This is a follow-up to my previous post related the DebConf videoteam using a new software stack for the next conferences: http://acaia.ca/~tiago/posts/c3video-for-dc-take-2/.
An outdated documentation for current subject is available at https://wiki.fem.tu-ilmenau.de/streaming/projekte/c3/28c3/crs/pipeline. Although the system may work differently nowadays, the basic idea remains the same. A newer, but incomplete documentation can be found in https://repository.fem.tu-ilmenau.de/trac/c3tt/wiki. Btw, CCC people from #voc at hackint.eu have been very kind and supportive.
I've set an instance of C3TT for DebConf15 in http://c3tt.acaia.ca/. If you want to play with it just ping me in #debconf-video at oftc. As you can see, we can keep a single external C3TT server for all Debian events, without much work left to the local side, doesn't it sound amazing?
Setting a new conference
Go to Projects, then Create.
In the project area we'll need to import the Tickets. Tickets will come from the schedule file, which is a XML as generated from frab. With a minor hack we've been able to make the schedule XML from DebConf Summit quite compatible to it (kudos cate!):
By importing the schedule from https://summit.debconf.org/debconf15.xml we'll be asked from which rooms we want to import the events. Usually those that have video coverage will be selected:
Then, we may want to exclude some talks that we won't provide video:
We're also required to adjust some Properties for a given conference. An example with some explanation of these properties is availabe at https://c3voc.de/wiki/c3tracker. For my initial tests the ones below seem to be enough:
The backend: basic understand
The screen UI mentioned above will run a set of scripts in background which will automate most of the tasks, preparing videos for cutting to deploying them in different online services.
Tab 0: A-recording-scheduler
Each 30 seconds it will check if there's any ticket in the state scheduled or recording. It's based in the start/end datetime of the talk, so the ticket will be kept as scheduled (current < talk start), or marked as recording (start => current =< end) or recorded (current > end).
Tab 1: B-mount4cut
Each 30 seconds it will check if there's any ticket in the state recorded. That means the talk is already finished, and the raw video file is available in the path which has previously been set as a property (Path.Capture) in the web interface.
For each ticket marked as recorded it will try to find the related video file in the capture path. The file format should be <room>-YYYY-MM-DD_HH-MM-SS.<extension>. The script will then use fuse-vdv to create a custom filesystem with some needed files for human interaction (fancy stuff!).
Here's an example of talks in a room called Heidelberg, after being recorded and auto-mounted by the B-mount4cut script:
The human interaction is just a short review process using Kdenlive. The reviewers will access these files via a glusterFS network share. There's even a Debian Virtualbox image provided for that, including all the necessary tools. I'm going into this right now and will report what I get in the next hours.
Hopefully the following scripts will also be covered, very soon-ish :)
Tab 2: C-cut-postprocessor
Tab 3: D-encoding
Tab 4: E-postencoding (auphonic)
Tab 5: F-postprocessing(upload)
DebCamp is fun.