[FFmpeg-devel] [RFC]lavc/adpcm: Support baby monitor adpcm_ms

[Carl Eugen Hoyos]

Paul B Mahol <onemda <at> gmail.com> writes:

> >> > So how should this be implemented?
> >>
> >> Perhaps predictor, if too big should be just clipped.
> >
> > Why do you think there is a too big predictor?
> > Two additional bytes are inserted at the beginning of
> > every frame.
> > Should the demuxer be patched instead?
> 
> Why you think that?

(You mean that there are two additional bytes?)
Because if I remove the two additional bytes that 
the encoder adds for each audio stream, the file 
decodes as expected.

> There is clearly overread if you ignore
> first two bytes.

How can I reproduce the overread?

Carl Eugen