Do you believe the offered explanation of the Webdriver Torso videos?

[snot]

See http://www.bbc.com/news/technology-27778071 .

As I write, these videos are being uploaded to YouTube at a rate of more than one per minute. How much testing is really required, here? Why are they so worried about video quality, when download speeds are cr*p? Who or what is actually making the claimed comparison at the rate of these uploads?

Just interested in your take.

[CBGLuthier]

1. My download speeds are not crap. I watch youtube on a 46 inch TV

and quite often it looks really good.

I also upload a 20 minute video every day and it takes a certain amount of post processing time. I imagine they are doing large scale tests of many different tweaked algorithms to see what works best. Considering the huge quantity of videos they process even small tweaks that save time will equate to large savings in processing power, which costs money.

I would imagine they are using software for the comparisons. Take the compressed videos and break them out frame by frame and do a comparison with the original to see how much is lost.

[snot]

2. ok . . .

I just did the math. Since the channel's inception, it's averaged 179 videos per day. You think they're tweaking the code that many times per day? And there are other, similar channels.

Maybe it's just a way to make sure things are still working, i.e., to detect if something breaks down? But in any case, why go to the trouble of creating more than 90,000 (so far) unique videos for all these tests – why not just use the same one, or few, over and over – wouldn't that make it even easier to detect/compare changes in quality or speed?

[CBGLuthier]

3. Do you have an alternate theory as to why they are doing this?

I can see value in not using the same test image for every test. Testing your software in a limited way like that is not really a very good idea.