Behind the Scenes with "Teh Funny"

Part of the reason for recording and posting the sessions from Northern Voice is to learn and share the experience of doing event videos, which I hope will encourage others to do more of the same. Here's the short story behind the Teh Funny. More extensive details are below.

The video was shot by three cameras, one of which was a cell phone camera. Audio was recorded on a standalone audio recorder. Slides were recorded directly from the presenter's laptop. These five recording sources were synchronized with our software and then edited in Final Cut Pro.

There's a lot more to it as I'll now explain. This is pretty detailed technical info, so hide your eyes if you find it too scary.

Slides

The slides were recorded onto a T61 laptop using an Epiphan VGA2USB and VirtualDub. I used the ffvfw MPEG-4 codec and recorded audio from the laptop built-in mics. That's poor quality audio but it was only needed for synchronization purposes later so it was enough. Unfortunately the quality of the video was not good. The Epiphan box does a good job about 90% of the time, but this was not one of those times. So I asked the presenter for his slides. They were in Apple's Keynote which has an easy way to export all the slides to individual images: look under Share > Export... I did use the Epiphan recording, but just to know where to place the slide images in the timeline.

Video

I recorded video from the back of the room using a Canon XH A1. It was equipped with a Rode shotgun mic for backup audio and to pick up some room tone and audience sounds. Another attendee recorded B-roll footage using a monopod at the front of the hall. (You can see her on the right in some of the early frames.) A third attendee posted live video to Qik which was downloaded later.

Audio

The primary audio was recorded using a wireless lavalier mic into a Zoom H4. Secondary audio was to come from the camera-mounted shotgun mic. These would be mixed after synchronization.

Post-Production

I dumped everything into Final Cut Pro and used our software to automatically synchronize it. It worked like a charm. The Zoom audio had a bit of drift but the software corrected for that and it was perfectly synced with the shotgun audio so they could be mixed. Normally at this point I would throw both audio tracks into The Levelator and then do the editing.

That would be the end of the story, but there was a problem. I'm fussy about the quality of the audio and I've had excellent results using this setup, but in this case the primary audio recording was beyond bad. First of all, the lavalier mic on the presenter's shirt kind of flopped over, so there was lots of bumps and clothing noise. Second there was electrical interference with the wireless system. I'd tested it all beforehand, but once the room was filled with dozens of laptops and all the AV systems fired up, something caused nasty hissing and zapping sounds.

I knew some of this was happening during the shoot, but I couldn't bring myself to interrupt the presenter to tinker with everything. So I decided to fix it as best as possible in post.

It was a nightmare. I spent a loooong time in Adobe Audition cleaning it up and I think I used every trick that program has to offer. There were three different noise profiles used at various times, lots of gating, some EQ, envelope shaping and the healing brush. The more I used Audition the more I appreciated what it can do. Not only are there lots of tools, but they have implemented the algorithms very well.

In the end though, it's garbage in and just slightly less smelly garbage out. After mixing in much more camera audio than was originally intended, it was done. The best that can be said about it is that the speaker is intelligible and it doesn't actually make your ears bleed.

Compared to the audio, the rest of the editing process was a breeze.

Lessons Learned

1. Watch where you clip on those lavalier mics. Get it as close as possible to the mouth but at all costs keep it from rubbing or bumping clothing or anything else.
2. Wireless mics can work well but if you have any reasonable alternative, go for it. In other sessions at this conference, we used an Olympus portable voice recorder with excellent results. With hindsight I wished I had fed the wireless audio into the other audio track on the camera instead of the Zoom and instead just propped up the Zoom on the lectern to record with its own mics (as backup audio).
3. If people are recording to tape, just grab the tapes from them at the end of the day. We ended up having to do a complicated compress-upload-download cycle that slowed things down a lot.
4. It's a good idea to have a USB key handy to grab a copy of the slides from the presenter right after the presentation. As much as I love the Epiphan box, the original slides are better quality and it can make a noticeable difference even at web resolutions.
5. By all means, go with multicamera/multiaudio recording. If you ignore my whining about the audio quality problems, you'll see that the process was amazingly simple despite the number of variables involved. By doing synchronization in post, the cameras and audio don't need to be connected to each other during recording. And by having a precise sync on the audio, you can get a pleasing mix of presenter and room audio.