Re: [Audacity-devel] Patch for bug 421: Freeze when libmad takes PCM audio

[Michael C. <mc...@gm...>]

On Mon, Jul 18, 2011 at 10:01 PM, Richard Ash
<ric...@go...> wrote:
> On Mon, Jul 18, 2011 at 2:04 AM, Michael Chinen <mc...@gm...> wrote:
>> The main question is if we want to give ffmpeg priority over libmad.
>> I don't have a problem with this, but I am biased towards ffmpeg.  I
>> also haven't come across mp3 cases where libmad handled something that
>> ffmpeg couldn't but this is purely subjective.
>> Does anyone have reasons for keeping libmad above ffmpeg?
>
> Libmad has been in Audacity far longer, so for a long time was the
> only available MP3 importer. The rationale for using it (in preference
> to other MP3 decoders) was mainly one about quality - libmad is a full
> accuracy tested decoder, and is capable of 24-bit PCM output (although
> no-one ever got round to using the latter in Audacity).
>
> So I would like to see some information on the accuracy of ffmpeg's
> decoder before changing the default for Audacity - a fair number of
> users would be upset if we switched to a less accurate MP3 decoder. As
> far as I know ffmpeg can't handle more than 16 bit audio in most
> circumstances.

I looked and couldn't find any info about the accuracy of libmad vs ffmpeg.
I do know some work has gone into ffmpeg mp3 over the last two years,
but I believe this was mostly for speed improvements.

So I guess we can't take the alternative patch unless someone spends
the time to do the tests.  If no one says they want to do this in the
next couple of days I'll commit the original (modified) patch.

Michael