Today's lispy gopher show had downs and ups. It almost, yet
bizarrely, not quite didn't work live. The fact that it
didn't quite completely not work live is a source of mystery
to me. How can it possibly have almost completely not
worked? Seems like it should (work-p episode) -> either true
or false.
On openbsd sndio I have three devices, snd/0 my speakers
(output), snd/1 my mic (capture), snd/mon : speakers but
it's a capture (the system sound monitor). Refer to my
previous (openbsd . sndio) posts. sndio and icecast are
supported by ffmpeg. I have no need for complex filters, due
to routing everything through my system sound and having a
system sound monitor. Here's most of my ffmpeg line for
icecasting, divergent enough from publius' that anything
wrong is my fault:
ffmpeg -f sndio -i snd/mon \
-c:a libmp3lame -ar 44100 -b:a 192k \
-legacy_icecast 1 -content_type audio/mpeg \
-ice_name "me" -f mp3 \
icecast://me:
[email protected]:port/me
Originally I thought legacy_icecast meant that it wasn't
icecast2, as suggested by the icecast:// and not http:// ;
however, legacy_icecast is for supporting any icecast prior
to version 2.4. Because this ffmpeg line has worked, I think
its flags are completely correct and something else was
going wrong today.
When I say today didn't work, there was long, total silence,
punctuated by 10 seconds of my speaker sound (during 9 Days
- Absolutely (story of a girl)) working normally. That 10
seconds ruins a lot of theories. claudiom commented on the
track choice, so I was confused as to whether it had been
working.
After random fiddling hadn't worked for half an hour, in
which tob created a discussion of getting toobnix.org
working as a non-javascript user, I just tried my original
line with snd/1 instead of snd/mon, and it worked (but only
broadcast my mic capture device, snd/1).
The same ffmpeg line is what I'd used at Synth Battle Royale
to play my system sound, though I hadn't plugged in my mic
in that case.
After the live show's technical failure (bar 3 minutes of my
mic right at the end), I replaced the icecast with
upload.mp3 in the original line, and recorded an abridged
version of what I intended normally, and uploaded it
appropriately named to clobber the previous archive.
During openmic, I streamed that mp3 via ezstream as well
without a hitch. It conforms to aNONradio's requirements,
and did in all cases.
there seem to be three places something could've gone wrong.
- snd/mon was angry in this particular case on my end
- http or extreme latency problems on my end
- http or extreme latency problems on the icecast end
My suspicion and hope is that it will inexplicably resume
working in the future. I plan to record the show, with some
coordination, during an openmic and then pre-upload it for
my slot since user interaction is important (lispy, after
all), but could happen during a spare openmic slot.
Unrelatedly, I've discovered I had my mic's gain negative
(logscale), which explains some of the disparity between
music volume and my voice in the successful recording.
*
I also used ffmpeg to direct my snd/1 to snd/0, which worked
in the recording and other cases
ffmpeg -f sndio -i snd/1 -f sndio snd/0
it's annoying to hear yourself echo though.