Vegas to Youtube for the masses

johnmeyer wrote on 6/15/2011, 9:01 PM

Amazingly good! Thanks!!

LReavis wrote on 6/15/2011, 9:03 PM

I just want to say again how much I appreciate what you and the others have accomplished - and so generously shared here on the Vegas forum. I've been using essentially this method (that I learned from the long threads) for quite some time now and still can't get over the quality of the encodes.

I recently posted some videos compressed at only 150 kb/s. The following link shows one of the results. You really need to watch full-screen in order to appreciate the quality at this low bitrate (but please ignore the image quality of the guitar video - I shot it years ago on hi-8). Warning to some of you - this video still is not finished, and if you've seen it before, don't expect to see the complete version. It's at:

http://www.torealize.net/2physics.html

I must admit that I spent a lot of time in Vegas in order to get the bitrate that low: I shot in good light in order to minimize noise, I used green-screen so that I could blur the background, which minimizes bandwidth demands for the leaves on the trees in the early closeup, I avoided unnecessary motion, etc.

Still, I'd like to get the bandwidth even lower. As someone pointed out to me, connections in internet cafes, libraries, etc., can be really slow. When played over my own website, the overseas intermediate web links sometimes are terrible, even though they seem good enough to play well in much of Europe.

So, when time allows, I plan to offer a small-screen version (which, I hope, also will play in Android prior to version 2.2, and other non-Flash-friendly environments). Then maybe I'll try for around 50 kb/s. Can't be much lower than that, I fear, because of the demands of audio. Need to do some additional experiments . . .

But first, I have a number of other videos that are in various stages of completion. Moreover, I'd like to go back someday and do a better job correcting color, etc. I suppose my body will conk out before I get it all done ~

In the meantime, thanks again to musicvid and the others - Larry Reavis

musicvid10 wrote on 6/15/2011, 9:08 PM

LReavis,

Jerry (amendegw) can give you more insight on ultra-low bitrates, he's done a lot of experiments with low-motion source using both x264 and HTML5.

Amazing how far we've come. 320x240 mpeg-1 wasn't watchable at under 1Mbs, back when we were encoding it for "video emails."

ushere wrote on 6/15/2011, 9:48 PM

thanks musicvid - very useful indeed.

i have to say that (working with hdv - pal) i've tried all three methods and though i agree that 'best' does do a 'better' job, the 'better' (with my material anyway) doesn't warrant the time / effort, especially for youtube, etc. which in most cases is an after thought with my clients.

this wont deter my pointing students to your video though and letting them experiment and so letting them make their own decisions.

as an aside - i now 'checkout' my online videos whenever i get a chance to on other peoples (private / corporate) pc's / mac's.... i sometimes wonder why we worry at all!!!
many laptops i've come across are set at other than optimum display ('oh, everything was TOO SMALL), lcd's are set to either 'vivid' or 'theatre', and the few board room set ups i've seen are still set to 'demo' mode from the store - bright and oversaturated.

of course, this is not to mention what happens with dvd / lcd aspect ratios either!

musicvid10 wrote on 6/15/2011, 10:12 PM

Leslie,
Side-by-side comparisons of Nick's "best" method with "better" using the same footage show it is indeed superior, due in large part to QTGMC deinterlace, which is amazing, but slow as snails by comparison. I suspect we could come close in HB by using Yadif, EED12, and McDEINT in some combination, but again at the expense of inordinate increase in processing time. Because of the time factor, the Handbrake devs have shown remarkably little interest in incorporating QTGMC into their project.

Since we began this undertaking knowing Youtube's quality limitations would stop us at some point from actually making a silk purse, I think we did pretty well, while keeping the processing time on par with our benchmark MC 2-pass in Vegas (actually beating it in some projects). I would be delighted if one or more of your students got a better deinterlaced product while keeping the encoding time in line with current numbers.

I feel your pain about the way your stuff is viewed. As producers, we have no control over the endpoints, but if we keep the product uniform, there is a theory in the business that the world will eventually migrate their tastes to fit our norm (excluding TV commercials, of course!)

