Slashdot Mirror
Video Codec Comparison
FonkiE writes "Doom9 wrote a good article: After more than 3 weeks of work and no free time during that period it has been done: The latest codec comparison is online. 7 codecs have been put through one of the hardest tests in the history of codec testing. The results: find out on your own ;) I had planned to change the presentation somewhat but certain events (forum problems and such) prevented me from completing this for the release. I plan to eventually supply an updated version of the comparison."
there is lots of things left out like bitrate/quality comparisons, some codecs, like realvideo do a much better job at low bitrates (200k/s) then say xvid at the same bitrate.
I keep hearing good things about Bink. Anyone have any experience with it, one way or the other? Seems to be used in a lot of games.
A signals-processing attempt to measure audio quality isn't useful in general, and especially when dealing with lossy codecs. The various measured distortion values aren't really interesting -- the only relevant result is audio quality. As such, the only interesting tests are blind listening tests.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
Well, if you have read the article, you can see that 3ivx (not 3ivX as the article states :) ) does not fare well. However, 3ivx does have one thing that the others do not have whatsoever... it was built from scratch for QuickTime compatability. The reason that this is a good thing is the versatility you can achieve with a QuickTime movie. I have personally ripped and encoded an anime movie, and was able to put both English and Japanese, as well as English subtitles, all controlled by a flash menu. The few OGMs I have seen have similar capabilites, but nothing quite as nice as QuickTime.
The video quality is actually pretty damn good, IMO. I suggest trying it out for yourself. Check my webpage for more relevant information.
I wish they tested that too, the encoder isn't free, but it is cheap. Then the question is, is the price worth it?
GPL Deconstructed
If you wanted to make video files that will have the best chance of being viewable in 10 or 20 years, what are the best file formats and codecs?
Are any file formats and codecs likely to be visible?
All these years and MPEG1 is still the only truly universal video format.
.mov or .avi you downloaded requires a codec you don't have.
1. MPEG1 is not encumbered by patent problems as MPEG2 and 4 are. http://www.mpegla.com Thus it is effectively free-as-in-beer by default.
2. MPEG1 is playable everywhere from the old Solaris and SGI boxen to the newest PCs.
3. No finding out after the fact that the
4. It is not a tool to let Microsoft, Sorenson, or Real dominate all online video and divide the web into gated communities.
I have a problem with re-compressing an already compressed source. MPEG-2 for DVD can run between 2Mb and 9Mb (8Mb realistically). I have always been a believer in the garbage in/out theory of compression (please don't make me explain). Each of these titles I am sure were encoded at diffrent bit-rates. I dare to say, the cleaner the source, the better the compression. In this instance I believe the best compressor is the one that can best smooth out the errors of MPEG2, not necessarily the best compressor. I would like to see how these do off of uncompressed source media.
I've been semi-following XivD for about a year, occasionally compressing one of my DVDs to see how it's doing. (which always seems to be: Great, but better the next week (i.e. a severe double-edged sword!).
One thing you know about Xvid is that those problems (the ones Doom9 found) will get addressed. Cheers XviD team.
This is my Sig, this is my Gun. One is for Slashdot and one is for Fun.
Myself I get most of my anime thanks to the existance of these compression formats. I'm a fan of DivX 5 simply because it's easy to get a hold of the codec and it works for all my historical DivX 3.x movies. I have nothing against the Xvid team but it's a bit annoying when I get the occational anime that won't play on anything else but the Xvid codec, requiring me to find it and install it into my system. And as the reviewer noted it's got a few bugs still. DivX 5 may not be the best but it's certainly the most convenient and standard to an end-user.
IANAL but looking at the Helix binary EULA there seems to be a clause disallowing this sort of thing.
c kw rap/eula-clickwrap
https://reguseronly.helixcommunity.org/2002/cli
Entry 2(a)(vii)
You may not make available to any third party the results of any evaluation or testing of the Software by You under this License. Any such forbidden use shall immediately terminate Your license to the Software.
Just a thought
The main issue missing here is that the various video codecs, conversion tools, and audio processing tools are not easy enough to use.
Ease of use should be the focus of new releases of virtualdub, tmpenc, flaskmpeg, avisynth, etc.
The format war continues, but ultimately the majority of the codecs are trying to implement the mpeg4 standard. Surveys and comparisons aside, if you trawl around you'll find that most "users" (ie not companies) are using either DivX3/5 or XviD.
Some clever individual has come up with ffdshow, which you can get off doom9.org, which will play either DivX or XviD without having either codec installed on your system. And at around 500k, it's a smaller download than either of those codecs
"By Grabthar's Hammer, what a savings."
For whatever reason, some programs mess up the spacing of the video and sound streams, for example, Variable Bit-Rate Audio often gives problems. The thing is that it isn't the video codec itself, just the delays getting sound and vision to run concurrently.
See my journal, I write things there
I don't know about anyone else, but for the most part (with the exception of 3ivX) I didn't notice enough image degrigation to matter. Especially at 24fps.
