View Full Version : What source will be used in the next codec shoot-out?
Valeron
17th December 2004, 15:56
Maybe you should include "Ghost In the Shell 2: Innocence".
I think it's a good source even do 2 CD Rip in high resolution.
Doom9
17th December 2004, 21:34
Matrix: Revolutions
SPR
Futurama
iwod
18th December 2004, 09:18
Originally posted by Valeron
Maybe you should include "Ghost In the Shell 2: Innocence".
I think it's a good source even do 2 CD Rip in high resolution.
agree, i think there is another title with 2D charcter and 3D backgound which is very hard to encode.......... can't remember its english name though.....
krackato
18th December 2004, 12:50
I've always liked that scene in Attack of the Clones where Obi-Wan is fighting Jango Fett in the rain.
Doom9
18th December 2004, 13:07
I'm sure there are hundreds of sources I could take, but those are the ones that are going to be used. In fact there's no going back now..
SeeMoreDigital
18th December 2004, 16:06
Hi Doom9,
Will you be generating any encodes at 1:1?
Cheers
Doom9
18th December 2004, 18:30
Will you be generating any encodes at 1:1?Not any but all. I know you don't like that but until the day every participant does PAR properly it'll be 1:1. Plus moonlight demonstrated that PAR support can be a curse.. even though I know have proper AR on their content, their filters now screw up other sources... wouldn't happen if PAR wasn't an issue and I'm afraid I still have not found any reasonable explanation for the PAR values used. Perhaps if you like PAR so much you ought to write a little guide where all the math is explained, starting with 4:3, 16:9 and 1:1 sources ;)
SeeMoreDigital
18th December 2004, 18:57
Originally posted by Doom9
... Perhaps if you like PAR so much you ought to write a little guide where all the math is explained, starting with 4:3, 16:9 and 1:1 sources ;) Touché Doom9 ;)
As you know, I've never made any secret of the PAR signalling settings I prefer to use for my 1:1 encodes. Together with the reasons for doing so!
Hopefully your new tests will show that the Qf performance values for most codecs has changed. And that re-sizing encodes is not always the preferred option.
Cheers
krackato
19th December 2004, 02:08
Originally posted by Doom9
Not any but all. I know you don't like that but until the day every participant does PAR properly it'll be 1:1. Plus moonlight demonstrated that PAR support can be a curse.. even though I know have proper AR on their content, their filters now screw up other sources... wouldn't happen if PAR wasn't an issue and I'm afraid I still have not found any reasonable explanation for the PAR values used. Perhaps if you like PAR so much you ought to write a little guide where all the math is explained, starting with 4:3, 16:9 and 1:1 sources ;)
Nothing in that post made any sense to me.
1:1?
PAR?
proper AR?
Man, I've been away from these forums for a while.
gino25
19th December 2004, 13:31
And what codec will you use?
I suggest
Divx
Xvid
Nero h264
Snow (of ffvfw)
rv10
vp6
Hello
Doom9
19th December 2004, 13:42
And what codec will you use?I like to keep you in suspense. But I can promise that it'll be interesting.
Regarding snow and ffvfw I do have a comment: I wished that there were a few standalone vfw (or DS if it has to be) interfaces for some of the codecs that come with ffdshow.. it's just so exteremely overloaded. Especially the x264 codec where there's actually a VFW GUI but developers have abandoned it and not taken up the offer from people who wanted to maintain and update that GUI. I doubt very many use XviD in form of ffdshow - there's a good separate interface available with every XviD build and I have my doubts that XviD would be so successful if it only came as part of a massive package and inside *nix based commandline encoders that have to be crosscompile to get them working on Windows. I understand that certain people prefer linux and I have nothing against that, but the XviD team is also mostly using linux.. yet their success is based on their Windows users.. the codec can be cross-platform but imho you need to provide your own interfaces specifically tailored for the featureset you're offering.
Sagittaire
19th December 2004, 15:25
Matrix Revolutions 1CDR (hard encoding) in 720*304
SPR 1CDR (very hard encoding) in 640*272
Futurama 4*180 Mo in 720*400
with
H264 Nero PP0
RV10 HFE1
VP6 PP4
WMV9 VCM PP1
XviD PP4
DivX 5.2.1 PP4
DivX 3.11 PP4 for reference
Snow, x264, 3ivx, theora are useless for me ...
Manao
19th December 2004, 16:19
Sagittaire : SPR is 1.85, not 2.35, and anyway, it's hard enough to put it on 2 CDs, don't you think ?
Krackato : PAR means pixel aspect ratio. 1:1 means that the pixels of the video we're encoding are indeed square ( that's what most of us do when we resize before encoding ). On a DVD, pixels aren't square, they are streched vertically, and the picture must be stretched horizontally when the video is playbacked.
Search around the forum for more information, or follow the links in SeeMoreDigital's signature.
krackato
19th December 2004, 17:48
I'm interested in h.264 and Snow. I know next to nothing about Snow.
Manao
19th December 2004, 18:44
krackato : there's a thread speaking of Snow on this very page. But the best right now is to try it by yourself. There isn't much documentation about it, except for the source code.
hellgauss
19th December 2004, 21:45
If you have a camera, i suggest to use DV source, not only mpeg2 source.
Bye
HG
SeeMoreDigital
19th December 2004, 22:12
Originally posted by hellgauss
If you have a camera, i suggest to use DV source, not only mpeg2 source. Unlike DVD, a DV source is unable to offer progressive video, so it really does not compare!
