Theora 1.1 Thusnelda Is Released

[blindfrog]

Is the container different for Theora and Vorbis? Did you mean the preferred file extension?

Same container different name.

[L33F3R]

No. Google wanted h.264 for HTML5 fot it's awesome quality/size and hardware acceleration availability.

i did not mean google's use of it. I should have removed that part of the quote.

[dashcloud]

Excellent! Free multimedia FTW! Now would be nice to see some unbiased tests comparing h.264, xvid, divx and theora thusnelda.

BTW for Theora videos the preferred container is actually ogv and ogg is for audio because it's well always been so no point to confuse peole and change it for example to oga. They're both technically same it's just for distinction.

I really like Theora and what it stands for, but putting h264 and Theora in the same test isn't really fair, for one reason: you'll draw a comparison to x264, quite possibly the best h264 encoder available.

The most recent comparison I've seen is the animation encoding test here: http://x264dev.multimedia.cx/?p=102
Once you've read the introductory text, check the graph to see the results of the test.

[whizse]

That (animated cartoons) looks like a very good example of what Theora still has problems with, "flat color surfaces and gentle gradients tend to disintegrate into an obvious/noticable pattern of flickering square blocks, even at high bitrates":

http://web.mit.edu/xiphmont/Public/theora/demo.html

[drag]

I really like Theora and what it stands for, but putting h264 and Theora in the same test isn't really fair, for one reason: you'll draw a comparison to x264, quite possibly the best h264 encoder available.

The most recent comparison I've seen is the animation encoding test here: http://x264dev.multimedia.cx/?p=102
Once you've read the introductory text, check the graph to see the results of the test.

The graph is worthless.

The problem is this:

For testing, I?ve chosen SSIM as the quality metric, since subjective testing is a nightmare and rarely gives solid numerical results anyways.

The guy does not get what matters when comparing results. The problem is that numerical results are immaterial. When looking at lossy video and audio the ONLY thing that matters is subjective testing. Lossy compression is psychotropic (not sure if it is the right word) in nature, meaning that they are purposely destroying or removing information that is not relevant to a human's perception.

That is it is entirely and 100% possible to win on a numerical result and be a inferior codec. Things like PSNR (or SSIM) are useful for programmers tuning a codec by comparing successive runs of a encoder, but that is about it.

Since your dealing with how humans perceive visually the only valid and scientific way to compare results is by a double blind study.


But since that is excessively expensive and difficult for this sort of thing your mostly going to have to make do with your own subjective judgement. The guy you linked to should of just stopped with the clips. He did a good job there. The SSIM analysis and graphs are misleading and, frankly, a waste of time.

Although it is interesting on a purely trivia level.

I am not saying that Theora is better then H.264.. there is no way that Theora will ever beat H.264 in a fair comparison.

[some-guy]

I am not saying that Theora is better then H.264.. there is no way that Theora will ever beat H.264 in a fair comparison.

Why?

Theora doesn't have hardware accelerated decoding, so performance can't be compared.

There will never be a fair comparison until theora get hw accelerated decoding

[RealNC]

If Theora can do what H.264 (with x264) can currently do, namely compress an 1080p BluRay to 4GB and still look awesome, then it might be accepted as a true alternative. It's not unlikely; Vorbis beat MP3 and matched ACC. Now it's Theora's turn?

[Apopas]

Vorbis beat MP3 and matched ACC.

With aoTuV we get better results than AAC.

[drag]

Why?

Theora doesn't have hardware accelerated decoding, so performance can't be compared.

There will never be a fair comparison until theora get hw accelerated decoding

The only performance I care about is bitrate to perceived video quality.

As long as the system is fast enough to play what I want it to play then I don't care about hardware acceleration, if the system starts choking then obviously thats bad.

[drag]

Here is how you can do about as close to a double-blind test on your own:


Directions:
* make sure you have mplayer installed.

1. copy this script to /tmp and name it "comparison.sh" and make it executable

#!/usr/bin/env bash
 
CDIR=$(pwd)
 
mkdir links
 
find ./ -maxdepth 1 -type f | while read i
do
   TF="$(tempfile --directory "${CDIR}"/links/)"
   rm "$TF"
   ln -s "${CDIR}/$i" "$TF"
done
 
find ./links|sort|while read i
do
   mplayer -FS "$i" > /dev/null
done

Be careful with that script.


2. Make a new directory called /tmp/test

3. Copy all the media files you want to compare to that directory.
A older Theora comparison can be found here if you can't find a better comparison:

http://people.xiph.org/~greg/video/ytcompare/comparison.html

4. Change to that directory and run the script like such:

cd /tmp/test
../comparison.sh



------------------------



How that works is that it creates a /tmp/test/links directory and creates randomly named symbolic links back to your files in /tmp/test directory. Then it has mplayer play them in alphabetical order according to the randomly generated names.


If you want to you can replay them over and over again using:

find links |sort|while read i; do mplayer -FS "${i}" > /dev/null ;done


This is nice because since mplayer is command line and your supressing output then you should have no way to know what codec or rate is encoded any file. (Other then you viewed the files prior the test and recognized visual artifacts that indicative of one codec or another)

Then you record your perception of quality according to the order you played them.


Then after your satisfied you know how the files rank in your comparison you can reveal the codecs and rates by using "ls -l links".

That way, as long as your honest, you can get about as close as a real comparison as your going to get. And since it matches up with how you percieve stuff then it will show you the best choices for yourself.



-------------

Oh. And the above works really well for audio files, too.

Although for audio files you will want to throw in the original uncompressed (or lossy flac comparessed) into the comparison.

People have shown that after listening to a particular type of audio compresson they begin to favor the "sound" that that codec tends to produce. For example mp3 tends to make the "highs" stand out, and people begin to favor that. So you can use this technique to "reset" your ear so you know what is actually more accurate sound reproduction, if you care about that sort of thing.

After all its all reproduction and will never accurately show the original performance... so the best sound quality is always going to be "live". Unless your a fan of Britney Spears or other pop acts since they actually just play back recordings at concerts!