@DDogg: Elementary streams and 1/8 thing

[r6d2]

@DDogg,

I've been trying your 1/8 thing. It works great. I use DVD2SVCD (SVCD mode), set media size to 555 or so (I have not been able to figure the best number yet), and Author with either TMPGEnc DVD Author or DVDlab.

Unfortunately, both of my authoring tools trial periods ended before I could make a final choice. I've tried a freeware GUI for DVDAuthor I found but it is on very preliminary stages and won't even let you save a project.

So I guess I'll have to buy a good authoring tool.

I've heard you say DVDlab is the way to go. Can you elaborate on that?

[DDogg]

I guess I should have said "for me". DVDLab works the way my head does, so I find it very intuitive unlike many of the others. Of course, I don't do subs and multiple audio tracks and neither does the base version of DVDLab. The upcoming Pro version does and looks very interesting. The price points on both are very low for the rich feature set each provides.

In just a few minutes with DVDLab I can create great looking menus as complex as I want, author and burn. Also, it supports importing SVCDs and that was important for me.

Like I said, it just works for me better than anything I have used. I should also say that I am not a 'pro' user and only use a very small portion of the features available in DVDLab.

[r6d2]

:thanks:

[r6d2]

@DDogg, I've found that Half D1 anamorphic encodes look really good both on my TV (of course, since CVD looks OK on TV anyway), but also on my projector, which was not something I expected.

I can fit about 4 average-compressibility movies per DVD. Have you tried that?

[DDogg]

I've done three at full res and held a Q under 30, but I probably used 704. They looked great. If you went to 1/2D1 I would imagine it to be fairly easy to do 4 at good quality, maybe even 5 or 6 if they were easy to compress and shorter. You are halving the bitrate for the same Q when you go to 1/2D1 so that makes sense. I am kinda surprised that they do look good on the projector, although I know I've been surprised myself on how good 1/2D1 looks on my older TV. They are much softer looking, but then you are the guy that can't tell bilinear from bicubic so you would not notice it. :-D

[r6d2]

Originally posted by DDogg
They are much softer looking, but then you are the guy that can't tell bilinear from bicubic so you would not notice it. :-D LOL

Well, I've done the following: FACAR with ResizeTo=2 (source is anamorphic, so this works if you set Anamorphic no borders as encoding mode in D2S). I'm using Lanczos since the resizing is "heavy" (2:1), and since it is done only in one dimension, no additional artifacts are created by the resizer.

So at playback my PC player (I'm, using MPC) resizes to 469 (352*4/3) and then to 640 to fix the NTSC PAR. The projector plays that cleanly, no typical CVD aliasing noticeable. It makes no sense to use higher resolutions since the projector resizes internally again. So 640x480 looks pretty enough.

I could go for AutoGK too, but these MPEG-2 encodes play also on my standalone so I have best of both worlds.

I've found that using 528 as width is even more perfect, but I have 2 problems doing that:

1. Authoring programs reject odd widths as they are not DVD compliant.

2. Subtitles are not properly centered by D2S on unsupported widths.

If you want to try, please give me your thoughts.

[DDogg]

I just stick to 704 as I don't use a PC or projector. Just your plain old cheap standalone. Speaking of cheap, I picked up one of those mini Cyberhomes for 34.95 the other day. Has progressive, component output, etc. Man, I hate what offshore manufacturing has done to the blue collar workers in the States, but as a consumer, it is hard to bitch about it.

[Nick]

@R6D2

One thing that might help you make your mind up regarding authoring softwares is that TMPGDA does NOT support Anamorphic ½D1 files! You have to patch them to full D1 with DVDPatcher, author the DVD, then patch the vobs back again!

For people like us who have discovered the beauty of the ½D1 anamorphic multi-movie DVD, this is something of a disaster!

@DDogg

I don't use DVDLab (already paid for TMPGDA and don't wanna have to buy another just yet) so cannot verify that DVDLab does support such files. Maybe you could do a quick test on that front...


Cheers
Nick

[jsoto]

maybe even 5 or 6 if they were easy to compress and shorter. Hey DDogg, seems you're the poorest man I've seen. I knew r6d2 was a poor man , but I didn't notice about you :D

Seriously, half D1 is a very good resolution. I normally use it for extras, with mdvdauth plugin (IFOEdit authoring). Not sure if I used it in Anamorphic (probably, I have to check my copies), but I never have a problem with this resolution, except my doubts about if bitrate constraints should be applicable or not
I did some tests, and I found a bitrate reduction pretty close to 50%. But I've also find some strange behaviors in the case of bad quality originals.
BTW, do you know if VCD resolution in MPEG-2 is supported by the DVD standard?. I'm using these settings in long, bad quality extras, but not sure if I'm out of the standard or not.

jsoto

[r6d2]

Originally posted by Nick
You have to patch them to full D1 with DVDPatcher, author the DVD, then patch the vobs back again!Yes, I also found that trick, and it is a real PITA when you have to do it too often.
I don't use DVDLab so cannot verify that DVDLab does support such files.Yes, it does. It has some settings about loose compliance you have to adjust, but after that it does the job.

