I created about 60 image inputs for a show and then layered on various NDI inputs and graphics using the multiview layers of each of those inputs. I thought it would be wise to use an image input to establish the background of each scene rather than placing the background image as the base layer in the multiview in order to retain as many layers as possible for other use. Why waste a layer on the background? All the images I used were appropriate for the production quality (1920x1080) and not overly high resolution.

As I continued editing and testing the show I started noticing a lot of weird problems during transitions; the background would kind of pulse or flare in brightness during the transition. With merge transitions this seemed to happen halfway through the transition. With fade transitions there would be a sudden snap in background brightness at the end of the transition. I was also seeing some odd behavior in how some of the NDI inputs were transitioning; such as fading away to nothing and then snapping back to 100% visibility at the end of the transition. I could not make sense of why this was happening in some cases and not others. Sometimes I was seeing brightness flares or snaps transitioning between inputs with the exact same image background; sometimes I wasn't. Sometimes I was seeing brightness flares or snaps between inputs with different image backgrounds; sometimes I wasn't. Where there were problems they seemed to occur consistently but I could not determine what was causing the problem.

Of everything I fiddled with, the only thing that solved this problem was to change the scene inputs to transparent colors, create separate image inputs for the backgrounds, and then insert those into the base layer of the scene input multiviews. Only then did I see consistent smooth transitions as I would have expected to see from the very beginning.

I'm sharing this experience here because I could not find anything in these forums or the manual about issues transitioning between image inputs. The behavior I was observing sure seems like a bug of some kind, but perhaps there is advice from vMix somewhere not to utilize image inputs in the way I was trying to. If there is, I have not been able to find it.