possible to save each color of crytomatte in different layers to open in photoshop?
thank you again
I’ve never exported cryptomatte and I don’t have Photoshop.
“Quality” is very much “in the eye of the beholder.” Or, as the case may be, in the actual hardware that the end-user will use to view the final result. And then, exactly how closely he is actually going to look at it.
On a still image, say that you intend to mount on a museum or art-shop wall, the finest attention to the most subtle characteristics might be justified. Whereas, in video, there are plenty of shots where you might be able to get away with murder.
For example, Star Wars Episode One actually shipped to theaters where one passing “crowd shot” in a podracer scene was actually occupied by painted Q-Tips cotton swabs. (Of course the shot has since been replaced.) The short-on-time filmmakers knew that, having already recently seen a fully-populated “crowd in the bleachers,” the audience would “see” the same thing again. And, we all did, until we saw the “making-of documentary” confession.
It’s a judgment call where “the law of diminshing returns” is important, as is the phrase “good enough.” Only you can decide where “enough” is, and it definitely varies from shot to shot.
1 Like
P.S. I strongly recommend that you should render into the (one frame per file …) OpenEXR or, as the case may be, MultiLayer OpenEXR file formats.
The OpenEXR file formats are specifically designed to capture data. As in, “great big files of floating-point numbers.” So that you can carry these numbers precisely forward all the way through your pipeline, right up to and including the “final cut.”
At which time you can then independently address the particular physical needs of each of the “deliverable files” – JPGs, PNGs, “movies of various kinds.” Considering each one according to the requirements and peculiarities of their software and hardware, compression and file-size considerations, and so on. While drawing from a single “pristine source.” It is only at this final step do you deal with things like alpha, compression, and other things that are specific to the characteristics of that display device (or devices).
is it possible to render without pass : combined ?
I turn off the combined and it still appears in the render ?
thanks
I believe it will always do the combined pass. I’m not sure if the checkbox in the layers panel does anything…
My experience is that you’re always gonna get the “combined” pass. You don’t have to do anything with it … but it will be there.
using option : 2D or 1D in texture nodes , example : musgrave , saves processing in render ? and viewport ?
and select which render I’m going to use, eevee or cycles, in node material output?
thanks
Are you using cycles or Eevee?
both
both cycles and Eevee
Then set that render output node to both.
As to 1d or 2d textures being faster than 3d or 4d, probably. But if you need a 3d texture a 2d one won’t work. If you need a 2d texture, a 1d one won’t work.
They have different uses. If your shader needs it, it needs it.