[r6d2]

Originally posted by jsoto
I did some tests, and I found a bitrate reduction pretty close to 50%.That much? Are you sure both resulting encodes have the same quantization? My tests show a 25% to 33% BR reduction with OPV Q=30 or so.

[jsoto]

That much? Are you sure both resulting encodes have the same quantization? My tests show a 25% to 33% BR reduction with OPV Q=30 or so. Yes, close to 50%, but I have to confess I did the tests with only a clip sample, good quality, low motion. No bitrate restrictions applied. (well, DVD one, max 9Mbit/s). Unfortunately I've lost in the last PC cleaning the spreadsheet I did, comparing VCD, CVD, SVCD and DVD MB per minute. IIRC, I used OPV Q=20. And No, I didn't check the quantization.

jsoto

[r6d2]

Originally posted by jsoto
IIRC, I used OPV Q=20. And No, I didn't check the quantization.If Q is the same, quantization will be the same, so it's OK. But the BR limits must be the same in order to compare apples to apples.

Don't worry about using DVD BR limits with Half-D1. It will play OK on DVD media/players.

[DDogg]

The reduction from D1 to 1/2 D1 is exactly, or damn close, to 50% to yield the same Q. Run a sample of D1 and another with the same settings right after it at 1/2D1. It ought to be exactly 50%, well maybe a bit or 2, one way or another. This assumes the first is just a mpeg2source line and the second just has the 1.2D1 resize.

[r6d2]

Originally posted by DDogg
The reduction from D1 to 1/2 D1 is exactly, or damn close, to 50% to yield the same Q.I have this result at hand:

Bad Boys II
3,43-GB 704x480 16:9
2,13-GB 352x480 16:9

Scarface
2,70-GB 704x480 16:9
1,69-GB 352x480 16:9

That's 38% reduction, both cases Q=30, KISS enabled, caeteris paribus.

[DDogg]

I'll have to do a test tonight on a full encode. All my tests have shown it to be haved, but then all my tests were on samples. It bears futher checking.

[DDogg]

R6D2, I did two full Q28 OPV encodes and these are the results. Are you sure you are not introducing some other factor in your scripts? This is typical of all my other test results on samples.

EncodeFull.mpv Frames 168718 D-ABR 2361 Size 2,076,347,292 bytes
EncodeHalf.mpv Frames 168718 D-ABR 1221 Size 1,074,429,880 bytes

Reduction = 48.4 %

Script for half:

mpeg2source("d:\dvdout\swat.d2v")
BilinearResize(352,480)
converttoyuy2()

I'll do the same thing again, but use Undot().Deen()
Maybe that is causing a different effect. Will post results later.

/Add: Ah, there is a slight difference when using filters:

EncodeFull_Filter.mpv Frames 168718 D-ABR 1984 Size 1,744,901,764 bytes
EncodeHalf_Filter.mpv Frames 168718 D-ABR 1131 Size 994,862,224 bytes

Reduction = 43%

[jsoto]

Ah, there is a slight difference when using filters: I didn´t use KISS in my tests, seems this is the reason.
jsoto

[r6d2]

Hmmm. I used Lanczos. Maybe that's the actual reason?

KISS should make no big difference because it affects both streams the same way.

The resizer, OTOH, only affects when it does resizing. If you use 704 with FACAR you just crop, and the resizer is neutral.

I'll test with bilinear.

[DDogg]

"I'll test with bilinear". That would be good, we need to pin it down properly.Originally posted by Jsoto
Unfortunately I've lost in the last PC cleaning the spreadsheet I did, comparing VCD, CVD, SVCD and DVD MB per minute. IIRC, I used OPV Q=20. And No, I didn't check the quantization.Remind me to murder you the next time I see you :) [I'm thinking, is that gonna translate properly to jsoto as a friendly joke? :)] I would have loved to have seen that data.

[r6d2]

Originally posted by DDogg
I would have loved to have seen that data. I did the same for the Idiot's Guide. You may want to look at it. All clips are done with bilinear.

[jsoto]

Remind me to murder you the next time I see you Actually you cannot. I already killed myself because of that. :)

Anyway, I did the same tests as r6d2 Idiot's guide. I'm going to repeat them. But I'm sure I was using BicubicResize and no filters.

jsoto

[jsoto]

My tests:
http://www.posunplugged.com/jsoto/resize_tests.xls

They confirm your results. KISS filtering affects them.

jsoto

[DDogg]

Looks like we were thinking along the same lines. Here (http://forum.doom9.org/showthread.php?s=&postid=488646#post488646).