Hi--
stp writes:
MP3 sounds marginally better than RealAudio compression. Anyone with working ears and headphones costing more than $3 can easily tell the difference between MPs and even a cheaps**t sound card.
Hmm. I think MP3 at 128kbps or higher sounds very much better than RealAudio. The "no loss of quality" claim isn't completely silly (only mostly)... If the source has already been severely compressed spectrally (or just doesn't have any very low or high frequencies) then it compresses very well. :)
I've written a subclass of MixedSound called ExternalSampledSound. It fetches samples from an ExternalStream (on a file or a socket, for example) while playing. It just does WAV decoding currently; it would be strightforward to make it understand MP3 (or QuickTime or whatever).
-C
-- Craig Latta composer and computer scientist craig.latta@netjam.org www.netjam.org latta@interval.com Smalltalkers do: [:it | All with: Class, (And love: it)]
Craig Latta wrote:
Hi--
stp writes:
MP3 sounds marginally better than RealAudio compression. Anyone with working ears and headphones costing more than $3 can easily tell the difference between MPs and even a cheaps**t sound card.
Hmm. I think MP3 at 128kbps or higher sounds very much better than RealAudio. The "no loss of quality" claim isn't completely silly (only mostly)... If the source has already been severely compressed spectrally (or just doesn't have any very low or high frequencies) then it compresses very well. :)
Well, perhaps I was a bit too negative, but especially at the dawn of the era of 24-bit 96 kHZ audio, it's depressing to hear so much hype about MP3. Most of what I've found on the net is heavily dynamically and spectrally compressed. As far as I'm concerned, the old maxim still applies: You can have 3 things in a compression scheme: good compression ratio (better than 1.5:1) real-time unpack (on standard HW) artifact-free compression (to good ears) (pick any *TWO*)
-- stp
Stephen Travis Pope | stp@create.ucsb.edu | http://www.create.ucsb.edu/~stp
...at the dawn of the era of 24-bit 96 kHZ audio, it's depressing to hear so much hype about MP3.
Yep; for me it's another reminder of how slow current networks and memory busses are.
-C
-- Craig Latta composer and computer scientist craig.latta@netjam.org www.netjam.org latta@interval.com Smalltalkers do: [:it | All with: Class, (And love: it)]
At 4:22 PM -0800 4/1/99, Craig Latta wrote:
Hmm. I think MP3 at 128kbps or higher sounds very much better than RealAudio. The "no loss of quality" claim isn't completely silly (only mostly)... If the source has already been severely compressed spectrally (or just doesn't have any very low or high frequencies) then it compresses very well. :)
Has anyone listened to the Voxware music (MetaSound) codecs? To my ears, the quality of their 96kps stereo codec is better than the 128kps MP3, and the sound is very decent even at 48kbps (although that bitrate drops the sampling rate to 22050 with the corresponding loss of those sparkling high frequencies).
But the truely stunning codec I've heard is the QSound one. If you have Quicktime 3.0 or better installed, you can listen to samples of this at www.qsound.com. They can stream CD-quality sound over a 28.8 modem! How do they do it? Apparently by expending a LOT of work in analysis during encoding. On a P133, it takes over an hour to encode 10 minutes of music. But decoding is real-time.
Quicktime includes decoders for Voxware, QSound, and MP3, and QT is available for both Win32 and Mac. Thus, it might make sense to add a Squeak primitive that can access the Quicktime codec library.
Incidentally, Squeak 2.4 includes a SoundCodec hierarchy that currently includes ADPCM (2, 3, 4, and 5 bit), GSM 6.10, mu-Law, and an experimental Wavelet-based codec. But none of these codecs approaches the compression ratios and quality levels of the commercial codecs mentioned above.
Pointers/Compression Expertise Wanted
The Squeak team would be *extremely* interested in pointers to publicly available source code, standards docs, or algorithm descriptions for high quality codecs that we could incorporate into Squeak. Something like the GSM source code from Jutta Degener and Carsten Bormann of the Technische Universitaet Berlin (which we're using for the GSM codec) would be perfect. Or, if any of you Squeakers is an expert on sound compression and might be interested in implementing a codec in Squeak (for translation to C), please let us know. Thanks!
-- John
P.S. The GSM codec is rather large, so it is packaged as a Squeak plugin. Plugins have been compiled for Win32 and MacPPC and the C source code for GSM will be included in the 2.4 release. P.P.S. We already know about the two ADPCM implementations on ftp.cwi.nl/pub/audio/.
johnm@wdi.disney.com wrote:
At 4:22 PM -0800 4/1/99, Craig Latta wrote:
Hmm. I think MP3 at 128kbps or higher sounds very much better than RealAudio. The "no loss of quality" claim isn't completely silly (only mostly)... If the source has already been severely compressed spectrally (or just doesn't have any very low or high frequencies) then it compresses very well. :)
Has anyone listened to the Voxware music (MetaSound) codecs? To my ears, the quality of their 96kps stereo codec is better than the 128kps MP3,
[ ... ]
But the truely stunning codec I've heard is the QSound one.
As per usual, John is right; there are a number of codecs that sound better than MP3 (or sound comparable at lower bit rates). Please also be certain that I never meant (in my earlier negative comments) to say that building better codecs into Squeak was a bad idea. I was merely firmly contradicting the marketing hype that says that MP3 is artifact-free at a 12:1 compression ratio.
With respect to John's call for PD codec sources, the problem is that so much of the good recent work has been at companies who want to license the technology as their business, so it's unlikely that they'd post a reference implementation to the web (or even publish a paper describing their techniques that would make it easy for anyone else to make an external implementation).
-- stp
Stephen Travis Pope | stp@create.ucsb.edu | http://www.create.ucsb.edu/~stp
johnm@wdi.disney.com wrote: With respect to John's call for PD codec sources, the problem is that so much of the good recent work has been at companies who want to license the technology as their business, so it's unlikely that they'd post a reference implementation to the web (or even publish a paper describing their techniques that would make it easy for anyone else to make an external implementation).
Sad but true. I do have a good book ("Introduction to Data Compression") that describes the general approaches used in MP3 and other modern codecs but I suspect that a great deal of careful work must go into tuning the precise parameters and filter coefficients to make a really great codec.
-- John
squeak-dev@lists.squeakfoundation.org