So, with a 1080i source, now I'm wondering if it's still better policy to send 720p, or just to start uploading 1080p. There are a few issues at stake... smooth playability of 1080p... upload time... future-proofing... Wouldn't want to lose some audience if YouTube decide to start differentiating HD and Full HD.
Nick, why don't you join my discussion in the other thread. I spent a day of angst over this, and came to the conclusion that "if" one has a 1080p monitor and "if" one has the horsepower to run YT 1080p at full screen on their browser without stuttering and tearing, then going to the extra time and trouble to transcode and upload 1080p might be worth it.
For the rest of us, 720p as you and I have concluded, remains the sane solution for now.
So the real question now is, does their 720p look better than our 720p?
Well, that's testable by sending them 720p and 1080p of the same thing then digging the 720p versions out of the browser cache and comparing them on the timeline or whatever. I don't suppose they're using Lanczos resizing so you'd expect a 720p upload to have the slight edge, but you never know.
Musicvid, do you mean the discussion in the big long thread or have I missed one?
I actually did this, re-render the 1080p video to 720p and upload it along side the 1080p version, this was just before noticing that the letters in the 1080p version didn't look like crap, and decided to re-upload the "render stress test" above.
It is not a test as such, just a pretty dark video, made significantly darker by YouTube I might add, as if they forced the cRGB to sRGB conversion after upload.
""...as if they forced the cRGB to sRGB conversion after upload."
It's actually the other way around. As expected, Youtube's Flash-based player expanded the levels by mapping. Look at the belle-nuit test chart in your uploaded example. Apparently rendered without the recommended Studio RGB filter, it shows clipped levels at both ends, 0-16 and 235-255.
"So the real question now is, does their 720p look better than our 720p?"
It would have to, in order to justify the inordinate amount of extra rendering time, upload/postprocess time, and load time on playback, amounting to about 2 hours extra time for my 3 minute clip at 1080p instead of 720p.
So the "other" real question is, is it worth it?
So which is better? Be your own judge, but the worse ME method generally gives a fatter-looking bird, all else being equal.
Thanks for pointing that out. I looked at the original mp4 I uploaded, and I have problems there, so I removed all colour correction from the tracks and rendered it as 32-bit Video levels, 32-bit Fill range and 8-bit.
Only 8-bit got it right.
32-bit video levels have the belle-nuit "16" at 6 and "235" at 232.
32-bit full range have the belle-nuit "16" at 21 and "235" at 214.
8-bit is spot on on all levels except "0" and "255" where I get 1 and 254 respectively.
"So the "other" real question is, is it worth it?"
Now there's a question :)
Render times don't worry me much, plenty else to do while that happens. Upload times are even less of an issue as I've got plenty of bandwidth to burn and an old laptop with bandwidth management.
Still no point producing pristine 720p let alone 1080p masterpieces if no one actually watches it at that resolution.
I had hoped YT would reveal how many viewers watch at each resolution but apparently not :)
If anyone knows of a way to get YT to reveal that statistic I'd be interested and it might add a meaningful perspective to this question.
The whole youtube, HD, and quality issue will drive you nuts.
I had a request from a relative to help him trim a few videos of his band. He showed me the stuff a photographer friend had taken (with DSLRS) and posted on youtube. "It's HD", he said. sure enough, it was 720p, and it was the most overcompressed mess I had seen in awhile (obviously, mashed up encoding before submission to youtube). I didn't know what to say - certainly, I didn't want to to put a damper on his enthusiam - but dam!
Especially with DSLR's, the Computer RGB->Studio RGB filter needs to be applied before rendering and upload. We've documented it to death over the past year. To show you how dangerous an NLE (in this case FCS) can be in the hands of the profoundly uninformed, I found this "artsy" shot on Youtube . . .
I think the lure of lucre to be made with video using DSLRS, for the stills photographer, can make them forget the fact that video is a whole 'nother animal.