RogerBW's Blog

Now that I'm involved in a boardgaming podcast as well as the role-playing one, and all recording is remote, I've been recording multiple tracks and editing them together.

There are sites that offer this: Zencastr for example, and for some guests we've used this. But the free tier only gives you MP3 recordings, and of course it's subject to change or withdrawal at any time.

I'm looking for a Javascript/etc. solution that will record and then compress to FLAC from within a web page, but I haven't found a good one yet. (Must also allow selection of inputs, and give some indication that it's running.) Windows used to have a sound recorder but I'm told that in recent versions it's missing, or only gives you lossy audio, or has a limited duration, or some such nonsense.

So what I mostly do is encourage people to use Audacity (available on every desktop platform, free, configurable in a standard way) to make a mono 48kHz recording of their own microphone, and to wear headphones so that that track only has their speech on it.

This offers some advantages over simply recording the chat session: it's less vulnerable to network problems, it doesn't record any bips and bongs made by the videoconference platform, and it gives me each speaker on a separate track. This helps to disentangle things when people talk over each other, but more importantly it lets me space them out across the stereo field to make it easier for listeners to keep track of who's who. (Generally the more someone speaks the more I put them towards the middle, with the GM right in the centre; for a given RPG adventure or campaign I maintain the positions from session to session.)

One potential problem is timing: even with modern machines it's not unusual for someone to become half a second off between start and end of a two-hour session. If that displacement is small, I just line the track up to minimise total error; if it's larger, Audacity has a "change tempo" which can squeeze or stretch audio to the right speed, but it takes a while even on a new fast machine.

Then each track gets its own individual noise reduction; I do any actual editing, normalise, set levels appropriately, and mix them together – then compress the overall result (but not any music between segments) rather than each one individually, because compressing one speaker tends to amplify their silences.

This doesn't take significantly longer than the two-track edits I've been doing before and I feel the result is worth the very minor effort. (Of course it's a bit more effort for the other participants, as they have to make their own recordings and then get them to me…)