All posts related to V2
User avatar
By tom
#364408
gotoxy av-media wrote:i thought by posting those issues it would be helpful for NL to improve their render-engine!
Could you please render the same scene with an improved engine and post the results to see what kind of improvement we need here?
gotoxy av-media wrote:i have found some more issues, which i would like to post in this forum...but i won`t, due to a lack of response...
Of course, you're free to do that. It's already our part to improve it and not your obligation.
By zdeno
#364418
gotoxy av-media wrote:....

i have found some more issues, which i would like to post in this forum...but i won`t, due to a lack of response...

regards
DECOY !

nice try ! respect !
#364432
first of all thank you for the feedback!

@Dario Lanza

What is the reason for NOT having the same AA as the alpha-channel?

@tom

There is no need to render with other engines, this is no competition!
It`s all about maxwellrender and how to get the best results by understanding how things work.
Take a look at the top of page 1 and here are my questions:

...would you please tell me what i did wrong?
...how to get useful mat/obj-id that doesn`t differ in AA in comparison to the Alpha-channel?!
...or what is the intention of these channels? ...and how to use them correctly?

@bubbaloo

what is the purpose of mat/obj-id?
By bograt
#364433
Vray had the same issues, as frustrating as it is material ID has never been great for making masks.
I think Bubbaloo highlighted what may be the problem; AA would surely blend the different hues rather than the luminosity.. maybe its time someone developed a 'multi-alpha'
#364436
@bubbaloo

yes, i use it for modifications in post, too!
It is limited for very thin lines...(AA)

and

yes, i had to render several alpha-passes for the different objects, which is very time/energy-consuming when you need an animation sequence!

...

@bograt

multi-alpha sounds nice!
just like the shadow-pass, which saves separate files for each light, but for mxms (mat-id) and for objects (obj-id)
User avatar
By tom
#364438
Unlike Alpha Channel, Id Channels are not receiving the same implementation of pixel subsampling driven by the sharpness parameter. So, currently the antialiasing coming from Id Channels cannot perfectly match Alpha Channel. Naturally the difference is more evident in thin lines where antialiasing is heavily visible. In other words, Id Channels are not tied to Sharpness. Maybe I can make a suggestion if you tell more about your scene, aim and workflow. Rendering a grid of thin mesh and complaining about "Id coverage of this object is not matching with alpha coverage ruins my work." sounds senseless.
#364441
Hmmm, I havent read the whole thread but I would like to add something constructive on the general subject.

- Use matt id, not object id, because it renders clean by sl6, theres no reason to render higher as far as I can tell.
- Render matt id's for stills after the main render, because its better to concentrate on the main render when we press 'render'.
- Use pure R,G,B for matt id's to avoid colour bleed between id's, so its generating 3 id's per render, (theyre very quick renders - sl6). So therefore its neccesary to set off extra renders to get extra masks, I usually need 2 matt id passes.
- Use Photoshop with gausian blur 0.3...!

And then they look like this;

Image

The shift lens is making the thin lines vertical ofcourse which helps.... In some 'wireframe' situations an extra alpha pass would be needed as has been mentioned.

Hope this info helps someone... Just wondering why object ID renders slowly, does that have full AA? is it less slow if rendered with the main render? probably...
#364476
Unlike Alpha Channel, Id Channels are not receiving the same implementation of pixel subsampling driven by the sharpness parameter. So, currently the antialiasing coming from Id Channels cannot perfectly match Alpha Channel. Naturally the difference is more evident in thin lines where antialiasing is heavily visible. In other words, Id Channels are not tied to Sharpness.
well summarized!
so, if i don`t touch the sharpness parameter (which i did!) - i would expect the same AA in Alpha and ID channels...
Rendering a grid of thin mesh and complaining about "Id coverage of this object is not matching with alpha coverage ruins my work." sounds senseless.
hmm, i don`t know what you are talking about?!
Hmmm, I havent read the whole thread but I would like to add something constructive on the general subject.
:lol: thank you!
By renbry
#364958
Thanks for the explaination Tom,

I might have problem in the misunderstanding that a pixel (think a bucket of information) shouldn't have pixel-filter-matched sub channels (AOVs) There might only be a few exceptions to not having pixel-filter-matched sub channels such as integer-based AOVs for diagnostic purposes (cpuID for a distributed bucket renderer)

I would always expect to have 'samples aligned' whenever i'd like to have an R,G,B or visibility A, or materialID, or objectID etc etc. Those values must sit in the same 'bucket' and be sample aligned; whether i'm rendering a wire grid or not...

Vray has AOV controls for 'filtering enabled' which lets the user not filter the pass in the case of z-depth or velocity.

Please make sure the samples are filtered and aligned from the beauty, through the AOVs otherwise they're fundamentally useless.

So, after something like three weeks of evaluation[…]

Fully working plugin for C4D

When you will be working on it please try to keep […]

Let's talk about Maxwell 5.2

Is there any update regarding CPU engine ? Som[…]

Hello, I agree with Jasper ... I still have this […]