That's totally due to Youtube's low minimum vbr and the sensitive flat nature of stringer's source. No way to get around it; it's not present in the uploaded mp4, and a higher bitrate in the upload doesn't fix it. Baaad Youtube!
Yes, my results are much better keeping the intermediate at 1080i, and resizing / decomb in HB.
Yes, that looks like my settings, EXCEPT I had the Vegas framerate at 29.97 (project setting).
Got to go do some work, but I will be around later this evening to post some Handbrake settings, also the levels for your poppies, which were tricky.
"...also the levels for your poppies, which were tricky. "Ha! I just followed the rules of the assignment - "The footage should be well-lit, not clipped, and contain motion as well as static detail. Tripod is a must. Prefer colorful subject material"
btw, one thing I noticed after I sent you the clip - my camera is searching for focus at full-zoom. Maybe no one else notices this, but I do. Hope no one attritutes this "glitch" to your test.
I knew I was going to have to tweak levels, but your poppies presented a challenge, although the luminance was within gamut.
Youtube would have muddied the saturated reds and blocked up the shadows somewhat unless we brought the chroma into range as well. What I arrived at (after some failed attempts) worked well after upstream postprocessing, especially on Vimeo.
You've somehow managed to clip the red and green channels, perhaps you somehow overexposed the shot to start with. The curve that you're using is probably not helping either.
Given that most camera today go over 100% by design it's oftenly better to roll the highlights off than to clip them. A curve such as the following may give better results
Another thing I'm not 100% certain of given that your original footage is probably a bit overexposed is it seems to have too much edge enhancement. If your camera permits dialing down the Detail setting to a negative number might help.
Excellent thread musicvid. I'm trying to determine which, if any, or your assumptions apply to the way I shoot. In general, I shoot progressive rather than interlace, and don't usually resize unless downsizing for DVD.
Following are my questions and/or assumptions regarding the points you posted. I'm posting these in an attempt to better understand the applicability of your findings to my workflow. Perhaps there are other Vegas users who also follow a workflow similar to mine.
-- That x264 holds up better below 10Mbs than does Mainconcept AVC.
I have also assumed this and always encode web media and home streaming video using Sony AVC.
-- That 720p is currently a better upload medium than 1080i, from an upstream postprocessing standpoint, and playability on most consumer systems and connections.
I tend to agree. My assumption here is that, all else equal, you can use a lower bit rate on 720p than 1080.
-- That yadif is a more versatile deinterlace method than blend or interpolate, in most situations.
N/A - no interlace here.
-- That lanczos3 resize is better than bicubic, with or without sharpening, in most situations.
N/A - no resizing here.
-- That uploaded media must be strictly conformed to 16-235 levels, to prevent clipping at both ends. Includes generated text and media, and FTB levels!
This surprises me since I thought computer displays could handle 0-255 (or at least 16-255). I would like to learn more about this. What is "FTB levels"?
-- That uploaded media should be streaming-ready.
Why? If you are talking YouTube and Vimeo, don't they add the streaming flag when they transcode? I use SmugMug and they don't stream - the clip downloads then plays.
-- That a DNxHD 220Mbs intermediate is probably overkill for 4:2:0 source, and is probably not worth the extra time and file size over 145Mbs.
Glad to hear that. I am rendering straight from AVCHD - no intermediate.
-- That the time spent encoding the intermediate and again in Handbrake is reasonable, compared to say, 2-pass VBR in Vegas.
It isn't so much the time that bothers me, as the extra manual steps in going to intermediate and Handbrake. So my question is: Do you think the Vegas->Intermediate->Handbrake workflow would yield a significant benefit for the case where there is no deinterlace or resizing required?
Here are my impressions to your questions in order:
1. Holds up better than Sony AVC, too. Also, Sony is CBR, meaning larger files for a given source with or without motion.
2. Yes, 65% lower bitrate, all other things being equal. That figures into the playability piece on many systems and connections.
3, 4. If you are starting with 720p, the biggest advantages offered by Handbrake (resize, decomb) are off the table. I would probably just render in Vegas in this case, to save a couple of steps.
5. Youtube and Vimeo both do an automatic cRGB mapping, whether your footage needs it or not. Thus if you give them 0-255, it will chop them, destroying highlight and shadow detail, and give an unwanted increase in contrast. This is true with all upload formats. Getting wysiwyg from these services is something we've spent a lot of time investigating. See video link at bottom.
6. If you give them non-streaming source, the upload servers must do an extra preliminary pass to find the metadata, taking up more time and resources than is necessary. This has a cumulative effect on overall server response.
7. I am unable to see any difference, either in the intermediate or in the final mp4 output. (I don't have an objective method to determine this).
8. Probably not (see 3. above).
Thanks for the level of inquiry. Although I don't claim to be right on all my assumptions, if it serves to get others thinking about it, we are all bound to benefit.
Ah! Those are exactly the same settings I've been using - save one. I've found that if I set video encoding to 2-pass VBR, I can get away with ridiculously low bitrates (often < 1Mbps) as long as there is not continously fast motion in the video clip.
PS: Obviously, the audio settings can be varied as well - not much need for very high bitrates if music is not involved.
Yes, I noticed that 2-pass VBR (non-turbo) is better at very low bitrates.
But for Youtube and Vimeo uploads at 6-10 Mbs, Constant Quality is a lot faster render.
I appreciate your work in squeezing the most quality out of very low bitrates. Few of us own the server space to offer huge HD files, thus the move to embedded players for many.