DebConf5TodoVideoStreaming
From Wiki
Debian conference todo: video streaming/recording
Contents |
[edit] Agenda for next meeting
- hardware setup, do we have everything we need now ?
- evaluation of different software solution
- what to test
- What formats/bitrates/sizes will we put online? Woody supported, Sarge supported, Sid supported, Mac supported, Windows supported?
- Who should we be bugging to get a definitive answer on how many rooms and how many hours will need to be filmed?
- Where will the streams be hosted? People.debian.org and debconf.org seem like they're overburdened already.
- should we attempt to stream the main room live even if the overall focus remains on recording? (If we have an extra camera, it wouldn't hurt to try)
[edit] what does HUT provide
Main auditorium (300 ppl):
- directional mocrophone, probably works from the control cabin
- control cabin ;)
- 4 mics for the speakers (fixed, wireless probably available)
- 2 big projectors from the bac of the aud.
- slides projector, with integrated camera
- 2 hooks for computer video outputs
- computer (with windows or mac os)
Small aud. (90 ppl):
- 2 mics for speakers
- 1 projector
- mooch believes no mic for the audience
- slides projector with camera
A room for editing in the basement of dorm building
- available 24/7
- 100 mb/s sustained between the dorm and the auditoriums
[edit] debcamp
- we absolutly need to be prepared before debcamp
- participate in the preparations ;-) (which at the moment takes place on debconf5-team@lists.debconf.org, on irc (#debconf-team on freenode) and here)
- put your name here (if you dont want that, tell us via mail): h01ger, ...
- we will use debcamp to setup our plans and test them
- we absolutly need to be prepared before debcamp
[edit] about the cameras needed
- "We need cameras. No, more cameras. and tripods!"
- we should aim for three cameras and three tripods at least. might be better to have one spare camera if another one fails.
- a dedicated streaming server (to produce the streams) for each camera would be desirable as would be 3ghz and 160gb harddrives
- how much storage space for the archived streams do we need approx. ?
- 300 GB ~ 24 hours of DV. Consider USB disks.
- hrobak recommends recording directly to harddrive. A laptop and a USB harddrive can record a day's worth of video. As an alternative he suggests sending the DV stream over the network to a server with big disks. That would help a lot with timely compression and publishing.
- could using a mix of NTSC & PAL cameras become a problem ?
[edit] other cameras + mics we got
lightsey brings with him
- 3 Cameras, all NTSC:
- Panasonic PV-GS70D (Good image quality, Good Linux compatibility, Mic input)
- JVC GR-D22U (Reasonable image quality, Good Linux compatibility, no Mic input)
- SCD103 Samsung (Poor image quality, bad Linux compatibility, Mic input)
- 2 tripods
- 1 Wireless mic (Audio-Technica Pro 88W)
- 1 USB2 external drive case (for 3.5" IDE drives)
- 4 or 5 3.5" drives ranging from 120GB to 300GB
Herman Robak brings
- 2 cameras, both PAL
- Sony HDR-FX1E (prosumer HDV video camera; good image quality, Linux compatible, minijack mic/line input w/manual level)
- Panasonic NV-DS28 (consumer model, modest image quality, Linux compatible, mic in)
- One tripod.
- 1 dynamic microphone with XLR connector.
- 1 Wireless mic (Chiayo m800, UHF)
- 1 2.5" 30GB USB2 disk, 1 2.5" USB2 disk enclosure and a 80 GB 2.5" disk.
[edit] operating the cameras
- who is going to organize the people who will operate the cameras?
- we need some person to do it
- Fabian is gathering team members, regular updates are sent to the debconf5-video list
- schedules / tasks
- For the video recording, do remember post-processing. In Oslo, we tried to do video recording, but when the recordings were done, no-one had thought about the need for post-processing, and the tapes just ended up on my desk. They are till there. :)
- Also, the audio recording level is important. Part of the problem with the tapes from Oslo was that the audio level was too low, so they would require a lot of work to make sensible video films from them.
- How do we ensure good audio quality?
- microphones hooked up into the server-camera setup
- someone to monitor the audio level?
[edit] software setup
- distribute the streams as local live-streams and for worldwide download afterwards via http and bit-torrent - or (how) should we aim for a worldwide-realtime-streaming solution...
- if we go for the latter, how much bandwith is available and how much is needed ?
- do we provide a separate, audio-only stream?
