does anyone have the same issue, that in a batch with at least 2 actions and a few nodes in general, if you change something in the start of the tree and the active context is somewhere at the end, the first action will generate a screenshot from all the flame interface. Mostly its from the batch, a few times the timeline.
It is printed in one of the inputs, so you can hide and unhide it or even restart the software It stays the same with the same screenshot. Only after changing the transparency (or I bet also coordinates), the right image is let through again. Sometimes the main interphase will also get an overlay from the actual image getting into the action.
I’m wondering, if sb had sth. like it and found a direct fix, if it is a common bug anyway or if I need to annoy our technican again to find something on the machine.
The error that flashes extremely briefly when I try and render mentions “hardware accumulation” but turning off motion blur and fiddling with sampling doesn’t help.
Currently the only way I’m able to get a working render out is to select my action outputs one at a time, shuttle back and forth until I get a good frame, then pre render just that output. If I change anything upstream, I have to do that again, otherwise I get something like this (I’ll file a bug tomorrow when I’m not so salty):
Somewhat intriguingly, when I tried to collapse the whole tree into a bfx, flame complained of a resolution mismatch on a color node that didn’t exist.
It’s exciting! I haven’t had a good WTF bug in a while!
I just found some temporally fix by accident. Turning the 3D view in the action from the active view on and off again will fix it until next time happening. (Edit: switching 1 frame in between will surely delete the wrong image after a lot of testing). At least way faster than prerenders or changing the layers’ transparency.
