Lossless encoders

Most compression smallest first

  1. Lossless Audio Info and download, Odin compatible and Linux, least player support best for archive, 2 Gui Windows apps, plugins Win/Lin
    Alternate download Rarewares || Original site
  2. OptimFrog slowest, slightly more players
  3. Monkey's audio much quicker, setting: Extra High, good player support, most pleasant sound with very small file size.
    Unix port || At Sourceforge
  4. Tak very fast, just behind Monkey in size, limited player, source included, Winamp plugin. Impressive!

Quick encoding

  1. True Audio TTA good for multi-channel audio, similar file size to Flac exceptionally quick || Sourceforge dl WLM os, plugins
  2. Tak Tom's Audio Kompressor see above
  3. WavPack almost the same as 1 and 2. Can be lossy,lossless or hybrid (2 files)
    Rarewares dl Win/Lin, also Flake/Flac
  4. Flake compress more and faster than original Flac

Other lossless

  1. Flac reasonably quick, file size large. Issue with files that could be played but not decoded to wav. Mplayer will crash on some later flacs, yet older Mplayer version plays. Can not decode from one version of Flac to another sometimes.
  2. Halac multi-thread encoder should be super fast. Untested, requires files from Itunes to function. Very little player support, new codec.
  3. WMA lossless encoder only Written outside of Microsoft, command line. Creates largest file size, nothing special to use for.
  4. Other: SAC ultimate compressor for size, take nap when encoding though. Purely theoretical at present. Mpeg4-ALS where is executable? No special feature. Hardly used. Alac Apple only, large file size. Apple must have it's own version, unneeded.

Lossy encoders by preference

  1. Musepack MPC, most musically balanced lossy compared to WAV. When using my method, prefer listening to MPC over WAV. Small file size. || Gui for MPC
  2. Opus examples similar size to MPC, 63% slower encode uses 48khz, excellent at small bitrates.
    Recommended settings || HydrogenAudio settings
  3. Ogg Vorbis Long time favourite. MPC has better sound/file size, replaced Vorbis for music. For spoken voice, audiobook/interview still great. See below for parameter setting.
  4. AAC at Rarewares Improvement over mp3, MPC still rocks!
    Qaac higher quality need later Windows version || Fraunhofer FDK AAC quality over Faac
  5. Helix mp3 faster than Lame, simple parameters. Helix unix || Helix at Rarewares. Also GoGoNoCoda (lame enhance with assembly code) fast, used GoGo before wiped MP3 off the disk.
  6. WMA from Microsoft, no thank you

Myth vs reality

Comments often made on the Web that bitrate equals quality. Which encoder used is more important and most encoders are tuned for specific bitrate range. 224kps for Mp3, Ogg, Aac sufficient, higher rate only adds to file size, little improvement in sound. Vorbis can go as high as 500kbps, yet ideal is 192-224.

Flac vs Monkey, Ape encoded with similar compression to Flac is close in speed. Extra compression takes more time. Decompression speed largely irrelavant with today's cpu.

From ripped WAV to completed codec

Using your favourite cd ripper, create a WAV file. All process below is done in a RAM disk, files only written to disk at completion.

e:\bird\waveclip %1 %2 z.wav
e:\bird\cdburn\wav2cdr --towav --fadeout 80 z.wav -w 9
ren 9.01 y.wav

Waveclip OS/2 executable %1=Wav file %2=time length. What is a Linux equivalent?
Wav2cdr from CdRecord or CDRtools package, can use --fadein, --fadeout 80 equals a second in time. -w add beginning of temporary file name.

Result:
e:\bird\waveclip file.wav :3-4:26.50 z.wav

Cut file at 3 seconds, end at 4 minutes 26 and half seconds. File z.wav is time edited!

z.wav with 1 second fadeout renamed to y.wav. Choose Y or Z and rename to original WAV. Delete the other. Wav2cdr can also cut either end of file.

