Protospace vs YouTube Thumbnails

So I’m adding a logo to Calgary’s Protospace teaser video, and none of the YouTube thumbnails look particularly appealing. I don’t want a split-screen of a hacker. I want a Protospace logo dag-nab-it!

Protospace hacker VS Protospace logo for YouTube thumbnail. The choice is clear.

At some point in early 2009, YouTube stopped grabbing thumbnails from 1/4, 1/2, 3/4 positions in submitted videos. Thumbnails were being abused, with brief images of pornography and misleading images placed at 1/4 1/2 3/4 locations in uploaded videos.

I can’t define pornography, but I know it when I see it. And Protospace logo ain’t no pornography! The Protospace teaser is rendered at 29.970 fps (NTSC drop-frame). I added a frame counter to the video to locate the new thumbnail locations…

And confirmed what I’d heard when Googling the subject… the thumbnail locations are now pseudo-random!

YouTube's proposed Protospace video thumbnails.

Due to the “random” part of the thumbnail grabbing, it took me a few tries before a logo lined up with a thumbnail, but I did manage it…

…and then due to a medical condition the doctors refer to as “borderline stupid curiosity” I began adding frame counts to other videos, uploading them, and documenting YouTube’s 3 thumbnail locations for each.

Notice that the TN (ThumbNail) locations are constant for extremely short videos, and become more pseudo-random as the video length increases. I wouldn’t be surprised if an exact formula could be nailed down, but I don’t see it.

If you can solve the pattern, or have more data points you’d like to contribute then please ping me. I can either add them by hand, or share editing rights to this spreadsheet. I suspect any pattern may have something to do with MPEG-4 and keyframes. Or maybe Dwipal Desai injected a bit of the pseudo into the random because he enjoys the thought of thumbnail seekers endlessly spinning their wheels… heck it’s what I’d do if I were him!

, , , , ,

  1. No comments yet.

You must be logged in to post a comment.