As usual, we strongly encouraged speakers to
record videos to lower everyone’s stress levels
and allow for captioning by volunteers, so that’s
what most speakers did. We were able to handle
a few last-minute submissions as well as a
live talk. Getting videos also meant we could
publish them as each talk went live, including
automatically putting the videos and transcripts
on the wiki.
We didn’t have obvious video encoding cut-offs, so
re-encoding in a screen was a reliable way to
avoid interruptions this year. Also, no one
complained about tiny text or low resolution, so
the talk preparation instructions seem to be
working out.
Automatically normalizing the audio with
ffmpeg-normalize didn’t work out, so Leo Vivier
did a last-minute scramble to normalize the audio
the day before the conference. Maybe that’s
something that volunteers can help with during the
lead-up to the conference, or maybe I can finally
figure out how to fit that into my process. I
don’t have much time or patience to listen to
things, but it would be nice to get that sorted
out early.
Next year we can try remixing the audio to mono.
One of the talks had some audio moving around,
which was a little distracting. Also, some people
listen to the talks in one ear, so it would be
good to drop things down to mono for them.
We think 60fps videos stressed the res server a
bit, resulting in dropped frames. Next year, we
can downsample those to 30fps and add a note to
the talk preparation instructions. The hosts also
suggested looking into setting up streaming from
each host’s computer instead of using our shared
VNC sessions.
There was some colour smearing and weirdness when
we played some videos with mpv on res. Upgrading
MPV to v0.38 fixed it.
Some people requested dark mode (light text on
dark background), so maybe we can experiment with
recommending that next year.
I did a last-minute change to the shell scripts to
load resources from the cache directory instead of
the assets/stream directory, but I didn’t get all
of the file references, so sometimes the test
videos played or the introductions didn’t have
captions. On the plus side, I learned how to use
j in MPV to reload a subtitle file.
Sometimes we needed to play the videos manually.
If we get the hang of starting MPV in a screen or
tmux session, it might be easier for hosts to
check how much time is left, or to restart a video
at a specific point if needed. Leo said he’ll work
on figuring out the configuration and the Lua
scripts.
I uploaded all the videos to YouTube and scheduled
them. That was nice because then I didn’t have to
keep updating things during the conference. It
turns out that Toobnix also has a way to schedule
uploads. I just need to upload it as unlisted
first, and then choose Scheduled from the
visibility. I wonder if peertube-cli can be
extended to schedule things. Anyway, since I
didn’t know about that during the conference, I
just used emacsconf-publish-upload-talk function
to upload videos.
It was fun playing Interview with an Emacs
Enthusiast in 2023 [Colorized] – YouTube at lunch.
I put together some captions for it after the
conference, so maybe we can play it with captions
next year.