And somewhere on Youtube, there actually lives a tutorial on how to polish a t*rd.
;?)

farss wrote on 6/15/2011, 10:49 PM

Two comments:

1) Not so certain about applying the Levels correction. See my thread about this. I think you can see why this is not such a good recommendation.

2) The YADIF de-interlacer is now available in Vegas, V10 only as it's an OFX plug. Using that which is drop dead simple to use I'm pretty happy with the results I get straight out of Vegas. It avoids all the futzing around and uncertainty of using the Smart De-interlacer.

3) Just a very minor point. It's 50i not 25i :)

Bob.

musicvid10 wrote on 6/15/2011, 11:08 PM

9:05

NickHope wrote on 6/15/2011, 11:08 PM

Excellent job, musicvid, and I've added a link to it in my guide. Thanks a million for the link to my channel too, although at the moment the link is broken. It seems to be missing the final "n".

It might be an idea to put all the download links etc. into the video description and an annotation telling people they can find them there.

Regarding processing time, on a quadcore-or-more machine I think the QTGMC>Megui workflow can actually be faster than the DNxHD>Handbrake workflow, once one has refined the settings and achieved a stable multi-threaded setup. That in itself requires patience and a willingness to experiment, so it's not for everyone.

Regarding deinterlacing quality, I think a large part of the success of QTGMC is in the way it utilises MvTools, the same plugin that johnmeyer and others have used for slomo etc..

musicvid10 wrote on 6/15/2011, 11:26 PM

Thanks for the headsup. I'll check the link before I collapse.
[EDIT: Fixed.]

Excellent idea about putting the d/l links in the description. On my list for this week.

No doubt you've got better deinterlacing than the Handbrake default. If I was a youngster like you I'd be . . . oh, well.

amendegw wrote on 6/16/2011, 3:07 AM

"Jerry (amendegw) can give you more insight on ultra-low bitrates, he's done a lot of experiments with low-motion source using both x264 and HTML5."As mentioned in many of my earlier posts, I've found that getting quality low renders is pretty much a trial-and-error process, but a good starting point is the Handbrake setting of CQ:RF=30, then adjust based upon your quality standards. I've started doing some tests on "adaptive streaming" where JW Player automagically adjusts the video quality based upon your bandwidth, but so far the results are disappointing.

As far as HTML5 testing, I've pretty much stopped testing in this area as there are too many "standards" out there and I don't have the hardware to adequately test. You can search this forum for "html5" to see what I've done to-date.

Any specific questions on either of these topics, I'd be happy to take a shot at them.

...Jerry

dlion wrote on 6/16/2011, 6:29 AM

quite excellent. thanks!

Randy Brown wrote on 6/16/2011, 6:56 AM

I'm starting a new thread for this, as someone pointed out that the posts at the bottom of the megathread take forever to load.
Not to mention...for me at least...I'd have to be careful so that my head didn't explode with confusion...I've tried but there is no way I can keep up with you guys.
IOW, this video tutorial is very much appreciated...Thank you!!!

musicvid10 wrote on 6/16/2011, 9:22 AM

Jerry,
Out of curiosity, what other compression "goodies" are you using (in addition to CABAC) in Handbrake on low-bit renders?

8x8 ?
weight-p ?
pyramid-b ?

For sure, video encoded with weight-p "on" will not open again in Vegas. But that may not be a biggie.

LReavis wrote on 6/16/2011, 12:48 PM

"i've tried all three methods and though i agree that 'best' does do a 'better' job, the 'better' (with my material anyway) doesn't warrant the time / effort"

As others have noted, the "Best" method is superior (it seems) mainly because of superior deinterlacing. I've been using 60p cameras for some time now (currently a pair of TM700s; before, a pair of Sanyos), and they make life so much simpler. Not only is there no interlaced video to get in the way of efficiency, but I also get easy slow motion. Highly recommended.

amendegw wrote on 6/16/2011, 1:44 PM

"Out of curiosity, what other compression "goodies" are you using (in addition to CABAC) in Handbrake on low-bit renders?"musicvid, I think I'm using the same params that are referenced in your tutorial. On your advice (in another thread), I fiddled with some other mods and couldn't see a nickel's worth of difference. Here's the settings:

...Jerry

musicvid10 wrote on 6/16/2011, 2:02 PM

Good thing about keeping the settings simple is it will play on almost anything.