Don't believe me? Try this: scroll through the screenshots (at about the rate of 1 image going from the bottom of the screen to the top per second - 1fps) and tell me if you can pick out the glitches in most of these codecs.
What's more, if anyone was walked into several rooms in sequence, all playing the same movie, but one being DVD, one being DivX3, one being WMV9, etc. I suspect nobody would be able to distinguish one from the other, provided they're encoded at one of the higher quality settings - even if they're intimately familiar with the film.
This is a load of garbage. DVD is a broken codec to begin with.
~/ssh slashdot.org ssh: connect to host slashdot.org port 22: too many beers
since MPEG2 requires such a high bitrate for decent looking video to emerge, I can give you five times better quality (for same size video file) using DivX ;-) because Divx ;-) is many times more efficient. Also note that MPEG2 is quite limited, doesn't work nicely with high resolutions, etc.
The trick to good encoding is in knowing how to use NanDub (or VirtualDub, or VegasVideo, or whatever you use..) and knowing ALL of the various little settings that can be tweaked. The default settings INVARIABLY SUCK, with the quantizer matrix being set for fast, crappy quality video output. The max. quantizers (if using SBC or DivX5) should NEVER exceed 9, for example; however, the default is usually 31! You shouldn't use the default settings and still expect near-DVD quality.
I'm writing a proper, updated tutorial on SBC and Divx 5 Pro right now, which I will submit to various newsgroups when it's done. In the meantime, check out Doom9.org for a rough idea of how to rip properly. They ignore some pertinent details (ie, filters - esp. contrast filter) but you should get a very good idea of the work involved to produce a decent rip.
I noticed even in this article, the WM9 encoded shots look pretty damn good compared to the others. I think the author may have been (possibly subconsciously) downplaying how good WM9 is because of the M$ connection.
I'd love to see something as good on Linux. Anyone know if you can use mplayer to encode using the WM9 encoder?
5x better than MPEG-2? Not a good MPEG-2! A cutting-edge MPEG-2 encoder (try Canopus ProCoder) using 2-pass encoding, and taking advantage of the modes than downloadble files support (like square pixel progressive encoding) could give pretty good results at the data rates the article is looking at - certainly nowhere near 5x worse. MPEG-2's biggest limitation is lack of a deblocking filter, so distracting block artifacts start showing up a lot below a certain point.
At NAB I demoed a 1280x720 24p 4 Mbps MPEG-2 file, and it looked pretty darn good. Not as good as the best MPEG-4 advanced simple, but it wasn't night and day, more like 2pm and 4pm.
My video compression blog
Unfortunately, it appears that Ogg Theora development is "mostly dead". The main developer has been stuck doing contract work (on the integer decoder for Vorbis, as far as I can tell) and can't get to it "for the foreseeable future". The mailing lists are almost completely dead, and, most tellingly, Xiph hasn't updated the theora.org page since January.
I doubt very much they'll have the 1.0 release next month as they have been saying since last June that they'd do...Alpha 1 was looking really promising, but Alpha 2 got pushed back twice (originally scheduled for early December 2002...then late December...then they stopped talking about it anywhere.) Last I'd heard was they were planning to skip Alpha 2 and go straight to Beta in March. Obviously that didn't happen. I do know Monty managed to get some (non-Theora-specific?) work done that will benefit Ogg Theora, but that was back in February, and nobody's talking about it since then.
There are hints that there are other people puttering with the code a little (and VP3 decoding support [the "video codec" part of Ogg Theora - I gather there are still a few "tweaks" to be worked out to turn VP3 into "Ogg Theora"] is slowly being worked on for ffmpeg, Xine, and MPlayer.) but I don't know if Xiph has enough attention on it to get anything out. (Support for VP3/Theora video codec going into Xine is mentioned - very briefly - in the latest "Ogg Traffic" newsletter which at least indicates SOMEBODY remembers that Theora exists. I think if they at least got out some documentation on the format (particularly the .ogg part - they say .ogm is 'horribly hacked' but until there's a "proper" standard available for people to work to, that's all we have for "video-in-ogg") it would help. (If encoding support for Theora in ffmpeg/mplayer isn't far behind, then adoption and work on it outside of Xiph will probably pick up pretty quickly.)
Kinda sad to see the project languish silently as it has for most of the year - some days I can't tell if Xiph will be abandoning Ogg Theora or ever getting back to it or what...
As a side note, back on the topic of "codec comparison", my playing with the one and only release of Ogg Theora way back when it was released (8 months ago!) gave me the impression that it can be a very nice format, especially for more compressed bitrate. Where most codecs seem to get "blockier" as they compress, VP3/Theora seems to get "blurrier" instead, which to my eye generally "looks nicer", despite the fact that it has lost as much actual information from the video as the "blockier" codecs (e.g. mpeg4). IF Xiph ever gets around to some file format documentation and VP3/Theora encoding support appears relatively soon, I can easily imagine Ogg Theora becoming a popular format for internet video and archiving home video.
Hacker Public Radio is our Friend