- User Since
- May 12 2014, 7:23 PM (321 w, 5 d)
Sat, Jun 27
Mon, Jun 22
Thu, Jun 18
Just tested 2.90 from the buildbot and it still crashes on Mint and CenOS
Wed, Jun 17
Sat, Jun 13
Fri, Jun 12
Houdini doesn't do flipping as well on 24 fps
Jun 11 2020
Jun 10 2020
Jun 6 2020
Jun 2 2020
Jun 1 2020
May 28 2020
Since we must not change icons in the outliner as it's quite logical to have them from left to right (eye -> monitor -> camera) may be we should stick to this concept in the rest of the UI?
I propose to reorder modifiers buttons completely 1 2 3 4 -> 4 3 2 1. That will be On cage -> Edit mode -> Real time -> Render. Then reorder icons in the particles settings back to Real time -> Render. That will be the most consistent and logical. Render is the last operation, so it belongs to the right.
May 26 2020
May 23 2020
Can confirm. 2.83 latest is also affected
May 6 2020
Summoning @Brecht Van Lommel (brecht)
May 2 2020
Unfortunately it's not resolved in 2.83 as sss pass has been completely removed and merged to the diffuse. Well, no pass no problem, but the issue is very critical as I described here D6848.
By the way, if take a principled shader without sss and mix it with a separate sss shader, then passes are correct and diffuse has the information. But it's only true for 2.82.
This is clearly a not very good decision regarding compositing. We cannot drive translucence and sss separately anymore.
The picture is from 2.82.
Is there a better solution for the eevee passes consistency? Because this one kills compositing quite much.
To make it clear, sss pass lets you increase or decrease sss factor on compositing, as well as the color of sss. Compositing is not about simplification. It's about giving all the possible information to a compositing artist. What we need is not reducing the number of passes, but in contrary add more as much as possible.
Apr 25 2020
Apr 12 2020
Apr 11 2020
Mar 27 2020
By the way, UDIMs are not autodetected if using ctrl+shift+t for simultaneous shader channels population with image nodes. But may be it's up to the node wrangler addon functionality.
Mar 25 2020
Mar 17 2020
Mar 11 2020
Hey! Great that you've fixed memory releasing, but the "bug" was about memory consumption on layer creation (copy). Which, I assume, should be just a number of flags, not a huge data copy.
Mar 6 2020
The expected behavior is that you hide types of objects from the opengl viewport only, but you still see them in render view mode. There's nothing complicated here isn't it?
At least all the other GCC work that way. Another way those buttons have the same functionality as the hide buttons in the outliner.
Mar 5 2020
Oh, that was my thing. I must drag by icons of the files, not by names. That confused me. Can be closed I think. Thanks!
Mar 4 2020
Mar 3 2020
Feb 28 2020
Feb 13 2020
Feb 4 2020
Feb 2 2020
Jan 20 2020
Jan 18 2020
Jan 17 2020
@mankysee (mankysee) I'd avoid ptxes by any means. Long story short, any scenario that I've experienced throughout my career with ptxes was a huge pain.
Jan 15 2020
Jan 13 2020
Oh, you are correct! It's visualy consistent, but screwed up in curves! That must be my fault and the way i'd recorded it. Thank you, we can go on. The motion blur is correct.
Hm, the bug is still there in my scene.
Here it is to try. Just render the video.
Jan 12 2020
Jan 8 2020
Dec 18 2019
@Jacques Lucke (JacquesLucke) Please reopen the task. You closed it because of T70164, but that one is claimed as incorrect (which is arguable) and closed. But this task is 100% correct and needs a lot of attention, as the current behavior will break your complex scene to the ground.
Dec 13 2019
Dec 10 2019
Oh damn, you're right. I fooled myself. So the only thing left is the hidden render subdiv. If it's by design then we can call it a day.
Oh, thanks to you I found that adaptive is not working at all. I've unchecked adaptive, set to 0, checked back and got jagged mesh on render, but now it's clear that changing dicing scale whether it's 1 or 64 does absolutely nothing. It stays jagged no matter what I set. Try it!
Dec 9 2019
Nov 28 2019
Nov 25 2019
I've found that this bug appears even if an object is still and only camera moves. Still present in 2.81 release. Found that on a production sequence render. This is unfortunate.
Nov 21 2019
Oct 31 2019
Oct 10 2019
Ok, I can see the statement, but not the answer why it's a bad design (except that it's automatic in ZBrush).
I agree, that the base mesh should stay intact, at least as an option. I don't see why non-automatic way can interfere with a serious sculpt work.
All we need is stable and correct work of the multires modifier.
Oct 7 2019
Pablo suggested poking. So i'm poking :) Increasing motion blur steps doesn't help.
Oct 6 2019
Yes, found that too. It doesn't cycle through the outputs because ctrl_shift_click doesn't work on selected node at all. If you take an unconnected node, select it and ctrl_shift_click it - nothing happens.
Oct 4 2019
Sep 25 2019
Ok, fair enough
Sep 18 2019
Yes, the nowadays latest build is rock solid, couldn't crash it. But the latest on the moment of posting was crashing as hell.
I suppose it fixed itself yay! This is how it works, right? :)
Thank you for your time.
Sep 15 2019
Sep 11 2019
Sep 10 2019
Sep 6 2019
Sep 5 2019
Oh, you are right! That was my mistake. Close that.
Sep 4 2019
I can confirm, And while it's not that noticeable on a plain spinning object, it's very bad on an animation like this:
Sep 3 2019
Sep 1 2019
I've tried it in 2.81 and seems like this bug is gone. Is it?
Aug 29 2019
Yes, i meant the unchecked box behavior.
Aug 19 2019
Aug 18 2019
By the way, not only that. A couple of subdivisions can lag a lot while mesh editing. 2.79 doesn;t suffer from that. A lot of people complain about subdiv performance on editing.
Don't know if it's related, but seems so.