Next is to Deglitch or remove damaged samples. Sometimes none, typical 10 samples or less. The most was 63. Windoze tool Odin/Wine functions.

d:\odin\system32\pec d:\odin\system32\deglitch -y %1

Using Shibatch tool normalize volume to 1.1 decibel below maximum. -twopass likely unnecessary.

ssrc --att 1.1db --twopass --normalize %1 z.wav

Sox tool modify pitch from 440hz to 432hz.

sox z.wav %1 pitch -.3176665363342928

File ready for encoding. First is Musepack

d:\odin\system32\pec d:\odin\system32\mpcenc --silent --quality 5.35 %1

Removed --silent parameter to show statistic.

Result:

MPC Encoder 1.30.0 --stable-- (C) 1999-2009 Buschmann/Klemm/Piecha/MDT
Built Feb 26 2009 20:33:36

 encoding file 'file.wav'
       to file 'file.mpc'

 SV 8, Profile 'Standard'

    %|avg.bitrate| speed|play time (proc/tot)| CPU time (proc/tot)| ETA
100.0  186.7 kbps 18.66x     3:59.0    3:59.0     0:12.8    0:12.8

Finally, Monkey Audio APE is created, temporary files cleaned up.

mac %1 %1.ape -c4000
ren *.wav.ape *.ape
del z.wav 

--- Monkey's Audio Console Front End (v 3.99) (c) Matthew T. Ashland ---
Compressing (extra high)...
Progress: 100.0% (0.0 seconds remaining, 19.1 seconds total)
Success...

Result:
file.ape 30538104
file.mpc 5576579
file.wav 42159648
Musepack for play, Monkey's Audio to archive

Other settings

You will find songs where the volume is outside normal, too low typical from the 1970's or latest songs too loud. Normalize can not adjust enough. Solution is manually adjust volume by using:
e:\bird\cdburn\wav2cdr --towav --fscale %2 %1 -w 9
ren 9.01 z.wav
:: range .7 to 1.7, .85 for lower typically sufficient

Parameter %2 is scale adjust, %1 is file name. You can add the --fadeout parameter. Usually 1950-1960 era is normal, 1970's quiet, 1980-2000's regular, 2010-onward has some crazy loud files.

Audio files that are only spoken word can be created exceptionally small and sound acceptable. Use Ogg Vorbis with:
oggenc -q1.4 --resample 12000 --downmix -o file.ogg file.wav

Vorbis quality range is -q-1 -q0 to -q10. You could have -q3.78 if desired. -q1.4 equals approximately 25 kbps, 12000 is hz, downmix is mono. All you need and super tiny.

LossyWav to lossless encoder

Very interesting project. LossyWav is a pre-processor. It scans file in blocks of 512, finds unused bits, marks as "wasted bit". Next, use a lossless encoder that will recognize "wasted". Most desirable codecs are: Tak, Flac and WavPack. 
(16 bit / 44.1kHz)
lossyWAV Test Codec lossless insane  extreme high    standard economic portable extraportable
10 Album Test FLAC  854kbps  627kbps 548kbps 477kbps 442kbps  407kbps  353kbps  311 kbps

As the list above displays, can have as much lossless/lossy as desired. -Insane not recomended, -Extreme is favoured for lossless, -Standard or -Economic for better lossy. File size with -Extreme lossless about 30% less. Significant. With all the encoders, use the parameter to set block size at 512 to match LossyWav. Otherwise the small file size is not achieved. Flac works, Flake ignores wasted bits same size as normal.

Some caveats: since WavPack can be lossy, better to use it's routine for smaller lossy files. Great for lossless though. As usual TAK compresses the most, is the speed king, and uses the least CPU cycles.

If using -High or -Extreme parameter, easily transcode to any other codec with same sound. Is bit-identical to a WAV file? No! Will you ever hear a difference? Doubtful on any hardware as wasted bits are just that. Wasted!

If concerned about bit identical WAV, LossyWav has the feature to create a correction file. Merge the 2 files back to an original WAV.

This page is at: Encoder lossless/lossy list

Enjoy your musical experience! The sound is more important than numbers on a test.