Inspired in part by the "How to Digitize a LP" posting at the Internet Archive, I think a how-to-record-an-event guide is in order. The below approach is the best blend of cost effectiveness, technical flexibility, and practicality that I have achieved thus far. Suggestions for improvement are welcome.
Record video, record!
My Panasonic PV-DV73D digital video camcorder has been a champ over the years. This particular camera provides good functionality (e.g., LCD panel, simple controls, quality built-in mic and light) without being a lug-over-the-shoulder burden.
Perhaps counter-intuitively, obtaining high-quality audio is more important than video in settings where words are important. Audio should be monitored during recording such that any problems (e.g., background noise, ground loops, and clipping) can be identified and hopefully corrected. In case the event is an iPod commercial, I recommend Creative EP-630 earbuds. Room acoustics and any specialized audio equipment should be tested and mediated beforehand. If standing next to an audio capture source, absolute silence is necessary lest viewers hear an errant Velcro strap, chuckle, or sneeze.
Multiple participants in the room may be accommodated either by tilt/pan/zoom of the camera frame or multiple stationary cameras. I use a cheap telescoping tripod but need to buy a better one as my tilting is jerky and thus visually distracting. Keep the camera lens clean with a lens cloth. Room lighting is invariably "as is." I have an extra large camera battery but do suggest using AC power whenever possible.
Events are best recorded either to digital video (DV) tape in SP mode or directly to a hard disk via Firewire IEEE 1394a live capture. Keep tapes absolutely clean, humidity-controlled, and far away as possible from magnetic fields.
Live streaming via Icecast is possible. (As a huge aside, I would love to see a mobile cameraperson icecasting from a Wi-Fi zone, perhaps using embedded technology like a football helmet.) I should also mention that my Cowon A2 is capable of recording audio/video from an arbitrary source, albeit at limited quality settings and in a proprietary format.
For those serious about filming anything and everything, Current TV's filming guide will transform you into the consummate semi-professional cameraperson.
March video, march!
Use of DV will require a Firewire-enabled system. I use a Belkin F5U506 ExpressCard. The following BASH command on a Linux (live) distro usually works for me:
[user@machine ~]$ dvgrab -format raw -interactive -showstatus somefilename
(-duration is also a handy option)
This realtime capture command will produce a series of 1GB .dv files rather than a monolithic single file. I prefer this behavior as 1GB is a practical transport size and possible file corruption is confined to a single 1GB chunk.
The demands on disk space are unforgiving. By my calculations:
4min, 51.5 sec standard DV = 1GB on disk; 60min = 1 tape in SP mode = 12.36GB
2 hr meeting = 24.72GB on disk, then ×2 for Cinlerra export + finished .ogg ≈ 50GB
500GB HDD @ $90 USD ≈ 20 2-hour events (w/o file intermediates) ≈ $4.5 / hour
md5sum is perfect for tracking file integrity; FIFO pipes are problematic.
Using full-fledged DV may seem to be overkill when the finished product is delivered à la YouTube, but consider the following:
- The encoding result is better when starting from non-compressed originals.
- You will have the full-quality original in case you record something that surprises you.
Compress video, compress!I recommend Theora video and Vorbis audio in an Ogg container file with an ".ogg" file extension. The
Xiph.Org Foundation requests using ".ogv" to denote "Ogg Video" but I think doing so undermines the reputation of Ogg as a multiplex-ready specification.
I do reduce video resolution but keep quality settings high and the frame rate untouched. I suggest using 352×240 resolution: moderate file sizes, sufficient to communicate body language, and not aesthetically overdemanding on speaker(s). I have no interest in switching to high-definition video.
When all is ready, I use the following command:
[user@machine ~]$ ffmpeg2theora -o uploadme.ogg -v 6 --optimize -x 352 -y 240 --aspect 4:3 -f dv -c 2 -a 8 --artist "Me" --title "The Title" --date "Last week" --location "Down by the River" --organization "Me" --copyright "The license" --nice 20 preparedsource.dv'prepared_source.dv' may be either '
cat *.dv | ffmeg2theora ...
-' input or an edited DV composite file; I use DV exported from
Cinelerra. I will save Cinelerra editing instructions for a future posting. In a nutshell, I am not crazy about Cinelerra (esp. the name itself and GUI non-standardization) and look forward to future
Lumiera development.
The command encodes in near-realtime on my 64bit 2.17GHz laptop, which is to say the entire process is time-consuming. Fortunately, two simultaneous encodings are possible on a dual-core system and
ffmpeg2theora is amenable to batch processing. The command will transform 720×480@29.97fps NTSC DV (4:3 DAR, 9:10 PAR) → 352×240@29.97fps Theora (4:3 DAR, 1:1 PAR), resulting in a slight blurring effect as the PAR is changed. The alternative would be to add letterboxing, but nobody will be able to tell the difference anyway.
Thusnelda (a spec-compliant Theora branch under rapid development) is a terific improvement to standard Theora. Unfortunately, it seems many decoders "in the wild" were not actually Theora-compliant to begin with, so Thusnelda is not yet an option...
Publish!The Internet Archive is the most innovative library I am aware of. Your video will find a long-lived and happy home sitting next to many, many others on the digital book-shelf.
Login → Upload → use the "files over 100 MB" link → "Create Item!" → upload to the FTP subdirectory → click the activation link → place item in "Open Source" movie collection (a misnomer) → edit item details.The "Metadata editor" is fairly straightforward. Do change "Format" to "Ogg Theora" on the initial submission so that thumbnails and the in-browser playback applet are properly displayed. Also, consider embedding authorship information directly in the opening sequence as visual metadata is good professional practice.
You are not "sharing" video insomuch as you are irrevocably publishing it. Please bear this in mind, especially in light of potentially-thorny legal issues inseparable from event recording (
e.g., recording consent, defamation, and copyright infringement).
Closing thoughtsMost of my friends--even civically-minded ones--have not given cultural preservation serious thought. I hope these technical instructions will prove helpful in enabling the broadcast and subsequent analysis of important yet relatively unknown events to many others, including those on the far side of physical, language, and temporal barriers.
-George
P.S.:
Elephants never forget.
[Updated link to Current TV on 17-Dec-2008]
