When using an HLS input transcoding processes can really be speeded up
compared to having a udp input simply because you have a few segments
cached and can transcode immediately as fast as CPU allows.
This means that ffmpeg will also produce the playlist and first segments
really quickly. This is all good.
The main problem is when we want to make the output also available in UDP.
Simply we use the tee muxer to trancode once and output to both hls and
udp. The problem is that UDP becomes bursty and is not "streamed".
We would like someway to use "-re" (realtime) output for the udp output but
not the HLS output.
Do you understand the reasonings and what can you suggest?
We thought that it might be possible to possibly use tee muxer to output to
stdout+hls and then launch a separate ffmpeg to do "-re" from stdout to
produce the multicast output. But isnt there a better way?
ffmpeg-devel mailing list