- software of interest - we have not decided upon the software we want to use
- http://www.flumotion.net/
- ffmpeg (see http://ffmpeg.sourceforge.net), was used to stream+record the 21C3 as described here: https://21c3.ccc.de/wiki/index.php/Videostudio, sourcecode available here: http://svn.gnumonks.org/cgi-bin/viewcvs.cgi/trunk/21c3-video/ (where ./scripts is of most interest)
- or dvgrab for hd-recording. also no gui ;)
- http://www.xs4all.nl/~salsaman/lives/index.html - a video editor programm - AFAIU it's not really what we need as it's designed for post event processing
- ESM: http://esm.cs.cmu.edu/ - looks interesting, but no source to download (only http://esm.cs.cmu.edu/forum/viewtopic.php?t=25), no linux client ports for archs !i386 and no (released) linux broadcaster at all
- http://www.flumotion.net
- http://users.pandora.be/acp/shotcut/ - software to select and mix DV streams
- If we are to do live mixing (picture in picture, titles, etc.) we most probably need an SMP box for the compositing, and maybe a separate box for compressing the output.
- cinelerra
- kino
- large file support
[edit] misc questions
- Anything else to document on video beside the talks?
- If we have volunteers, transcripts would be ideal, because they make the presentations google-able.
- Otaniemi Underground Broadcasting S. ? (the university's amateur tv station)
- Oh, and it would be lovely, if we (the team doing the video recording) could get the sildes before the talks/presentation. So we can "cut" them into the video-recording directly after that talk or maybe even live - depending on our (free) software solution...
- It would save time if we could do the editing on-the-fly.
- Picture in picture, to show the speaker in the corner of the slides would be neat. But such compositing requires simultaneous decoding of 2 DV streams, and DV decoding is a CPU hog: A laptop is not powerful enough.
[edit] misc notes
- http://www.debconf.org/debconf5/news/call_for_papers says "The presentations will be recorded, and may be broadcast over the Internet. Any copies of the presentation will be made available under a license such as the MIT/X11 license."
- currently there is no money for insuances if people lend us their cameras - people must at least know this
- our first "job" will be to cover the debian day, which will happen on the 9th of july and will last around 8 hours!
- around 20 talks, each 65min => 210gb + 80gb (debian day) for the raw DV data
- if we need to keep the raw data after debconf for post-processing, we need some harddiscs to carry them home. as we have to give back the servers on the 19th or so...
- lightsey said : "15GB per hour should be enough for the raw video and the web encoded versions. If disk space becomes an issue, we can delete the raw video and keep the original DV tapes instead. I'd like to keep the original tapes of everything if it's possible, but they cost around $3 each."
- do we have 100mbit or 1gbit between the media servers and the fat server ? dedicated 100mbit ?
- More testing: A DV switcher with display. Must decode several streams realtime, display them, and let the operator switch. Outputs DV or compressed video.
- <h01ger> hrobak, <hrobak> h01ger: Could we not get a live feed from the projector? <-- ? projector = video beamer ? ahhh - excellent idea! :-) but i guess we need a vga mixer/splitter then... but now we can ask for those ;) <h01ger> right? <hrobak> Yes, that's what I meant.
- Herman's suggested division of chores: One camera person and one mixer person for each auditorium, with a dedicated 100 Mb network link between them.
- The camera operator has two cameras, a laptop and maybe a directional microphone to point at the audience.
- The mixer person switches between cameras, microphones and sound inputs, adds titles and monitors the sound levels. The mixer may sit in a noisy machine room.
- we should provide a area in the audience which doesnt get recorded+streamed.
- we have to leave the (cleaned )building on the 18th, and we (probably.. not confirmed yet) have to give back the hardware on that day as well - so on the 17th we got to have moved/backupped/whatever all the data..!
[edit] Things that need testing
- How feasible is it to grab from two DV cameras on one laptop?
- How can we reliably grab video on a laptop, and stream DV to a server?
- Compositing several video sources in real time. E.g. showing a close-up of the speaker overlaid on a capture of the screen image displayed by the projector. Does this require an SMP box? Can the same box manage to do compression as well, in real time?
- smp - lightsey heard the call "is mencoder (or whatever we will use) cabable of utilizing smp machines?" and answered "From what I understand...mencoder no, transcode partially, ffmpeg yes."
- wireless mics in a room with >100 laptops and cellphones - before the 9th ;)
<< Back to DebConf5Todo