If you want to see some great looking images using current affordable technology, take a look at SuperBit DVD.
Cheers
Doom9
19th December 2004, 23:44
before you make any more suggestions, let me quote myself (I don't like to repeat myself)
I'm sure there are hundreds of sources I could take, but those are the ones that are going to be used. In fact there's no going back now..
And no, I don't have a DV cam and no plans to buy one in the immediate future. Buy something that records interlaced? Shudder. Interlaced is evil ancient technology that we ought to have kicked out of our homes a long time ago.
hellgauss
20th December 2004, 00:18
I also don't like interlaced video, however there are many people that use dv camcorder in their holydays, and use video codecs to put video on cd/dvd. Today a dv-cam is the only (economic) way to produce digital video. since the source is quite different from a dvd movie i think it should be interesting see how codecs work with a lot of motion vector and a dirty but nearly-uncompressed source.
However if there is no going back, doesn't matter, i just want to say my opinion.
bye
HG
Sergei_Esenin
20th December 2004, 01:00
It really would be nice to have a full size, high quality source in roundups these days. Downsizing to low resolutions is fine for low resolution or SDTV displays, but today many people are using high-res displays and eschewing downsized video. Some are even encoding full HDTV resolution transport streams to 720p, 540p, or similar. It would be refreshing to see a high-quality representation--if not an HD res, at least full DVD res would be a nice change. Heck, my *TV caps* have better resolution and detail than most people's DVD backups these days..
dragongodz
20th December 2004, 01:27
i bet Doom9 drops and shakes his head when this sort of things starts up again. i dont blame him either.
listen its quite simple and has been gone over before with the previous codec shootouts. Doom9 will use the sources he wants to use and do the tests the way he wants to do them. so do not bother throwing more suggestions for something different unless you can come up with a truly strong case for it, not just "it would be nice to see". if you dont like that you are welcome to do your own shootout and post the results, maybe then you will understand how much work goes in to it.
MSlv
21st December 2004, 21:54
I really hope he uses some hard to compress DVDs. This way, the results would be more dramatic and exciting.
:D
FXsan
22nd December 2004, 00:42
It clearly involves a great job.
I would personnally also apreciate to see results about MPEG-2 and MPEG-4 ASP, as this would enable to have other (reliable) source than manufacturers themselves.
A consensus seems to be that an equivalent quality is obtained in MPEG-4 ASP with 30% less bitstream than MPEG-2, and in H264 with 50% less than MPEG-2.
Thus are these figures correct, and are they at 1, 2, 3 Mbps ?
Anyway, 4:2:2 still better than 4:2:0 ...
Doom9
22nd December 2004, 01:05
I'm gonna try and write up the hours used for each task, so you'll see how much work is actually involved, and so that you can get a rough estimate what adding a source or adding a codec actually means in terms of hours spent.
MSlv
22nd December 2004, 07:45
How do you chose which settings to use for each codec?
temporance
22nd December 2004, 09:27
Originally posted by FXsan
A consensus seems to be that an equivalent quality is obtained in MPEG-4 ASP with 30% less bitstream than MPEG-2, and in H264 with 50% less than MPEG-2.Not a consensus -- this is what the proponents of H.264 say. In my experience, both MPEG-4 ASP and H.264 achieve similar quality at ~50% MPEG-2 bitrate. ASP and H.264 produce different degradations so personal preference becomes important when comparing these two. At the high bitrates you mention, the differences become smaller.
Teegedeck
22nd December 2004, 10:21
That about summs up my experiences from testing AVC vs. XviD, too. Both deliver 1:1 copies at about 50-40% of the original MPEG2.
So I would expect that the results of Doom9's test heavily depend on the compression ratio he chooses.
SeeMoreDigital
22nd December 2004, 10:39
Originally posted by Teegedeck
That about summs up my experiences from testing AVC vs. XviD, too. Both deliver 1:1 copies at about 50-40% of the original MPEG2.
So I would expect that the results of Doom9's test heavily depend on the compression ratio he chooses. What would be interesting to see is how well both Mpeg4/AVC and Mpeg4/SP/ASP cope with high-def 720p and 1080i sources. And what sort of detail the two can preserve at the same bit-rate.
Sadly, at the present moment, AVC encoding can be a rather slow affair, even when encoding from Mpeg2 DVD sources... So the thought of generating an full HD movie in AVC might prove too much for most people!
Cheers
daveidmx
22nd December 2004, 17:15
if anyone wants to play around, i have a 3CCD DV camera that does telecined 24p video
Neo Neko
22nd December 2004, 20:36
Originally posted by Doom9
And no, I don't have a DV cam and no plans to buy one in the immediate future. Buy something that records interlaced? Shudder. Interlaced is evil ancient technology that we ought to have kicked out of our homes a long time ago.
Doom.... Interlace is not going away. The highest high definition profile is 1080i. I meaning interlaced. And as far as DV cameras go. You can purchase non-high def progressive camcorders. I think a DV test of codecs would be interesting myself. But not pertinant to a DVD shootout. ;) Perhaps for another time and another person.
Doom9
22nd December 2004, 20:57
The highest high definition profile is 1080i.actually there's 1080p, with a 30fps max (it's right here: http://forum.doom9.org/showthread.php?s=&threadid=76655 ;)). But when we're talking about movies 30fps is more than sufficient.
I've been thinking about getting a DV camera, but there's few progressive ones yet and I'm definitely not going to get an interlaced model. I rather have no camera than a crappy interlaced one.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.