The library introduces noise in unclear circumstances. There does not
seem to be any problem on VLC side though. I do not have time and
expertise to troubleshoot this.

Conversion from FI32 to S16 is already supported by the Laurent's
generic audio format filter. Conversion to FI32 are useless for lack
of (functional) FI32 encoder or output.

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

It only makes sense for i_original_channels.

Support for more samplerate than 44100 and 48000
Support for classic ISO bands instead of Winamp ones
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

This module is a wrapper around an audio mixer.
Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

Close #6514

I am not sure this filter ever worked as intended.
It was providing serious distortions.

This is now limiting seriously the space of parameters but at least it
does not explode anymore. Should close #4887
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Although this was not reported, this suffered the same problem as A52.

This closes a heap overflow on corrupt files.
Pointed-out-by: Clément Lecigne <clemun@gmail.com>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

It is good enough for normal people with only half the CPU usage of
medium quality SinC.

Close #6143

This reverts commit 5c261b04.
SRC is CPU usage too damn high!

This filter allows channels to be remapped. Potential uses include
downmixing, or to correct non-standard channel orders.
Modified-by: Jean-Baptiste Kempf <jb@videolan.org>
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

This is probably temporary, since SRC is more computationly intensive than speex_dsp
But this close #5781

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

This should fix buffer underflows when changing the playback rate.
We probably should instead reallocate the buffer and resume resampling
when the situation occurs, but that would be a lot more complicated.

We already have the integer list for that.

On rare/unreproducible occasions, level 3 seem to induce lots of noise.

(VLC simply lost one sample - no memory errors.)