Editor's note: This article was originally published in Jan. 2005. It has been updated to reflect minor but important changes in two software packages referred to in that original tutorial.
Call me a lemming, but once I understood what podcasting was, I jumped aboard as quickly as possible. Podcasting could be seen as just another way to push audio to people who want to listen to it, but it's really a unique combination of subscription and publication.
My stumbling block? Turning interviews into podcasts using entirely digital production methods that would allow me to make phone calls and record directly to MP3 for simplicity. I worked it out, but not until I went through a lot of trial and error that I'll save you from.
Podcasters, who are as varied as bloggers, create their broadcasts from interviews, personal observations, phone calls, music snippets, or random nonsense. The most popular seem to be discussions and strange programming -- news read in rhyme while tap dancing, for instance -- that's totally unavailable in familiar forms. The secret sauce is RSS: podcasts are syndicated, meaning listeners just have to find them once, and receive them every time a broadcast is issued.
In my line of work as a freelance reporter in Seattle, I spend most of the time that I'm not testing hardware, software, or services -- or actually writing -- on the telephone, interviewing people from companies, fellow authors, folks deploying equipment, and analysts. I write most frequently about Wi-Fi and related wireless networking, and have found a pretty fervent interest in that industry.
I decided that podcasting interviews and audio that helped elucidate about topics that I cover, in the first-person, direct from the source, would help my site better meet its audience's expectations.
I've tried to record from landlines and cell phones in the past, and it's all compromise. I happen to have an office with other freelancers that's mostly concrete and has windows facing north. We're in the north end of Seattle, meaning that cell reception is impaired both by a paucity of towers and the bunker I work out of. And I don't have a landline, nor have I been particularly impressed with nonprofessional -- read: not super expensive -- products that allow recording from a landline.
I turned to Skype, which offers both computer-to-computer voice over IP (VoIP) with extremely high quality, as well as a call-out service for the public switched telephone network (PSTN) called SkypeOut. This offers cheap rates and generally good quality (see Skype on Mac OS X: A Hands-On Approach Part 1 and Part 2).
Skype was the ideal place to start because it would provide a pure digital stream that didn't need to be digitized later. I could record a conversation live and ostensibly in high fidelity with little or no editing later. Or so I thought. The devil is in the details, and he had me by the ankles, thrashing me around until it was resolved.
I'd read Phillip Torrone's excellent guide to podcasting, which was one of the earliest articles after podcasting became a realistic possibility. It's already a classic.
But I differed from Phillip in not wanting to use as many pieces of software, later convert to MP3 after initial recording, and in rejecting GarageBand for speed and simplicity's sake. I wanted something simpler, but also more powerful.
Here's the toolkit I assembled:
Soundflower (free): This virtual audio device (in 2-channel and 16-channel flavors) lets you mix sound in the computer by routing both input and output sound to the same virtual device. This becomes important in a moment. Phillip turned me on to this fine product.
Audio Hijack Pro (demoware, $32): Rogue Amoeba Software created this software to capture audio from any individual application or input device and route it to an output device or a file. The Pro version includes the ability to record directly to MP3.
Audion: This discontinued MP3 tool works in Panther and is now free -- which makes it the cheapest way to edit MP3 files natively without converting to AIFF and back to MP3.
USB headset: I picked a $30 unit with noise cancellation from Logitech, which has a whole lineup of USB headsets. A USB headset ensures no feedback and a much clearer sound for your end of the conversation when recorded. (Your interviewees will appreciate how crisp you sound as well.)
Skype: The last few releases of Skype include audio controls directly within the program: you can choose the source input and output, making it easier to set up the specific reroute needed so that you don't hear yourself.
Audio Hijack Pro can create a separate audio stream for the output of any program. So you can take Skype's output and pop it into a file. You can also take any system sound input, like a built-in microphone on a PowerBook, audio in, or a USB mike, and record from that as well. (Audio Hijack Pro has a zillion other features, including scheduled recordings, removing dead air, and splitting files. If you're recording audio from programs, check out their full feature list.)
What initially stymied me, and I was unable to convey for quite a while correctly to the fine people at Rogue Amoeba tech support, was that I wanted to accomplish two separate goals:
Because the Macintosh treats each sound stream as a separate item, hijacking the audio from Skype only records what the other person is saying; your microphone input is not mixed into that output. Very nifty for separation and quality and full duplexing of sound; not so nifty for my purposes.
Soundflower was part of the solution, allowing me to pass sound through from both input and output to a single audio stream that I could record, but monitoring it was a problem: I heard myself in a slightly delayed echo in the headset earphone.
Rogue Amoeba finally gave me a solution: their pro software includes a number of audio effects, and one of them was the ticket to making this work.
Download and install Soundflower, which requires a restart, and Audio Hijack Pro, Skype, and Audion, which do not. Buy a headset and plug it in. (The demonstration mode of Audio Hijack Pro allows 10 minutes of recording per file after which it inserts noise, so it's fine for testing.)
