of my series of Minecraft videos is now up on YouTube.
It's been almost a full two months since episode 9. In part that was due to being busy with work, but it was also because I'd gotten into a bit of a funk about the videos.
As I've mentioned before, YouTube compresses the hell out of videos. Even if you upload fairly high quality it can end up looking like crap once YT is done with it. I spent quite a bit of time researching the settings others had used and doing test uploads to find the best recipe for getting YouTube to leave my video quality more or less intact and by episode 8 it seemed like I'd gotten it figured out. But then I uploaded episode 9 and it looked like crap again. The thought of having to spend yet more time probing the internals of YouTube put quite a damper on my enthusiasm for making videos.
This weekend I finally set myself down to start the investigation and very quickly noticed something rather odd: the quality of episode 9 was not nearly as bad as I'd remembered it. To be sure, there was some blurring of details not to be found in my original, but it wasn't all that noticeable.
I received a further surprise when I compared my videos against those of other Minecrafters whose video quality I admired and found, with one exception, that they were of similar quality to mine, displaying similar blurring of details.
Pleased by these discoveries, I recorded episode 10 (and re-recorded, then re-recorded again), edited it and uploaded it. But once that was done I was still left with the question of why my impression of my video quality had been so dramatically out of synch with reality. Had my internal critic really gotten that far out of hand?
As it turned out, there were three factors involved.
The first factor was that I was, in part, comparing apples to oranges. Note how I said that my video quality was similar to that of other Minecrafters, "with one exception"? That exception was KirrokCraft, whose video quality was still noticeably superior. No blurring of details for him! However, what I'd neglected to take into account was that KirrokCraft uses a stationary camera, separate from his in-game character, while I, like most other Minecrafters, use the default in-game camera which looks out through my character's eyes. As a result, my field of view is pretty much continually in motion. Even if my character is standing still I'm still looking around which means that the camera view is continually changing. Video compression works by just storing the differences between frames and it is the compression of those changes which leads to the blurring and other artifacts which impair video quality. Since the camera view in my videos is continually changing, there are lots of differences between frames, which meant lots of opportunities for compression artifacts to creep in. With KirrokCraft's stationary camera, the camera's view remains static. Any character or creature moving around in the field of view might generate some compression artifacts immediately around it, but the rest of view is unchanging, resulting in overall crisper images. If I used a stationary camera I would get equally crisp images.
The second factor had to do with way in which YouTube processes a newly uploaded video. YouTube generates separate copies of the video for each of the different resolutions that it supports: 240p and 320p (low quality), 480p (standard quality), 720p and 1080p (high definition). As soon as at least one resolution is available, YouTube makes the video available for public viewing. Since it's faster to generate the lower quality copies, those inevitably show up first. When I saw that episode 9 was available I immediately brought it up to check it out. I have my YouTube defaults set to display videos in 480p by default, so it brought up the 480p player. But the 480p version wasn't ready yet, so instead I ended up with the 320p version, scaled up to fit the 480p player. No wonder it looked like crap! Had I checked again 10 minutes later I would have seen the higher quality version. That also explains why the paramour didn't think the video quality was as bad as I did: zie'd viewed it later, after the higher-quality versions were available.
The third factor was that I was more sensitive to degradation in the quality of my own videos because I'd seen them in their original form. When watching other Minecrafters' videos I didn't have their source footage to compare against, so I judged them solely on their own merits.
Hopefully this marks the end of my obsession with the visual quality of my videos. Perhaps now I can turn some of that energy to improving the quality of their content. :-)