I use the nellymosser decoder and it works quite well except in one case. It seems that this codec does not allocate datas for white spaces. So when i encode files with "no sound" in part of the video, it packs all part of the video sound at the beginning of the video. If i have a 30 sec video with a blank running from second 8 to 16, So, something like that: 1 5 10 15 20 25 30 |----|----|----|----|----|----| Video: vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv Sound: xxxxxxxx yyyyyyyyyyyyyy The recoded video with: /usr/bin/ffmpeg -i in.flv -acodec libmp3lame -ar 44100 -ab 32 -r 30 -f flv -y out.flv Will have those datas 1 5 10 15 20 25 30 |----|----|----|----|----|----| Video: vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv Sound: xxxxxxxxyyyyyyyyyyyyyy Furthermore, the playback of the generated flv seems to have problems. (hanging, slow down & so on)