1. Before running Audio Hijack Pro or Skype, set up System Preferences > Sound. Set the Input to your USB headset and the output to your USB headset.
2. Launch Skype. Select File > Preferences, and click the Audio tab. Set the input to your USB headset. But set output to Soundflower (2-channel).
3. Launch Audio Hijack Pro. We're going to set up three separate hijacked streams, which is slightly complicated, but makes this all work as expected. Select Session > New to create a new audio stream and Session > Rename to rename it.
a. Create three streams named (to keep them at the top of the list and in order) "1 USB in to Soundflower 2ch," "2 Soundflower 2ch to file," and "3 Skype to monitor."
b. For "1 USB in to Soundflower 2ch," select Audio Device from the Source menu. Select your USB headset from the popup menu for Input Device. Select Soundflower (2ch) as Output Device. Click Hijack.
c. For "2 Soundflower 2ch to file," select Audio Device from the Source menu. Select Soundflower (2ch) as Input Device. Select Soundflower (16ch) as Output Device. This last step means that the output of this mix isn't played back at all; it's just a dummy method to record a file. Now click the Recording tab and set up the options. Typically, you want to record to uncompressed audio using AIFF (auto) in the Recording tab of this stream. An AIFF file takes 10 times or more space to store, but it's the best way to start. You can also record to an MP3 file with high quality, edit it in Audion, and compress it in iTunes, as described later. Click Hijack.
d. For "3 Skype to monitor," select Application from the Source menu, and choose Skype from the list of applications. A recent release of Audio Hijack Pro lets you set the target device for an application instead of just dumping it to a file. Because we've set Skype to dump its output into Soundflower (2 ch), we need to use this stream just to re-route the Skype output--your caller--into your headphones. In the Advanced dialog box, from the Target Device menu, select Soundflower (16 ch), which will just dump the sound. Click OK, and then click Effects tab in the lower right of the screen. Click in the upper left blank spot that reads Click Here to Insert Effect. Choose the Auxiliary Output submenu from the 4FX Effect menu. From the Editor that appears choose your headset as the Device; leave Source as default. Click Hijack.
To test this setup, make a Skype call to someone you know who won't be annoyed by you testing the service. You shouldn't hear yourself, but you should hear the sound of the Skype phone ring and the other person. They should hear you, too. (Note that you will need to keep the "3 Skype to monitor" set to Hijack as long as Skype is set to Soundflower (2ch) for output; otherwise, you won't be able to hear the other party.)
Click Record in the "2 Soundflower 2ch to file" stream to record while you speak to someone. Just record a few seconds of both of you talking to each other and click Record again to stop the recording.
To listen to the playback, click the Recording Bin item in the left pane of Audio Hijack Pro. You can select the file you just recorded and click Preview to have it played from within the program, or select the file and click iTunes to transfer it to play in iTunes.
If you cannot hear the other person or the recording lacks both parties, try quitting Audio Hijack Pro and Skype and launching them again. Occasionally, the audio settings lag behind the choice in a program, but relaunching seems to clear the state and enable your choice.
Unless you're very quick on the timing, you'll want to have a little lead in that you record. Audion lets you trim that lead in for MP3 files; if you recorded in AIFF format, you can use a variety of software, including SoundStudio Pro (http://www.felttip.com/products/soundstudio/).
You may want to record at a higher MP3 compression rate than you ultimately reduce your audio to. I haven't yet found the secret sauce that produces both a compact and high-quality audio recording of voices only from within Audio Hijack Pro. I'll record with the MP3 (Medium) setting, which is 128 Kbps in stereo using CBR (constant bit-rate). This provides a very high quality for the level of audio I'm sampling from.
I then use iTunes as my conversion engine. You can set iTunes to a low MP3 conversion factor, but be warned that you need to remember to reset this to your preferred ripping setting later for importing music, or you'll be a bit more than mildly disappointed.
These settings produce a slightly muffled, but perfectly decent sound, and they cut a 6-minute Skype interview from 5.5 MB to 1.4 MB. I'm still trying to find the perfect balance of size and sound quality. (Post your suggestions below.)
Podcasting may be a flash in the pan, but the volume of material being created and distributed shows a growing trend for people to take the means of (audio) production into their own hands and become narrowcasters rather than narrow listeners.
By adding "phone calls" (both VoIP and internet telephony varieties) into the mix, you can produce a podcast that sounds as much or as little like a radio show as you have the desire and ability to make it.
Glenn Fleishman is a freelance technology journalist contributing regularly to The New York Times, The Seattle Times, Macworld magazine, and InfoWorld. He maintains a wireless weblog at wifinetnews.com.
Return to the Mac DevCenter
Copyright © 2009 O'Reilly Media, Inc.