RealAudio Cook codec

[Mephesto]

Does anyone know the quality standing of that Cooker/Gecko codec used on old RMs? Like, is it better than MP3? My vague recollection tells me it was intended as a speech codec. Is there a command-line encoder available out there?

[El Heggunte]

Well, the "thing" cannot go above 96kbps, so if that's "good enough" for music, depends a lot on the ears of the user

I presume the so-called Helix Producer can be called via command-line; Easy RealMedia Producer would be another option.

[Hoser Rob]

Well, the "thing" cannot go above 96kbps, so if that's "good enough" for music, depends a lot on the ears of the user

I presume the so-called Helix Producer can be called via command-line; Easy RealMedia Producer would be another option.

96K? All right for a table radio I guess. Sure as hell not better than mp3, which I don't consider that great either. I only convert flac/wav music to mp3 or aac if the recording stinks anyway.

I just had a quick peek on the web ...BTW google is your friend ... and apparently it's included in libavcodec, so it'd work with ffmpeg. Don't ask me what the command line would be though.

[poisondeathray]

it's only available for decoding in libavcodec , not encoding

[Hoser Rob]

it's only available for decoding in libavcodec , not encoding

Oops. Oh well.

[Mephesto]

Thanks El Heggunte for Easy Realproducer, it allows encoding with bitrates higher than 96 btw.
Properly decoding it is now a problem though, FFAudiosource doesn't reproduce it faithfully. Directshowsource does but it gets normalized by the LAV splitter. I tried to set FFDshow (libavcodec) to be the decoder where I can easily uncheck the volume normalization but then Avisynth refuses to open it via Directshow.

I don't think a listening test was ever done for Gecko. Upon quick inspection it does seem to be a little better than MP3 but I can't tell without a proper test and this can't happen if one is much louder than the other.

[El Heggunte]

Thanks El Heggunte for Easy Realproducer, it allows encoding with bitrates higher than 96 btw.

Above 96kbps it's not Cook Audio anymore, if I'm not mistaken. Either it's ATRAC3, or AAC.

Properly decoding it is now a problem though, FFAudiosource doesn't reproduce it faithfully. Directshowsource does but it gets normalized by the LAV splitter. I tried to set FFDshow (libavcodec) to be the decoder where I can easily uncheck the volume normalization but then Avisynth refuses to open it via Directshow.

Then you should manually build a GRF file that includes ffdshow, and feed DirectShowSource with it.

I don't think a listening test was ever done for Gecko. Upon quick inspection it does seem to be a little better than MP3 but I can't tell without a proper test and this can't happen if one is much louder than the other.

Another codec that has been "overbashed" IMHO is ATRAC3. I think it's a silly error to still consider stereo MP3 @ 128kbps encoded by LAME as *the paradigm* for ABXing lossy audio codecs, now that storage space is cheap and broadband Internet access is not a "privilege" anymore.

[Mephesto]

You were right, it can't go beyond 96 or it becomes AAC. I've done the test though. Gecko passed with flying colors. It beat the shit out of MP3 and outclassed opus, but wasn't better than AAC. Here are the results.

ABC/HR Version 1.0, 6 May 2004
Testname: Mazgal aac/cook/mp3/opus

1R = M:\mzg96cook.wav
2L = M:\mzg96mp3.wav
3L = M:\mzg64.opus.wav
4L = M:\mzg64cook.wav
5L = M:\mzg64mp4.wav
6R = M:\mzg96.opus.wav
7L = M:\mzg96mp4.wav

---------------------------------------
General Comments:

