Late this evening I started to explore what the youtube second encoding actually does to the quality of the video. And unfortunately I found out that it is quite disastrous. I created a useless 1080p test video which you can watch at youtube, if you want.
At frame 320, that is after 10.68 seconds, when I am just turning the camera at high speed, I looked at three frames: the original frame of the 2.7 GB video with lossless compression by Fraps. The H.264 CRF=20 encoded frame (37.3 MB) which I uploaded to youtube. And the frame of the youtube video at 1080p (19.2 MB).
Here are the three pictures. For example open them in different tabs of the browser and quickly switch back and forth to spot the differences. Make sure that any zoom is off.
The difference between the original and the H.264 compressed pictures are visible, but negligible. The difference to the youtube version, however, is not negligible at all. And that's the 1080p version. The 720p version, let alone the 480p version look simply terrible. At least when I think of all the effort I put into creating the perfect quality/bandwidth solution.
Does anybody have an idea?