Vegas8 and Vegas11 big quality difference Sony AVC

Infidel wrote on 6/4/2012, 4:38 PM

I personally happen to use Vegas 9 but apparently the same problem is there. This question is for my friend, who was on V8 but has since upgraded to V11. Since he has both versions on different computers, he was able make this comparison. And since we're talking many different computers I don't think system specs come in to play here.

The problem is this, Vegas8 used to render great Sony AVC files. But after V9, and including V11, the very same renders with the same settings produce a slightly larger file that has very noticeable worse quality. So bad in fact, that in Vegas 11, he was forced to render with the MainConcept engine, only this produced files 3x the size to get the same quality!

It doesn't look like I can post attachments or screen shots here but here are the settings:
V8: AVC, 1280x720, CABAC, 29.970 (NTSC), pixel aspect 1.000, bit rate 512,000. Field order none.

V11: AVC, 1280x720, CABAC, 29.970 (NTSC), pixel aspect 1.000, bit rate 512,000. Field order none.

That's kind of redundant, the settings are the same, for both video and audio.
The point is, the Sony AVC renderer created such vastly different output between V8 and V11 it made the V11 render unusable and that's why he switched to MainConcept, but again, that creates three times the file size to get the same quality!

Let me iterate this issue one last time in another way. In order to render videos as good as Vegas8 AVC, he has to render with MainConcept at 3 times the file size to get the SAME quality.

Anybody have an idea what happened to the AVC render engine? Or what this could be?

If I could describe the quality, imagine the V8 engine rendered an image for the video, the pixel resolution would be nice and it would be a solid still image as the video plays. In the V11 render, the pixels are obviously degraded, and then this "still" image seems to jump around and pixels move and change during playback. In V8, a still portion of the video the pixels would not be jumping around and changing, it stays pretty still.
The point is, the same render engine is producing quite different renders. Not only is the V11 render worse quality, the file size was actually about 200K larger in a test video of 1 minute, 11 seconds long. Obviously something changed in the Sony AVC render engine, and changed for the worse!

Help!

Comments

amendegw wrote on 6/4/2012, 5:20 PM

"It doesn't look like I can post attachments or screen shots here but here are the settings:This question comes up so often, I've created a cut-and-paste to help new users:

The short answer is: [img=http://dl.dropbox.com/u/20447760/Jazzy4.jpg]

The longer, better answer is in the sticky

The best answer is:

You can use the windows Snipping Tool instead of SnagIt & the web service of you choice instead of the free Dropbox service.

...Jerry

PS: Search this forum on "HandBrake"

videoITguy wrote on 6/4/2012, 5:54 PM

The question you pose about differences in AVC encoding are interesting. Certainly it is possible that the encoder did change between versions 8 and 9 or 11 but I doubt it. This is the first time I heard of this being brought up so I would like to have more input from a broad number of users.

I only use SCS VegasProVersions 8 and 9 (don't care to work with 11). Here is what I have noticed between those two versions as a change(?). The render dialogue settings have to be created manually in a custom template-saved render box in Version 9 to "recreate" the settings that were used originally in Version 8. I no longer trust the templates given in 9 and always use SCS Pro 8 as the reference. Hence my templates that I want to use exactly correspond in 8 and 9.

You may want to check that as a variable in your cross-version testing that you appear to be doing.

Serena wrote on 6/4/2012, 6:36 PM

this thread will assist:http://www.sonycreativesoftware.com/forums/ShowMessage.asp?ForumID=4&MessageID=812547codec comparisons[/link].
Look at the fine detail in the 3 renders shown. Certainly the Sony AVC is the worst of the three methods (at low bit rates).

videoITguy wrote on 6/4/2012, 6:57 PM

Serena's reference thread also contains a very significant finding from "Johnny Roy" that points out that default templates were seriously modified between different versions of VegasPro. There may have been intentions by SCS that were well placed but AFAI know they have never been explained by SONY.

Again I shall repeat my previous statement above, I use the template settings of version 8.0b for "standards". Introducing manual copies of every little adjustment in the template of later Vegas versions has been worthy of the time creating them.

rmack350 wrote on 6/5/2012, 12:44 AM

I don't know if this applies here but i think vegas started doing some wacky things with deinterlace modes somewhere in the later vp9 revisions. I suspect its not the codec that changed, it's vegas' defauts.

Rob

rmack350 wrote on 6/5/2012, 1:03 PM

It might help to know what you're working with, in addition to what you're rendering out to. I don't think the AVC codec has changed but Vegas most definitely did midway through VP9.

Rob

Infidel wrote on 6/6/2012, 10:12 AM

Here is a side-by-side compare of the V8 and V11 renders blown up to 2000% zoom.

