Fluendo, DV capture & other bits and pieces
Hi! I'm back from the cold and catching up with things quickly. Quick
run over my findings with firewire, fluendo, theora and friends.
- Still liking Fluendo - helpful to setup the flow we want, run it and
monitor it, specially for a live streaming setup. If we don't do
streaming, Fluendo can be shelved.
- If we decide to only do capture, and postprocess, Kino is the best
tool. It will perform dvgrab captures with good visual feedback and UI
controls.
- Theora compression is much heavier than expected. It's unfeasible to
run "straight" Theora for realtime unless we have really high-end
cpus.
+ We'll have to review the boxes attached to cameras and see how
powerful they are.
+ Assuming they are x86, we'll want to compile or package the MMX
version of Theora, which can do realtime (reportedly 3-4 times faster
than non-MMX). This is what people are doing at GUADEC, and I'll be
testing this on x86 this week.
+ Compressing several versions simultaneously is not realistic
(unless my tests prove otherwise). We were expecting to compress one
for HD and one for live streaming.
+ Considering compressing just for live streaming + capturing to HD
directly in parallel. Kino and fluendo can read simultaneously from
/dev/raw1394
- Firewire is tricky. We'll clearly want
+ latest 2.6.12rcX kernel
+ modprobe raw1394 manually (or a hotplug that does the raw1394 loading)
+ chmod /dev/raw1394
+ Good FW cables, and perhaps secure the connectors in place
(disconnects on camera moves are killing me!)
- Implications of local HD storage
+ ~13GB per hour * ~6hs of presentations = 78GB per day per camera
+ It all adds up to 1.6 TB (which apparently we'll have!)
+ Need one big storage server
* 1.6TB is hard to "take home and edit". We'll have to cut it at DebConf.
+ FW/USB drives to move files around or Gb Ethernet (seems we'll have
Gb Ethernet)
- Streaming to internal/external audience
+ Networking of the encoding machines important
+ Fluendo allows us to direct the streams to a central server and
serve from there. Need good connectivity from all the cameras to the
server.
+ Stream directly from encoding machines for Debconf users?
+ Stream directly from encoding machines to Intarweb? (if addresses
are routable)
+ We'll save the compressed stream and publish @ end of day
- Debconf website updates
+ Edit the "talks" page to add links to the live streams, stored
presentations, slides & links
- Camera operator workflow. Must be easy -- well resolved and with
strong visual & audio feedback.
+ Use Kino instead of dvgrab
+ Have a fluendo mgmt console to control start/stop
I'll post a plan detailing the DV capture strategy and the Fluendo
streaming strategy. We can have either or both. I'm hoping for both ;)
cheers,
maritn
--
To unsubscribe, send mail to debconf5-video-unsubscribe@lists.debconf.org.
Reply to: