[flac-dev] FLAC Mathematical Details

[Declan Kelly]

On Wed, Feb 08, 2012 at 08:49:37AM -0800, giles at thaumas.net wrote:

> On 7 February 2012 21:59, Martijn van Beurden <mvanb1 at gmail.com> wrote:
> 
> > Flake, another FLAC-encoders can use variable
> > block length and has a algorithm to decide the length, but this is
> > outside of the -0 to -8 presets, as these are all fixed block length.
> 
> How much does the variable-block mode improve compression?

I did a lot of tests about a year ago, and Flake definitely has tighter
compression at the higher presets (they go up to 11, probably because 11
is one louder).

I don't have the exact results to hand, but I do remember that it wasn't
as much of an improvement as I had expected, and it took a lot more time
to compress.
Playback compatibility is not guaranteed with any of Flake's additional
variable-block preset levels. I tested playback, and most software that
I tried wasn't able to decode the tightest Flake compression options.
The dealbreaker for me was no support for audio above 16-bit/44.1k.


I regularly re-compress files to save space, and I have noticed that a
lot of distributed .flac files were not compressed to the maximum level
possible using version 1.2.1 of flac/libflac (or an older FLAC version
was used, or both).

What I should really do is test battery life with different levels of
compression, as for a lot of applications that will matter more than how
full the storage device can get (per hour of stored music).

Latest version (according to the flake-enc website) is 0.11 (2007) but
at least the Flake maintainer is active on this mailing list :-)

-- 
-Dec.
---
"Mosaic is going to be on every computer in the world." - Marc Andreessen, 1994