[vorbis] as if you haven't seen this yet

[Rick Franchuk]

On Tue, 17 Jul 2001, Oddsock wrote:

> in case anyone is interested, I have built a win32 program which will 
> perform a double blind listening tests..It will allow you to play each 
> sample and allow you to rank them in order of quality.  The program will 
> then send the results to a centralized server which saves it off to a DB 
> and gives immediate feedback to the user the outcome of the test.
> 
> A few things though :
> 
> - I'd like to make it self contained (i.e. with an installer) which means 
> including the wav samples with the listening test. With 3 1 minute WAV 
> samples, that's 30MB..
> - I don't have good samples..I'd like to use samples that show obvious 
> differences b/w each of the codecs...
> - I would like to do a 64kbps face-off with a 64kbps WMA8, MP3Pro, and 
> Vorbis sample.  I can do the WMA8, and MP3Pro, but I need to wait for Monty 
> to finish the RC1 encoder to provide a comparable OGG sample.

Just a thought, but you might want to make it so that the program itself is
small and allows people to download a new sample (or samples) as they run it
the first time. This gives the dual advantage of being a smaller initial
installation (thus not scaring away potential testors) and allowing for
arbitrary samples to be added to the testing regimine as additional
free-for-use clips become available. The size of the .wav is going to be the
same if you take it as part of a package or download later anyways (you'll
probably want to use an entropy encoder on the sample to squeeze it some,
naturally).

As for obtaining samples, I would imagine that there's all manner of house
bands that'd be willing to make a track public domain in order to get some
free publicity (encompassing a wide range of musical styles). Speech clips
would be really easy to pick up too.

If you wanted to get REALLY slick, you could let people stick in thier
favorite CDs and have your program rip a track, allowing the user to rate
based on a musical sample they know intimately. Your program would record what
CD and track it was (using CDDB-ish things?) for us to identify against. I
don't know how useful that would be in terms of double-blind testing, but I
could see it being very handy for identifying tracks where vorbis (or any of
the encoders) have issues. It'd require a patient test agent for sure. Still,
judging encodings of a song you listen to a lot makes it easy to pick out
where things don't sound quite right... ask any Led Zeppelin fan who's
listened to an mp3 of "Been a Long Time Since I Rock And Rolled" and you'll
get definately less-than-optimum reports back ;)

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.