I know you guys have been talking about "defaults", but I know that all the render settings are the same, so what defaults do you mean? Somewhere deep in the configs or something?

https://dl.dropbox.com/u/2018803/V8-V11-Compare.jpg

You can see the V11 render is quite different, I don't know if, based on how it looks, you might have a clue what it could be.

I told my friend about the Handbrake tests and based on the types of videos he makes, he just doesn't feel like having to render twice. Are there any Handbrake plugins to render directly from Vegas?

Any other ideas?

Laurence wrote on 6/6/2012, 11:23 AM

It's hard to tell for sure, but that screenshot makes it look like it might be some sort of cRGB/sRGB color issue. Is the color intensity or colorspace different between the two versions?

As far as Handbrake and two renders goes, I've been doing the extra render for a long time and it really is worth it. The quality is just so much better. I use the Microsoft Expression encoder for wmv encodes for the same reason. At high bitrates of course it doesn't matter, but who wants that? The whole point of AVC is to get the size way down.

rmack350 wrote on 6/6/2012, 12:54 PM

I can't tell anything from those views of pixels. Yes, the VP8 render looks like the edges are sharper and areas of solid color are averaged. The VP11 image looks like the edges are blurred but that maybe there's a little detail retained in solid color areas. But no one will look at your video this way.

I don't know what you're starting with. Stills? 720P video? 1080i video?

The default deinterlace settings would be within your project properties, not within your render template. If your project template is interlaced then there will be deinterlacing happening before the render to 720p. Also, if there's any resizing going on then you need to render as Best rather than Good.

I'm sure there are other considerations as well.

I supposed if I really wanted to take the time I could reload VP8 and just try to reproduce this. But I'm not dying of curiosity.

Rob

videoITguy wrote on 6/6/2012, 2:10 PM

I am reminded of the magazine publisher that tries to produce an authoritative article showing the differences between this and that photo of some process...the limits of publishing will not be able to show the differences. Way too many variables to introduce here.

I do believe that there may be substantial differences in encodes from a given source - but now the variables...what degree of difference are you seeing on what device?? I have very good production monitors calibrated to view video edits..but as much as I try, I can not for the life of me..absolutely make a given viewing the same on playback on all screens at the same time. Just close.

Laurence wrote on 6/6/2012, 5:54 PM

Keep in mind that the main reason most of us use AVC is for uploading video to Youtube or Vimeo. The encode that we upload will be re-encoded for display on their sites. All the artifacts in your encode will cause even more artifacts on there encode. The more pristine the quality of your upload, the more passable will be the quality of their encode. That is why I always use Handbrake for anything ending up on either of these services.

ushere wrote on 6/6/2012, 7:04 PM

@ laurence - what settings do you use in handbrake for youtube?

farss wrote on 6/6/2012, 7:53 PM

"The more pristine the quality of your upload, the more passable will be the quality of their encode."

That's true however I use the Sony AVC codec to provide 720p at >6Mbps.
My video is already pretty clean and if I've shot interlaced I use the Yadif de-interlacer. I just don't see any upside in this scenario to doing the HB dance.

There is always going to be "things" that YT makes a hash of. They're encoding at a low bitrate and no doubt favouring speed over quality. Doing our utmost to avoid content that can be turned into blobs and artifacts starting with how we shoot and light our content is always going to be my #1 priority. Following that up with not doing things to it in post that'll upset YT's encoders and we should be able to get good outcomes

Bob.

Laurence wrote on 6/6/2012, 9:56 PM

>@ laurence - what settings do you use in handbrake for youtube?

1280x720x29.97fps. Constant bit rate of 20. If it is interlaced I use the decomb filter. If I have a mix of interlaced and progressive footage, I render an interlaced master and use the decomb filter the same as if it was all progressive. This works really well. I always check the "web optimized" tab so that it gets processed a little quicker. I change the audio to stereo instead of down mixing surround into Dolby surround since my projects are all stereo. Pretty basic really.

Infidel wrote on 6/18/2012, 11:41 AM

Regarding the comparison I posted, the "video" is indeed a still in this case. Basically it is an image that sits there as an audio file plays.
If you look at the VP8 part, it stays exactly like that through the video, thus being a true "still", it doesn't move, shake, distort, flicker or do anything but sit there.
The VP11 does none of that, not only does it seem "blurred", but the pixels move, and flicker and does not represent being "still" at all.

Since my friend can still render things in both versions, is there any kind of test you'd like him to do? I can have him render some kind of test image or a quick clip of something, I don't know what else to say. The consensus seems to be that he is just missing some settings in order to match VP8 in some sense. Is there a screen shot of settings I can have him take so you can see? I believe he would say he has tried looking everywhere to make sure the settings are right.

Not sure what I can post to keep troubleshooting this, or might it be the case that the VP11 version of the renderer just isn't as good and that's all there is to it?

Thanks