diff options
| author | Alex Sukhanov <alx.sukhanov@gmail.com> | 2013-11-14 17:56:36 -0800 |
|---|---|---|
| committer | Michael Niedermayer <michaelni@gmx.at> | 2013-11-15 11:54:02 +0100 |
| commit | 86b3435fc01049cfa9dc0257ddbf1f531f7cb67a (patch) | |
| tree | dc8492fd5fc0d0b7b8e39b112da78c864c66c893 /libavcodec/lsp.h | |
| parent | 995f450b44be02cc159ac825f038309e7a3de51e (diff) | |
| download | ffmpeg-86b3435fc01049cfa9dc0257ddbf1f531f7cb67a.tar.gz | |
af_aresample: Fix timestamp of first padded PCM audio packet
Problem:
ffmpeg generated video file which had two audio packets with the same timestamp: last original audio packet and first padded audio packet.
Timestamp of first added audio packet by 'apad' fitler had the same value as last original audio packet. The problem was in 'aresample' fitler, which used next pts instead of current one.
As long as 'apad' and 'aresample' filters have separate mechanisms of timestamp calculation, they got the same values.
Command line:
ffmpeg -i <input_filename> -shortest -apad 512 -af asetnsamples=n=512 -b:a 1058400 -ac 1 -ar 44100 -async 0 -acodec pcm_s16le -sn -f matroska -y <output_file>
Fix:
Call swr_next_pts() function before swr_convert()
Tested:
FATE tests passed.
Fix has been tested in our Transcoder regression framework on ~10k test videos. It's about ~500k transcodes.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavcodec/lsp.h')
0 files changed, 0 insertions, 0 deletions