Chris Cannam posted: > Because sfArk and sfpack compress soundfonts much better than > zip/gzip/bzip2 can. This is absolutely correct. The reason is that text files repeat character sequences exactly whenever words and other character combinations are used over and over again. The "normal" compression methods build tables and utilize this precisely repetitive nature effectively. Audio files repeat character sequences only approximately, so are not recognized as being almost the same even though they sound the same, hence are regarded as new sequences. True binary files (e.g. stripped executables) look pretty much random, also, to one of these compression algorithms. The random appearance of audio files is one reason why MP3, sfark, etc. were developed. To distinguish between these latter techniques: As most people know by now, MP3 is a lossy technique which means that information is lost never to be seen again. sfark is not. These soundfont compression techniques are a compromise between loss of information and effective compression. Some out there may be using MP3's in place of such compression schemes as sfark's, believing that MP3's are as good or better due to the good compression ratios obtainable. This is at a fairly heavy cost. I would advise against it generally, which is the reason I'm spending time posting this. MP3 is fine for distribution over limited-resource channels, but not so fine for soundfonts/samples. Hope this helps someone out there. Now I really am going to try to get caught up.... Regards to everyone, Dave.
