[FFmpeg-user] Raspivid with ffmpeg on both sources (youtube & rtsp)

[Moritz Barsnick]

On Mon, Jun 19, 2017 at 20:17:34 +0000, Nicolas Winandy wrote:
> So, I looked in the ffmpeg documentation and found an interesting thread on multiple outputs : https://trac.ffmpeg.org/wiki/Creating%20multiple%20outputs.

That's probably the right place to learn.

> Then, I have tried the piped processes and another ffmpeg call to the inital command:
> 
> raspivid -o - -t 0 -vf -hf -fps 30 -b 6000000 | ffmpeg -ar 44100 -ac 2 -acodec pcm_s16le -f s16le -ac 2 -i /dev/zero -f h264 -i - -vcodec copy -acodec aac -ab 128k -g 50 -strict experimental -f flv rtmp://a.rtmp.youtube.com/live2/<SESSION> | ffmpeg -f h264 -i - -vcodec copy -f rtsp -rtsp_transport tcp rtsp://localhost:8888/live.sdp

That's not correct, and certainly not what the wiki page recommends.
The ffmpeg call after the final pipe isn't being fed anything.

Let us see what you are doing:

$ raspivid -o - -t 0 -vf -hf -fps 30 -b 6000000 | [...]

I can only assume that that's correct, and gives a raw H.264 stream on
stdout. ("-t 0" isn't documented. Does that imply "infinite"? Can't you
just omit it?)

$ [...] | ffmpeg -ar 44100 -ac 2 -acodec pcm_s16le -f s16le -ac 2 -i /dev/zero -f h264 -i - -vcodec copy -acodec aac -ab 128k -g 50 -strict experimental -f flv rtmp://a.rtmp.youtube.com/live2/<SESSION>

You don't need "-strict experimental" for the aac encoder anymore. BTW,
why not omit the audio? Does YouTube need it?

"-g 50" doesn't have any effect with "-c:v copy".

Apart from that: Does this part work (if you omit the part after)?
Well, fine then so far.

$ [...] | ffmpeg -f h264 -i - -vcodec copy -f rtsp -rtsp_transport tcp rtsp://localhost:8888/live.sdp

This would do the same thing to a local RTSP stream (if the part in the
middle weren't there). Does this work by itself?

I guess, with some Unix shell magic and/or "tee" (the program, not the
ffmpeg muxer), you could probably feed both ffmpegs' stdins from
raspivid's stdout. But since you asked about the ffmpeg way:

Both your desired outputs use the same "copy" encoding, which is cheap,
so you can just do two encodes to two outputs:

$ raspivid -o - -t 0 -vf -hf -fps 30 -b 6000000 | \
  ffmpeg -ar 44100 -ac 2 -acodec pcm_s16le -f s16le -ac 2 -i /dev/zero -f h264 -i - -vcodec copy -acodec aac -ab 128k \
  -map 0:a -map 1:v -f rtsp -rtsp_transport tcp rtsp://localhost:8888/live.sdp -f flv rtmp://a.rtmp.youtube.com/live2/<SESSION>

If you eventually need to do actual non-copy encoding, you may want to
use the "tee" muxer, where you only need to encode once (assuming both
targets desire the same parameters).

Cheers,
Moritz