About this document
Written June 2009 for DVswitch Version 0.8.2
This document is an overview of what is involved in doing video at a large event. It is meant for people who are experienced DVswitch users. Trying to video a large event without being familiar with the tool chain will likely be a waste of your time and cause disruption to the event. If you are helping with an event and want to be familiar with the ecosystem, be sure to read the other docs too.
Recording and or streaming the content live over the Internet is very doable with COTS (Commodity Off The Shelf) technology. It does require some preparation, but given a conference is generally a high concentration of content, economy of scale kicks in and the effort is well worth it.
Weeks or months in advance you should do the following:
There are legal and politeness issues concerning pointing a camera at someone. You also may want to make use of the venue's equipment, like sound system, VGA projector, lights and network. Where “make use of” is as simple as “touch.” Use your good judgment.
Here are what some events have done:
There is marketing value in the videos. Letting it go to waste is a shame.
To title the videos, it is best to have the same titles as published in the event materials (website, printed.) For one or 2 talks, this can be done by hand, but much more and it makes sense to have a formal database.
- Number of event rooms, and hopefully some sort of ID for each room
- List of talk titles, and ideally datetime, duration and location, and ID(s) of the external systems so that they can be upldated with links to the video.
- Presenter(s) name and contact info, URLs referenced in the talks
- Categories, license.
- Tables and chairs – Some events sell out, so seating is limited. Some events do not provide the attendees with tables. The Video Crew needs a table, a place to sit and a place for the tripod. Make sure that is taken into account when room capacities are determined. It helps if the camera and operator are above everyone else, so having a platoform for the tripod and tall chair is good.
- Video Production room. You need a secure room that has some space to work and can have some equipment hooked up and running (so it needs tables, chairs, power and maybe AC) some expensive equipment will be left in this room, so it needs to be locked when no one is around.
If you are desperate, you can setup the equipment and let it run all day, but there is a high probability that the results will be worthless, so get a team of people to help out. Ideally you want the same people for the whole event, and ideally people only somewhat interested in attending the event. Some interest results in a good attitude, but if they are too interested they are more likely to ignore the video process. You want to fill the following positions:
- Project Manager: responsible for making sure everything is ready: equipment, team, permission, data, server logins (blip, icecast, etc.) Gets to resolve disputes over licensing and codec crusades. Makes sure the rest of the crew knows what they need to know to do their job. Assigns crew to locations.
- Network manager: One person has to make some decisions. They should have some clue about the dvswitch suite, the more the better. If they say “use the wifi' the are not the right person.
- Producer: Responsible for the quality of the content. Lighting, sound, camera angle, coaching the people doing running the equipment.
- Camera Operator (Cam Op): keep the camera pointed at the subject. Make sure the camera is on and the lens cap is off. Maybe take direction from the producer.
- Sound Board Operator (Sound Op): makes sure the levels are high but not clipping, and somewhat consistent. If the presenter moves away from the mic, they boost the level. They get to bring the audience mic up and down depending on their feelings about the noise coming from the audience.
- Video Director (DVswitch Op): operates dvswitch. Decides when to cut, what stream to use, when to apply affects like PnP.
- Reviewer: looks at the resulting video and tags it as junk, talk, problem.
There can be some overlap, but if you have one person dedicated to each of these tasks, the results will probably be better. There is benefit to the same person doing what they are best at, but it is good if everyone has a little cross training too. This helps when someone can't make it and people need to be redeployed, and it also gives people an idea of what helps the other people. A CamOp probably won't realize how annoying it is to be zooming and panning all the time as they tweak things to make it “just right” but if they get a turn trying to work with that feed, they will appreciate calmness.
For each Event room:
- 1 or 2 cameras/tripods
- 1 or 2 computers (if 2, then networking between them)
- 1 TwinPact100
- Sound (feed from PA or mic for speaker, mic(s) for audience, mixer)
Camcorder needs to be comparable with dvswitch.
Not using a tripod will result in terrible video. Even the cheapest static tripod will make a huge difference. Adding tilt and pan makes another huge difference. After that, better tripods (like fluid head) will result in better video. How much better is a matter of opinion.
TwinPact100 – so far the only device I would consider using for a conference. It is transparent to the presenter, and it works 100%.
Sound should be more than just a feed from the main PA. You want to be able to capture noises that are not fead though the house sound system, like laughter and applause. Ideally audience questions get asked into a mic, but there are times when someone will shout something out, and often it is something worth hearing. The levels coming from the main may not be desirable, so being able to adjust them on the fly can make a big difference. Too high a level will cause clipping, and that's very bad.
- Backup storage
- File Server
- Transcoding nodes
- Management App Server
- LAN for connecting systems (either in a central processing room or out to the event rooms if live streaming)
- Internet access if you plan on streaming or uploading from the event.
