The CfgMgmtCamp 2019 Recording Setup
As in earlier years, Sig-I/O takes care of some of the lecture recordings at various conferences like CfgMgmtCamp, LOADays, HackerHotel, Eth0 and Techtalks at IT-Gilde, Revspace and Bitlair.
During one of these recording sessions, JJ Asghar (@jjasghar), asked about the video setup and asked if there was a blog-post about how it worked. This will be that blog-post.
The basics for the recording setup being used is founded on the use of Open Source / Free Software and affordable hardware. Originally a Firewire based camera setup and dv-grab were being used, but since new machines with firewire have been getting harder and harder to find, a switch to something more modern was needed.
The current recording setup consists of the following:
2 USB3 HDMI capture cards
A camera with HDMI output, and preferably some form of audio input and connected to the HDMI capture card.
HDMI Splitter between the presenter's laptop and the projector, with the second output going to our HDMI capture card.
A reasonably fast/modern laptop (i5-6th gen or newer) with USB3 ports running Linux.
Open Broadcasting Studio software (obs-studio.org)
Currently, 3 different camera's are being used, with their own pro's and con's. For our requirements, the following features are ranked most important:
HDMI Output (SDI would also be appropriate, but it much more expensive)
XLR Audio inputs (for connecting to the sound-mixer / microphones)
Dual SD-Card slots for continuous-recording (nice to have)
Low-weight, so the camera can be oriented vertically (9x16)
AC/DC power input, so we can run continuously without using batteries
The specific camera's in use:
Canon XA10, nice and small, internal 64GB flash and dual SD-slots, 2 XLS's and minijack audio input, basically everything we need. (No option to have XLR+internal mic running at the same time though).
Panasonic AG-AC90, large, featurefull, good optics, 2 XLR's and internal mic, all switchable. 2 SD-cards and good audio tuning/interface.
Sony (model unknown), records to tape only so no backup to SD-cards, HDMI output, proprietary batteries and hard to find an AC-adapter for, not recommended.
In many larger venue's amplified audio is essential, so an audio hookup using an XLR cable is preferred (with the mic-signal from the speaker). This will allow us to get a clear and crisp audio input for our recording.
For smaller venues or if there is no audio equiptment present, we usually use some Samson Concert 88 series wireless microphones/beltpacks, as these are decent and affordable, though not as good as the (way to expensive) senheisers.
When a speaker used video and/or audio in their presentation, this will also be received when it is sent over the HDMI port (not always the case). One or more seperate handheld microphones are nice to have, so the presenter can be introduced or mic's can be put in the room for questions.
USB HDMI Capture
Affordable USB-3 HDMI capture cards are available from the usual chinese websites, and are mostly clones of Magewell capture cards. Expect to pay about $60 to $110 for them, and steer clear of the $20 ones, as these are SD-TV capture cards with an HDMI port. The quality on those cheap ones is beyond useless.
The 'good' ones will be marked with words such as: HDMI USB3.0 1080P HDMI Video Capture Card for Windows/Linux/Mac USB UVC UAC.
Basically, these are HDMI input ports which present themselves as generic USB webcam's and audio sources to the computer, and they work with the generic webcam drivers (USB Video Class, USB Audio Class) available in Windows, Mac and Linux.
Open Broadcasting Studio
The final piece of the pie is OBS, a bit of open-source software meant for webcam streaming. In my setup I usually hook up the 2 video/audio sources, name one 'Camera' and the other 'Speaker' or 'Projector', add a logo and titles (using the chatlog feature).
By orienting the camera vertically, the 2 video-streams can be combined more efficiently, leaving out less empty space. OBS allows you to configure various scenes beforehand, and quickly change between these at runtime. It's handy to have a full-screen projector version, some picture-in-picture scenes, and a default scene with everything.
In OBS you can also configure your livestream, for example to Youtube or Twitch (and various other sites/methods).
At most events (with enough bandwidth) we livestream to youtube and simultaniously record to disk. The camera itself is also recording constantly, and this recording can be used in case of issues with the computer-recording or as an audio-source then needed.
After the event is completed, the recordings are usually edited using kdenlive to cut out the idle bits before and after the talks, long questions where there isn't a microphone in the room and for delays in demo's / mistakes.
A title-card is added which will give some info about the video and the event. When editing is complete, the video is re-encoded as VP9/Webm for upload to youtube and the event website.
Upload to youtube
Uploading to youtube is done manually, with texts and titles copied from the event schedule. In case of CfgMgmtCamp, most speakers have a twitter-account, and a tweet is sent out with the link, title and hashtags for the video.
When the edited video's have been uploaded to youtube, the livestream versions are usually taken offline, as these are of lower quality.
Improvements for next time
For the next conference, some improvements that can be made:
Recodings in more rooms, even if it's just a static unmanned camera, as long as it's on a tripod or fixed mount, audio input would also be preferred, unless the camera is really up close and personal, so the internal mic would have enough signal.
Whitelist MAC-addresses for wired uplinks beforehand (Venue blocked youtube)