Podcasting Live to Tape

tl;dr: Don’t use Skype, don’t do double-enders, do it like a live radio show.

podfatherIn 2004, Adam Curry laments on his second Daily Source Code that Mac software for doing podcasts sucks (and the first show was recorded in his car). Well, 10 years later, the situation isn’t much better, and Skype isn’t helping. I’ve been there, and I don’t want to go back.

By recording one good version as you are doing the show, the post production effort shrinks considerably. Double-enders even need additional setup: This means instructing guests on how to do that, making sure they record the same mic they also use for the voip connection, e.g. Skype. They also will have to deliver the file via some kind of file sharing mechanism after the show. And while that’s centainly doable, all this gets really old really quick.

So here’s our battle-tested method of doing podcasts live-to-tape. We’ve been doing our show somewhat like this for years now, week after week, with a minimum of editing, and with very nicely sounding audio.

Don’t use Skype

As to why Skype still uses the SILK codec and only offers manual quality control to their broadcast customers, I really don’t know. I think Skype TX should’ve rolled out by the end of the year, but it still hasn’t.

Skype still adds artificial noise to the audio, messes with dynamics and adds a very noticeable latency. And because Skype tries to get around every horrible network in this world, you sometimes get a horrible relayed connection that just randomly cuts out or degrades quality. Just don’t use Skype unless absolutely necessary.

Bizarrely enough, Skype participated in developing the awesome Opus codec, but still doesn’t use it in their flagship product two years after announcing it. So far, the most reliable and widely available software that uses Opus is Mumble. I’ve written about Mumble as an alternative to Skype here and here. Other SIP-based software exists, but nothing great (yet). Luci Live is broadcast-grade software that unfortunately has its receiving end only on Windows or in dedicated broadcast-priced hardware.

Opus sounds great in two very important ways and is highly configurable and optimizable. Firstly, its inherent latency is very low and you can optimize Mumble parameters and server location to optimize for network latency. Secondly, audio quality is nearly indistinguishable from uncompressed audio at very low bit rates. Anything above 70kbit/s will make hardly any difference, especially after reencoding your final MP3 or AAC podcast file. Also, this will fit through just about any pipe. We have one regular participant behind an somewhat shaky LTE connection as far as VOIP is concerned, but simply doubling his buffer to 20ms cleared most of a tiny bit of audible crackling. On wired connections, there hardly ever is a dropout.

So now you just killed an entire flock of birds with one stone.

  • Thanks to the low latency, you just stopped talking over each other, just like you’re sitting in one room, having a conversation. That’s what it’s about, right?
  • When somebody loses their connection, no more “dialing” and “picking up”, they just join the room and are back.
  • Thanks to the audio quality, you dont have to do double-enders any more. No more audio drift, no more echoes. The audio latency “truth” only exists in your master recording, just the way you heard it while recording.
  • No guest will ever again forget to record their side and you won’t have to puzzle pieces together.
  • Boom, your live stream suddenly sounds great.
  • You get a free chat service in Mumble, and you don’t need to hunt it down in every new version. Hinthint, Microsoft.
  • No more editing for Skype-related garbage audio or double-enders*

* Now obviously, if something goes really wrong, if you lose your internet connection and have 5 minutes of dead air, by all means, set a marker or inspect the waveform afterwards for awkward silence, and edit out that part. If you really hit some topic in the discussion that leads nowhere, excise it and throw it out, but don’t micromanage every single line somebody said. If there’s one small dropout in a show, let it play out, no listener will mind, if it keeps the show going.

Do it Live

You shouldn’t plan on “we’ll fix it in post”. Imagine there’s a live studio audience that will start throwing things at you when you mess up. Imagine there’s an angry pointy-haired boss breathing down your neck if you don’t start the show on time. This should keep you on your toes and should make your show more fun to listen to.

I’m talking about the regular roundtable discussion show with two to six people talking to each other. Now this might not always apply to an overproduced magazine, but even that can be done live and to great effect.

And Tape it

This also explains my philosophy on recording: I believe computers just aren’t made for handling audio reliably, especially when done over scary USB, or when handling too many sources in one box. Do it in hardware. Do a live mix on a simple hardware mixer, plug in analog mics and your computer that runs Mumble with good old wires. Record that mono or stereo mix on a little hardware recorder. Multitrack recording certainly has its applications, but for the majority of podcasts I think it’s overkill.

The only time I ever came close to losing an entire show was when I foolishly used Soundtrack Pro to record. Recording something close to 1,500 hours of shows on this little Edirol R-4 never lost a single second of audio. Similar modern recorders like its descendant, the R-44 or smaller sub $100 handheld units write to solid state memory and should be exhibit similar soundness. Soundness, get it, because it’s an audio recorder. Ah, never mind.

If you’re live streaming the show, consider getting a little dedicated box that won’t futz with your computer’s audio. Another 50 dollars for a Raspberry Pi-type computer and sound interface or some old laptop and you’re done.

Even if modern radio doesn’t sound like it, podcasting is deeply rooted in radio. People talking to each other. Or one crazy person talking to himself. Hi Adam! Listeners listening to it. Radio is supposed to be done live. Do your podcasts live. And tape them.