Backup – If the value of the video is worth all this trouble, then it is worth having a backup. There is a lot of data to back up, and not a lot of time for it to happen, so make sure you have sufficient bandwith. Currently esata drives are a good choice, or a 2nd file server.
File Server – Central place for the files. It should have enough bandwidth to handle the demands of the nodes. It can ealily be the bottle neck if a few rooms of files start getting moved to it at the same time. 1 room will produce abut 100gig for 8 hours of recording (note that it is likely that breaks and such get recorded too) 1 gigabit eathernet connection will be saturated for hours.
Transcoding nodes – trancoding will take a while, the more CPUs that can work on it, the sooner it will be done. Events generally do not run 24 hours a day, so when the dvswich computers are not running dvswitch, they can be used for transcoding.
- Management App Server – Somewhere will be a database of metadata linking talk titles to video. This is also a sort of process schedular to make use of the nodes. The schedular has a list of avalible nodes, it remotely runs the trancoding command on the node, the node gets access to the .dv files over NFS.
- LAN for connecting systems (either in a central processing room or out to the event rooms if live.) Much like the backups, bandwidht into the server will be the bottleneck. There will be mutiple clients trying to dump 100gig of data into it, and that takes time (like 100gig/hour). Having more than one ethernet port is not a bad idea. Another good idea: backup to esata drives, then dump from drive to server. Exactly what you do depends on what technologies have been developed and what you have abliable.
You may not have access to the computers unill eveyone avies at the event, hopefully one or 2 days before the event starts. It is good to automate much of the setup. Once they are in use, you want things to be easy and fault tollerent. If things go bad, you want to be able to recover quickly.
For configing the PyCon 09 recording machines:
- install Debian from CD, give each box a hostname of pycon-N (N=1,2,3...)
- add http://debian-multimedia.org to /etc/apt/sources.conf
- apt-get install pycon-video dvswitch dvsource dvsink ffmpeg
- Install Debian from CD
- apt get install nfs-server apache django postgresql python ffmpeg
- ssh key exchange with all recording machines.
- create a django app for video file management.
- create various scripts to rsync, generate titles, transcode, upload, tweet, export csv of talk ID and URL.
The last 2 steps are slated to be rolled into a new and improved app. Some of the scripts have hard coded login credentials, so they need to be at least sterilized before they can be made public. Hopefully the new app will just take their place.
As recording is taking place, it is wise to have note paper handy that is pre-printed with Date, Room, Talk Names and times. Things will happen like “AC came on next to mic” or “presenter was 10 min late” This helps answer questions later like “why does this sould like a wind tunnle?” or “why is this 30 min talk only 20 min long?”
It really helps for the crew to be very noticeable. It helps the project manager quickly find them which helps things like "is everything ok?" It also is a back stage pass and a "If I say I can here an odd noise, it needs to be fixed."
Ideally the mic stands, cables, tables, chairs, tripods, etc will be setup the day before the event starts. Sometimes this is not possible, like there is some other event taking place in that venue. At a minimum, you should plan to spend an hour setting up a room, and you should be done setting up before the first attendee comes into the room. During the setup will be safety hazards (un secured wires) and you can work much faster when no one is in the way.
Also ideal is if the people running the equipment do the setup. This helps them be familiar with exactly how things are setup, and will be able to react to problems much quicker. It also helps paralize the setup process. 8 rooms at 1 hour each = not enough time it the morning for one person to do it.
Tear down is much less critical, but if the crew can take care of it, the Production Manager will be very appreciative.
About an hour before the first talk starts, the team should gather and figure out who didn't show up and adjust assignments accordingly.
- Hand out recording note sheets
- Ops go to the event rooms
- Cam Ops do white calibration
- Sound Ops do sound check.
- DVswitch Ops make sure all the streams are coming in, verify that sound is good.
- DVswitch Ops clicks “Record” and tells Session Chair the cameras are rolling.
- Ops operate their equipment
- Producer writes notable events on recording sheets.
- Project Manager checks every so often to make sure all is going well.
At least once a day:
- Collect the recording sheets
- raw files backed up and moved to file server
- file names get added to database
- files on server are reviewed, identified as either junk or associated with a talk
- all the files associated with a talk get combined and converted into something suitable for moving over the Internet.
- converted files get reviewed and approved for upload or marked as a problem.
- approved files get uploaded
- URL of upload gets pushed to upstream database of talks
- problems get addressed.
Once you are done uploading and confident no more work will be done, create a high bitrate version of the files and save to an external HD. Currently that is 'better' than blue-ray disk, but as prices fall and more people have them, that could change.
It is a lot of work, but hopefully you see it isn't rocket science. If you are familiar with DVSwitch and understand this document, you are well on your way. If you can then recruit a good crew and take possession of the equipment well in advance, you are over the hump and the rest is down hill.