On 07.11.2013 04:53, [email protected] wrote:
Hello! I've been using libavcodec 0.10.2 for decoding audio in a
certain application (calling avcodec_decode_audio4() to use the
"mp3" and "mp3float" decoders) and now I've tried to upgrade to
version 2.0.2. Everything builds fine without changes in my code,
but for some reason I get Mickey Mouse kind of sound after
decoding. I've tried other releases and come to the conclusion
that release 1.0.8 works straight off, but release 1.1.x (I've
tried 1.1.7) and later will produce this effect.
From reading the "Old FFmpeg Releases" page on the website I can
see that 1.0 was branched on 2012-09-28 and 1.1 on 2013-01-06. So
the change I'm looking for is probably somewhere between those
dates. But I also see that the 1.1.7 release use libav 9.10 and
1.0.8 use libav 0.8.3. What does that mean, which versions are you
referring to? Where can I find out what the differences (or rather
"changes of concept") there are between these "libav" versions?
Has there for instance been some change in the sample format,
sample rate or whatever between these version, so that I need to
call avcodec_open2() with some more or less undocumented option in
order to get the same behaviour as before?
I've tried to find release notes or any kind of documentation
where I can find out which changes were implemented and what I
need to do to cope with this. Could anybody please spread some
light on what might have changed, not the whole changelog for each
and every change of a comment or whatever but for a broader
picture. Is something like that available?
Better yet, perhaps somebody knows exactly what I'll have to do to
avoid this effect?
BR,
Mathias
Sorry Mathias,
You will have to make changes because the decoding API has changed.
See audio decoding example in examples directory. The major
difference is that what is contained in a decoded frame is not 16-bit
linear interleaved samples. The output is now codec dependent. You
have to use the API to get the sample format, and planar/not planar.
avcodec_get_frame_defaults
<http://ffmpeg.org/doxygen/trunk/group__lavc__core.html#ga1cb4e0fd7b8eb2f56d977ff96973479d>(decoded_frame);
avcodec_decode_audio4
<http://ffmpeg.org/doxygen/trunk/group__lavc__decoding.html#ga834bb1b062fbcc2de4cf7fb93f154a3e>(c,
decoded_frame, &got_frame, &avpkt);
if (got_frame) {
/* if a frame has been decoded, output it */
int data_size = av_samples_get_buffer_size
<http://ffmpeg.org/doxygen/trunk/samplefmt_8c.html#aa7368bc4e3a366b688e81938ed55eb06>(NULL,
c->channels
<http://ffmpeg.org/doxygen/trunk/structAVCodecContext.html#ac1e6c2cd1269caa7570575725c682a49>,
decoded_frame->nb_samples
<http://ffmpeg.org/doxygen/trunk/structAVFrame.html#a02f45ab8191aea1660159f1e464237ea>,
c->sample_fmt
<http://ffmpeg.org/doxygen/trunk/structAVCodecContext.html#a1bdba69ea111e2a9d03fdaa7a46a4c45>,
1);
}
If you are only dealing with one codec you can modify your code to use
this output sample format directly. Otherwise use libresample with in
rate = out rate to convert from c->sample_fmt to desired sample
format. See samplefmt.h and swresample.h.
Hello,
Most Audio outputs Plananar stuff now a days and it sounds like Mickey.
Like abo said MP3 outputs S16P and Ogg/Vorbis FLTP. So just playing them
like 0.10 is out of question witth out major source overhaul. When I
developped Mixxx input plugin I also suffered this one and broken out
little example app to solve this problem with resampling audio to wanted
(As FFmpeg has real good capabilities to do that). I uploaded it to
Github: https://github.com/illuusio/ffmpeg-example
Look at 'example2.c' there is this dirty resampler example that can use
swresampler/avresampler which is provided. It should work from FFmpeg
versions 0.10->2.1.
Sincerely,
Tuukka
_______________________________________________
Libav-user mailing list
[email protected]
http://ffmpeg.org/mailman/listinfo/libav-user