Just noticed the Youtube video got 200 hits in the first 24 hours. Didn't expect that. If Youtube offers me revenue sharing, I promise it will go into a Doggie Biscotti fund for Jazzy.
;?)

amendegw wrote on 6/16/2011, 2:51 PM

Jazzy is just enjoying his 16-235 minutes of fame!

...Jerry

musicvid10 wrote on 6/17/2011, 10:27 AM

After receiving a few requests, all of the download links have been added to the Description below the video.

Since this saves people from having to pause and copy the URLs by hand, I feel a bit remiss for not having thought of it.

farss wrote on 6/18/2011, 3:52 AM

Musicvid said:

"9:05"

OK, at first I was going to say something else but I decided to stew on it for a while.

I think I now understand what you're trying to do, clue is "failsafe".

Excellent idea if I'm right, kudos to you for taking that approach. Problem I still have is you need to say something different at the outset otherwise you're going to confuse a lot of already uncertain people.I'll not say what or I might let the cat out of the bag in your other thread.

Bob.

musicvid10 wrote on 6/18/2011, 9:39 AM

"I think I now understand what you're trying to do, clue is "failsafe".

Bob,
Totally correct, and thanks for being willing to take a different look.

At its simplest, my approach is WYSIWYG with regard to Vegas preview, and "failsafe" in the sense that existing video levels will never be clipped on playback. In that sense, one could easily apply the label "conservative."

Approaching it this way requires thinking of it completely backwards from the way a professional would; you could call it "match the output to the preview" rather than "match the preview to the output." My approach is geared to the other 98% of Youtubers who shoot out of range, are clueless about levels and probably wish to remain that way. Each approach comes with its own built-in set of pitfalls, but clipping on playback is not going to be one of them using my method.

You are now the fourth person who has unraveled the reason behind the mini-poll. And the results sofar (even on a pro forum) are very suggestive of the reason most services and players do what they do to our dynamic range. Doesn't mean their approach is "right," just that they will always choose one that gets them the fewest complaints.

No, I'm not switching camps just yet, I'm going to keep challenging others to take an alternative look, even if it seems backwards to the informed minority. I think Jamie Oliver would approve.

Congratulations on your own sleuthing and professional approach to the issue, and as far as explaining it better at the outset (from which I could obviously benefit), I will give weight to your input for the next version of the tutorial. And there will be one, given the number of changes coming for Youtube and Vegas down the road. Feel free to send your narrative suggestions to my username at gmail.

Glad we finally achieved a meeting of minds on this. Although you came in late and from around the corner, I would like to consider you a team member. I owe you a sixer of cold microbrew from the Rockies.

musicvid10 wrote on 6/23/2011, 1:13 PM

@ Mike (rs170a),

Thanks so much for embedding the tutorial on the "Cow."
So far over 600 hits total and counting. Jerry has had a couple of hundred on his own site as well.

Almost all responses have been that they like it or don't understand it at all . . .

VanLazarus wrote on 6/23/2011, 1:31 PM

@musicvid. Thanks again for this excellent tutorial and all the research you've done.

A few quick questions... Is most of the benefit of the Better and Best methods regarding deinterlacing? If I take progressive source material are all the methods closer in quality?

I abhor interlaced video and avoid it like the plague and hope to rarely have to work with it.

amendegw wrote on 6/23/2011, 1:41 PM

Project page has 850 hits (and climbing). That does not include the 360 hits on the earlier, first draft page.

I sent musicvid & Alistair a PM, but I'll repeat it here, "When they accept their Oscar, don't forget to give a shout-out to Jazzy the Dog!" [chuckle]

...Jerry

musicvid10 wrote on 6/23/2011, 2:43 PM

"Is most of the benefit of the Better and Best methods regarding deinterlacing?"

That and resizing. If you are starting with 720p there is not much advantage over doing it all in Vegas. The differences begin to show up in high detail in motion and at lower bitrates.

Vegas to Youtube for the masses - a Tutorial

Comments