Advantages of MPEG-1 (VCD) VBR?

[absinthe]

Lately I've been encoding full-movie backups to a single disc. I started using the TMPGEnc wizard, but quickly found that I could just plug the numbers into CCE and get the same result (of course, I have to run pulldown manually, if necessary, and mux manually ... but anyway) ...

Now is it just me, or does MPEG-1 video not come out better than MPEG-2 at these lower-variable-bit-rate encodes? Even though a 1-disc SVCD of a 110-minute movie will come out very acceptable, there is usually some noticeable ghosting and blockiness.

But I can squeeze a nearly 2-hour-long film onto 1 VCD (MPEG-1) with 2 passes in TMPGEnc (128 kbps audio) and the result is extremely good. That is, of course, unless the screen is almost totally black, and then there is the usual blocking effect. But I certainly find the VCD result better than the SVCD. This is especially true if the source was not pristine to begin with.

Am I alone here?

Anyone have any tips on how to sharpen up low-VBR SVCD encodes for 1-disc movies?

-abs

[jshumate]

I found your post to be extremely surprising. Quality is all in the eye of the beholder. I am NOT saying that you are wrong at all. I'm just pointing out that your idea of quality and mine (or someone else's) might vary quite a bit. My practical experience has been that unless you encode with CCE, MPEG-2 video below 2000 Kpbs is going to be pretty bad. You don't say what you are using for encodes. I have a SVCD that's about 50 minutes long that was encoded at approximately 1900 Kbps with CCE and it looks almost as good as the original DVD I ripped it from. Now my SVCD is very low motion, which probably helped. I cannot for the life of me imagine that you are truly able to encode MPEG-2 video for 110 minutes and get it to fit on one disk and look acceptible unless maybe you are watching in a small window on a PC. My experience has been the exact opposite of yours - MPEG-2 is pretty lousy at bit rates below 2000 Kbps and MPEG-1 is much better at those bit rates.

You can do what you propose with MPEG-1, but I don't think the final result is going to be very good. VCD standards do not allow for VBR video, so what you propose doing will not give you a valid VCD. That doesn't mean that DVD players won't play it, just be aware that you are not following the standards. http://www.dvdrhelp.com has a list of VCD, SVCD and DVD standards. My experience has been that MPEG-1
video at bit rates below VCD standards (1150 Kbps) is truly awful. MPEG-1 is pretty good at bit rates above VCD standards, though.

Maybe you would be interested in KVCD. It uses weird GOP settings (like 25 or more frames per GOP) and VBR encoding to squeeze a lot of video into a small space. There is a web site at http://www.kvcd.org.
Like I said, quality is subjective, so maybe KVCD is exactly what you are looking for. I haven't been real impressed with KVCD, for whatever that's worth. If you really just want to squeeze as much quality video as possible into a small space, DivX is probably the way to go. More and more DVD players are starting to support it now.

The only person you have to please is you, so if you like your results, that's all that matters. Just be aware that what you propose doing with MPEG-1 does violate VCD standards, which in turn means that some DVD players may not play such discs. Playing them on a PC will not be a problem.

[Nick]

Another thing to take into consideration is resolution.

SVCD is considerably higher resolution than VCD - 480x480(or 576) versus 352x240(or 288).

So in the case of a VCD each frame contains considerably less data to compress. Therefore to look upon a VCD vs SVCD encoding test as an MPEG1 vs MPEG2 test, is not strictly correct, since they are not like-for-like comparisons.

It may be interesting to try MPEG2 but at VCD resolution to see how it compares. Although whether you'll end up with any sort of playable disc is unlikely - it would be purely for academic purposes.

[absinthe]

I think I may have misled you with my third paragraph above, which I've edited to make clear that I meant that 2-pass VBR encodes of MPEG-1 looking better ....

Anyway, obviously I'm not talking about stellar, DVD-like quality here. I consider anything on a CD to be second-rate, anyway. I think the bottom line I was getting at is pretty much in line with what you mentioned: that squeezing movies onto a single CD, if you're going to do such a thing, is best done with MPEG-1 video; it just comes out looking a whole lot better than if you do the same thing with MPEG-2 (i.e., there are less artifacts).

I played around with KVCD for a long time. I've probably made 100 of them. It's a useful matrix for DVD, but I just got tired of doing it. The "size prediction" process is such a pain in the *ss. (But if you hang around the forum at KVCD.net long enough, you'll also notice that many people there prefer MPEG-1 for their encodes, too).

Thing is: I know a chap who routinely burns long movies onto a single SVCD. I'm talking sometimes as long as 2 and a half hours. I've seen many of these, and the quality is just unbelievable. That's what got me started experimenting in the first place, but he swears up and down that he does not use the KVCD matrix and that everything about his discs is totally compliant. I've never had a problem playing any of them. He says he writes his own scripts, and I do know that he routinely does 9 passes in CCE.

Anyway ... I'm still trying to get mine to look that good.

-abs

[jshumate]

Thanks for the follow up . It was very interesting. Do I think that doing a 2 pass VBR on MPEG-1 would improve your quality? Yes, but 3 passes might be better. I would love to analyze a SVCD like you mentioned from this guy. If you have a copy and can run the video through Teco Ltd's Bit Rate Viewer (http://www.tecoltd.com), I would love to see the output. I'd be interested to know his resolution, minimum, maximum and average bit rates, frame rate, and what the GOPs look like. If you are able to post this information or send it to me directly (it's been a while, but I think I set up my account here to let people email me if they want to), I would be very interested to look at it. CCE can give some incredible results, but I've never tried it below 1900 Kbps. I moved to doing DVDs some time ago and the VCD/SVCD world doesn't hold a lot of interest for me any more, but I'd be interested in seeing analysis of this guy's SVCDs because maybe I can learn something from it.

[absinthe]

Originally posted by jshumate
I would love to analyze a SVCD like you mentioned from this guy. If you have a copy and can run the video through Teco Ltd's Bit Rate Viewer. . .
I can do exactly that. I'd like to know more myself. Check your PM's.

thanks,

-abs