Osmocom Conferences (OsmoDevCon, OsmoCon, OsmoDevCall)

h1. Video Release Process

Unless the presenter has an objection, talks are recorded and published afterward. This page details the current process.

Note that the process uses "Davinci Resolve":https://www.blackmagicdesign.com/products/davinciresolve/ for the editing. It's not OSS but this is what I'm used to and have already setup for other jobs, so this is what I use.

h2. Source material

The BBB instance raw stream dump can be downloaded just by looking up the link in the page source. Or simply using the script below and feeding it the meeting ID which is in the URL when viewing the raw stream dump through the BBB interface.

Example, for this particular BBB URL: 

https://meeting5.franken.de/presentation/10e07e078472a1876b5c23655123bfd50bc4b187-1705518054974/video/webcams.webm

In the script below, set MEETING_ID to "10e07e078472a1876b5c23655123bfd50bc4b187-1705518054974".

Once downloaded, ffmpeg is used to repackage the video streams in `mp4` container and transcode the audio from OPUS into FLAC to bring them into containers/formats that the linux version of resolve can load. (Note newer version of resolve should support MKV now, but that hasn't been tried. Sticking with the "tried and working" process for now).

<pre><code class="shell">

#!/bin/bash

set -e

ODC_SLUG="$1"

MEETING_ID="$2"

curl -o "tmp_webcams.webm"   "https://meeting5.franken.de/presentation/${MEETING_ID}/video/webcams.webm"

curl -o "tmp_deskshare.webm" "https://meeting5.franken.de/presentation/${MEETING_ID}/deskshare/deskshare.webm"

ffmpeg -i "tmp_webcams.webm"   -vn -c:a flac -sample_fmt s16 "${ODC_SLUG}-audio.flac"

ffmpeg -i "tmp_webcams.webm"   -an -c:v copy                 "${ODC_SLUG}-webcam-vp9.mp4"

ffmpeg -i "tmp_deskshare.webm" -an -c:v copy                 "${ODC_SLUG}-screen-vp9.mp4"

rm "tmp_webcams.webm" "tmp_deskshare.webm"

</code></pre>

h2. Edit

Just do a normal edit, trimming start / stop and anything that shouldn't be in there (like technical issues or whatever).

Some notes / guidelines :

* Project Settings

** 1280x720 24fps

** Davinci YRGB (not color managed) 

** Fairlight -14 LUFS target loudness 

* Intro:

** Use a Fusion composition (two examples attached), tweak as needed (font size, ...)

** About 10 second long

** Start speaker audio intro ~ 1-2 second in

** Transition to slide: Dip-To-Color-Dissolve, 1s, Ease In-Out

* Q&A:

** If there is no such slide, use a Fusion composition for video

** Cut out any long blanks

* Outro:

** Check there is a "thanks for watching" or similar

** Fade to black ~ 3 second

* Floating Head:

** Magic mask with expand and feather edge. Add alpha out on color page from the magic mask node.

** Add power window to control the crop

* Audio:

** FX: Dialogue processor filter with default for "Male VO"

** EQ: Lo cut at 100 Hz, Hi cut at 10 kHz

** Adjust audio level in Q/A section for the individual speaker

* Master Render

** Video: Grassvalley HQX 720p

** Audio: Linear PCM 16 bits

h2. Local encode

Once the master render is done, the file is transcoded into a few formats more suitable for online viewing using ffmpeg using the script below :

Theses are relatively low bitrates but perfectly fine for "slides" type of content with voice-over in 720p.

(Note the script is design to work on a machine with a NVidia card and driver to have hw acceleration of h264 and h265 encoding)

<pre><code class="shell">

#!/bin/bash

set -e

ODC_SLUG="$1"

ODC_RENDER_PATH="$2"

ODC_RENDER_MASTER="${ODC_RENDER_PATH}/osmodevcall-${ODC_SLUG}_master.mov"

ffmpeg \

	-hwaccel cuda -hwaccel_output_format cuda \

	-i "${ODC_RENDER_MASTER}" \

	-c:v h264_nvenc -b:v 1M -pix_fmt yuv420p \

	-c:a aac -b:a 96k \

	"${ODC_RENDER_PATH}/osmodevcall-${ODC_SLUG}_h264_420.mp4"

ffmpeg \

	-hwaccel cuda -hwaccel_output_format cuda \

	-i "${ODC_RENDER_MASTER}" \

	-c:v hevc_nvenc -b:v 512k -pix_fmt yuv420p \

	-c:a aac -b:a 96k \

	"${ODC_RENDER_PATH}/osmodevcall-${ODC_SLUG}_h265_420.mp4"

ffmpeg \

	-i "${ODC_RENDER_MASTER}" \

	-c:v libvpx-vp9 -b:v 400k \

	-c:a libopus -b:a 80k \

	"${ODC_RENDER_PATH}/osmodevcall-${ODC_SLUG}_vp9.webm"

</code></pre>

h2. VOC upload

The final step is to feed the master render to the VOC rendering pipeline so the talk can be published on https://media.ccc.de .

* Generate schedule file

** Using a fork of the VOC tool : https://github.com/smunaut/voctosched in the @osmodevcall@ branch

** Edit the @osmodevcall.tsv@ to add the new talk

** Run @./schedule.py -c osmodevcall.ini@ to regenerate @schedule-extended.xml@

** Upload that file on some public facing HTTP server

* Upload master

** The generated master file needs to be uploaded on some public facing HTTP server for the VOC to fetch

** Currently the tool above assumes it will be at https://downloads.osmocom.org/videos/osmodevcall/ or https://downloads.osmocom.org/videos/retrodevcall/

* In the VOC tracker

** Use the "Import" button to import the updated XML from its source. Only import new tickets in the confirmation screen.

** Once imported, edit the top level ticket for the talk and change state from "staging" to "staged".

** The "Recording" subticket will then trigger and go to "Recording" state while its downloading the master file.

** When done, it'll go to 'Recorded' state at which point, you need to click the 'cut' button. No actual cutting is needed, just select language and confirm with the "I finished cutting" button.

** And it's done. All the rest should proceed automatically until all the version are in the "Released" state