---------------------------------------
1R File: M:\mzg96cook.wav
1R Rating: 4.4
1R Comment: Minor distortion in the upper shelf and slightly gravelled melody that took concentration to 

properly hear
---------------------------------------
2L File: M:\mzg96mp3.wav
2L Rating: 3.4
2L Comment: Dull and smeared
---------------------------------------
3L File: M:\mzg64.opus.wav
3L Rating: 3.6
3L Comment: No upper cut like 3 but theres noise in the melody instead of smearing
---------------------------------------
4L File: M:\mzg64cook.wav
4L Rating: 3.7
4L Comment: Stereo reduced and some smearing but no upper shelf cut like 2 or noise like 3
---------------------------------------
5L File: M:\mzg64mp4.wav
5L Rating: 4.0
5L Comment: A little bit dull but I couldn't detect any flaws in the melody
---------------------------------------
6R File: M:\mzg96.opus.wav
6R Rating: 4.0
6R Comment: A much reduced form of distortion than 3. The melody is a little pre-echoey but difficult to 

notice. Cant decide if its better than 5.
---------------------------------------
7L File: M:\mzg96mp4.wav
7L Rating: 4.9
7L Comment: Cymbals are louder, not distorted just louder.
---------------------------------------
ABX Results:
Original vs M:\mzg96cook.wav
    10 out of 10, pval < 0.001
Original vs M:\mzg96mp3.wav
    10 out of 10, pval < 0.001
Original vs M:\mzg64.opus.wav
    10 out of 10, pval < 0.001
Original vs M:\mzg64cook.wav
    13 out of 14, pval < 0.001
Original vs M:\mzg64mp4.wav
    10 out of 10, pval < 0.001
Original vs M:\mzg96.opus.wav
    10 out of 10, pval < 0.001
Original vs M:\mzg96mp4.wav
    10 out of 10, pval < 0.001

Cook seems like a really good codec actually.

[Mephesto]

Okay I did another test. Last song was a very tonal tune, this one has heavy percussion covering the entire spectrum.
Cook at 96 kb/s profile actually outputted 106 kb/s and 64 did 70 so I adjusted the other codecs to have the same filesize as the Cook files as I couldn't adjust Cook to precision.

Cook failed miserably this time, being worse than even MP3. I guess I was right that Cook was intended as a speech codec so it handles tonal audio better than broadband.

Interesting to note that opus was very nearly transparent at 70 kb/s.

ABC/HR Version 1.0, 6 May 2004
Testname: grcitrance aac/cook/mp3/opus

1R = M:\grc70.opus.wav
2R = M:\grc106cook.wav
3L = M:\grc70mp4.wav
4R = M:\grc106mp4.wav
5R = M:\grc106mp3.wav
6R = M:\grc106.opus.wav
7R = M:\grc70cook.wav

---------------------------------------
General Comments:

---------------------------------------
1R File: M:\grc70.opus.wav
1R Rating: 4.9
1R Comment: A very difficult artifact to spot in the same place as 4 only it's softer instead of brighter. Reason I had so many wrong guesses was because it took me until sample4 to find out there was an ABX-able spot in the audio.
---------------------------------------
2R File: M:\grc106cook.wav
2R Rating: 3.2
2R Comment: Dulled down, stereo reduced and really smeared
---------------------------------------
3L File: M:\grc70mp4.wav
3L Rating: 3.4
3L Comment: More dull than 2 but stereo better preserved and less smearing
---------------------------------------
4R File: M:\grc106mp4.wav
4R Rating: 4.9
4R Comment: One high note was a bit more accentuated. Was very difficult to first notice.
---------------------------------------
5R File: M:\grc106mp3.wav
5R Rating: 3.9
5R Comment: A mix of smearing and distortion in the upper shelf, but very well blended to resemble original.
---------------------------------------
7R File: M:\grc70cook.wav
7R Rating: 2.0
7R Comment: Severely washed out and almost mono
---------------------------------------
ABX Results:
Original vs M:\grc70.opus.wav
    42 out of 60, pval = 0.001
Original vs M:\grc106cook.wav
    14 out of 15, pval < 0.001
Original vs M:\grc70mp4.wav
    10 out of 10, pval < 0.001
Original vs M:\grc106mp4.wav
    10 out of 10, pval < 0.001
Original vs M:\grc106mp3.wav
    10 out of 10, pval < 0.001
Original vs M:\grc106.opus.wav
    17 out of 31, pval = 0.360
Original vs M:\grc70cook.wav
    10 out of 10, pval < 0.001