Musepack
Musepack, formerly known as MPEGPlus, is a lossy audio format that is strongly based on MPEG-1 layer II (mp2) but has evolved far past it. It was created by Andree Buschmann for one purpose: quality, and it has one of the finest tuned psymodels for lossy audio in the world. Musepack is a VBR only encoder like Vorbis and has absolutely no way of forcing an average bitrate on it. Musepack also has the fewest problem cases among all of the modern audio codecs. While Musepack is pretty much a dormant project right now (although mppenc 1.15s, 1.15t, 1.15u, and 1.15v have been released recently) there is hope that it will continue improving and hopefully fix its biggest flaws of being unable to be muxed with video and its slow seeking (it seeks about as fast as mp3 though). Musepack files get an .mpc extension but the old .mp+ extension is still valid. Musepack is now open source unlike in the past, but also as of now has no portable hardware support. It is on the Rockbox list of formats for the H100 series, and can be played back at realtime now.
Good Musepack encoders: Pretty much any of the SV7 encoders
Bad Musepack encoders: None that I know of
The current StreamVersion is 7, which is the container. SV8 is hoped for in the not too distant future. A tool that will allow you to remux current SV7 files without loss into an SV8 container has been promised once SV8 is complete. As of now Musepack only supports 1 or 2 channels so it is not for use with multichannel.
Supported input formats (SV7):
Channels: 1 or 2
Bit depths: 1 to 32 bit linear PCM
Sample rates: 32 kHz, 37.8 kHz, 44.1 kHz, 48 kHz (44.1 and 48 are highly tuned)
Musepack like Ogg Vorbis has a quality slider. Qualities 1 through 10 are valid. --quality 5 AKA --standard is the setting that was created for transparency. Going over --quality 6 AKA --xtreme is considered overkill. The --xlevel switch was once recommended to reduce clipping artifacts induced by encoding but as of 1.15s it is hardcoded into the encoder. It will be completely unneccessary once SV8 is done.
Profile Options (Quality Presets):
Code:
--telephone or --quality 1 lowest quality, (typ. 32... 48 kbps)
--thumb or --quality 2 low quality/internet, (typ. 58... 86 kbps)
--radio or --quality 4 medium (MP3) quality, (typ. 112...152 kbps)
--standard or --quality 5 high quality (dflt), (typ. 142...184 kbps)
--xtreme or --quality 6 extreme high quality, (typ. 168...212 kbps)
--insane or --quality 7 extreme high quality, (typ. 232...268 kbps)
--braindead or --quality 8 extreme high quality, (typ. 232...278 kbps)
I personally recommend --standard or --xtreme. --insane and --braindead are completely overkill as well as --quality 9 and --quality 10. Decimal qualities are allowed like --quality 5.25 like Ogg Vorbis. Also Musepack is the fastest of the the modern lossy codecs on both the encode and decode side of things.
Musepack uses APEv2 tags by default but APEv1 and ID3v1 tags are also valid. ID3v2 is not recommended for use with Musepack.
Musepack also has a different kind of version numbering system.
1 decimal point (i.e. 1.0, 1.1): stable
2 decimal points even (i.e. 1.12, 1.14): beta
2 decimal points odd (i.e. 1.13, 1.15): alpha