g729 has 10 bytes per packet and g729b has 12 bytes per packet (2 silence insertion descriptor + 10 bytes of data) Sample file contains stream with 11 bytes per packet. ACELP.KELVIN looks like modification of g729. My suspicion is that ACELP.KELVIN uses other coefficients than G.729 or has other small differences. If your file would contain such 2-byte Silence Insertion Descriptor packets, the output file would have to be longer (but it is too short even if the input does not contain any Silence Insertion Descriptor packets). Afaik, the difference between G.729 and G.729B is that G.729B is able to send (via network) the information that the sender is silent which is a shorter package than an actual audio package, this cannot be done in a wav file, see ticket #7604. My question was if you used the file that you attached here to create the output file (that you also attached). Are there any specific way to confirm that? No decoder for stream #0:0, filtering impossible

Guessed Channel Layout for Input Stream #0.0 : stereo Could not find codec parameters for stream 0 (Audio: none (5 / 0x0135), 8000 Hz, 2 channels, 17 kb/s): unknown codecĬonsider increasing the value for the 'analyzeduration' and 'probesize' optionsĦ62478440-short.wav: could not find codec parameters

Estimating duration from bitrate, this may be inaccurate Sample file, original codec and reference output attached.įfmpeg version N-41686-g91459bd Copyright (c) 2000-2014 the FFmpeg developersīuilt on 12:25:50 with gcc 4.9.2 (Ubuntu 4.9.2-0ubuntu1~14.04)Ĭonfiguration: -extra-cflags=-I/home/alex/work/ffmpeg/include -extra-ldflags='-L/home/alex/work/ffmpeg/lib -L/usr/lib64' -bindir=/home/alex/bin -extra-libs=-ldl -enable-gpl -enable-nonfree -enable-version3 -enable-libfdk_aac -enable-libmp3lame -enable-libvorbis -enable-libvpx -enable-libx264 -enable-libopencore-amrnb -enable-libvo-amrwbenc -enable-libopencore-amrwb -enable-libspeex I would be nice to have decoder for this codec.

Ffmpeg concat vorbis encoder error