Recording and editing Two's Complement

Like so many, I started a podcast – Two’s Complement – during the pandemic, with my good friend and colleague Ben Rady.

Some folks asked how we made it sound so “professional”, and so I’m sharing here how we do it.

Microphone and setup

Most importantly: get everyone to record their audio locally. Audacity is free and works perfectly for this. Make sure you test that you can’t hear each other’s sound in the mix. Start recording, make sure the right microphone is in use by tapping it and observing the wiggly lines and then get others to make noise and ensure you hear them but don’t see the wiggly lines on your side.

I use a Blue snowball mic on a boom with a pop filter, but Ben uses a simple headset mic, not unlike anything you’d use for normal video conferencing.

We use a Google meet to talk to each other (and a guest if we have one), and make sure again that the recording is working while Google meet is up and running.

Of course, we ask folks to stay close to the mic, near enough that the wiggly lines approach – but don’t hit – the top of the recording chart. We ask people to minimise external noise, though it’s amazing what you can edit out.

At the end of the recording we ask everyone to export their side of the conversation as an MP3 and we upload them all to a shared google drive.

Editing

I use Audacity to edit everything. The first thing I do is bring in all the MP3s into one workspace, one above each other. I then align them so the conversations are lined up.

I like to position each speaker in stereo, using the little L<->R slider on each channel. That helps make it sound more like a conversation.

I can’t remember if it’s on by default, but I have Audacity set up so I can click in the timeline and just start playing at that point.

Once lined up, I’ll find the beginning and end of the “take” and cut them off. Hotkeys are essential here!

• F1 is “select mode”, click and find the beginning of the podcast.
• Shift-J will select backwards to the beginning of the cliop
• Ctrl-Shift-K selects the same region across all the channels
• Backspace deletes
• Shift-K selects to the end of the clip

I do a super rough cut out to start with.

Next up some filters to make things sound nicer, for each track:

• Select an area of blank sound on a track (you should have lots of these where the other person is talking). Go to Effects->Noise Reduction and then click “Get Noise Profile”. Now select the whole track (there’s a button on the far left to do so). Go back to Effects->Noise Reduction and hit ok. I find the defaults work, and the amount of background hiss it reduces is magic.
• Pick Effects->Amplify and accept the defaults. This should bring everything up to “max” volume without overdoing anything. We will later address the fact folks get louder and quieter (compression) but we’ll do that after editing.

Now the long spot: actually editing. I find it quite therapeutic. My flow is typically:

• Listen to the podcast.
• If I spot an egregious pause or “umm” or “err”, I select it (F1 and drag over it). If it’s only affecting one track I’ll hit ‘Z’ (which expands the selection to the next zero crossing, slightly reducing pops). I’ll then hit ctrl-shift-K to expand the selection to all tracks, then hit backspace to delete both tracks together (keeping them in step).
• If there’s background noise from a person who’s not speaking, I’ll select the offending part of that track and hit Ctrl-Alt-K which cuts out their track but unlike Backspace does not move everything down, instead leaving a “hole”.
• If there’s more surgery required between the two speakers, I “cut around” the blank spaces where each person is silent, using Ctrl-Alt-K, then I’ll use the Move Tool (F5) to slide the remaining uncut parts where each person is speaking around relative to each other. This takes the most care as it’s easy to get things out of sync. But it’s the most powerful thing to do, allowing for removing people accidentally talking over each other and so on.

Once all the editing is done, I then “compress” the audio – this has nothing to do with data compression but to do with normalising the loudness. I select each track in turn and then Effect->Compressor with

• Threshold (at which loudness the amplifier kicks in) as low as reasonable. I’m currently at about -20dB.
• Noise Floor (below which no amplification happens) as high as can be. Mine’s set to -60dB but I fiddle with this to try and get rid of more noise.
• Ratio (how much amplification happens in the quiet bits) - Play around - mine’s at 4:1 right now. Too low and quiet bits are still quiet. Too high and EVERYTHING SEEMS A BIT ANGRY.
• Attack and release times as low as possible.
• Make sure “compress based on peaks” is set.

It’s easy to go overboard, and remove every “lip smack” and “umm” and “err”. Doing so makes for an exhausting listening experience; humans do um and err, and they can be a way for the listener to get a short break. So don’t edit too heavily. The first few episodes of TC suffer from this, if you want a “how not to”.

Editing on music and such is probably enough work to put in a separate post if anyone is interested.

Transcripting

It’s important to have a transcript for hearing impaired folks, non-native speakers, for SEO, and to help people understand what you’re talking about without listening in. We use Temi, which has worked out fairly well, costing us around $10 an episode.