My only problem now is that when I right click the window capture and choose "resize output (source size)", so that I don't get any black bars in the recording, the recording is a bit blurrier than the actual thing. I've tried playing with output settings (using GPU video encoder, changing the rate control to CQP, changing the CQ level), but to no avail. It's always slightly blurrier than what the actual window looks like.
So if anyone knows how to record a non-maximized window without losing quality, please let me know.
The guest OS has a weird display size, to allow for VB's menu, the host's taskbar, etc. *That's* the original. It's not any of the standards that you're used to. Scaling at all, if you have content at exactly Nyquist, necessarily blurs it a bit.
Nyquist comes up a lot more in the context of audio, in which case you're pretty much only sampling in time. But it applies to video too, where you're sampling both in time and in space. From one pixel to its neighbor can also be considered a sample rate, and all the same math applies exactly the same way.
So if you have a single black pixel in a sea of white (common for text, even if the colors are different), then you probably won't get that sharpness back if you scale at all, even back to what it was. If you want to get back to what it was, then *every* feature everywhere must be strictly greater than (not equal to) one pixel in every direction.
And even *that* is not particularly accurate, because physical pixels have sharp boundaries so that they effectively count as many pixels all in the same area that are all controlled the same. Kinda like the squarewave example here, that is forced into perfect alignment even when the math says otherwise, and so *that* distorts the image too:
www.xiph.org
Maybe *that's* the point of stupidly-high resolutions: we don't actually see them any better, but they allow for visually-lossless manipulation. Of course, that requires, as I said above, no feature anywhere to ever use just a single pixel, but with the processing and manufacturing that we have now, it's actually becoming feasible to do that.