We aim to reduce the runtime of segmentation / coding from wav to aac, segmented for streaming HTTP, using ffmpeg to segment and create a m3u8 playlist, using all the cores of our machine.
In one experiment, ffmpeg directly segmented a wav file in aac with libfdk_aac, but it took quite a while to complete it.
In the second experiment, I had the ffmpeg segment of the wav file as is (wav), which was pretty fast (<1 second on our machines), then use GNU in parallel to run ffmpeg again to encode the wav segments for aac and manually modified the .m3u8 file without changing the duration. This was done much faster, however, when listening to the audio output, you could hear spaces of "silence".
I initially tried the second script using mp3, and the result was still the same. Although I read that lame adds padding during coding ( http://scruss.com/blog/2012/02/21/generational-loss-in-mp3-re-encoding/ ) does this mean libfdk_aac also adds addition during coding?
Perhaps this question is related to this question: How can I encode and segment audio files without spaces (or sounds) between segments when it is restored?
source
share