A Workflow for Podcast Audio in Audacity
For some time, I have been using this workflow for processing raw audio into clear, clean, production quality files for ebooks and podcasts.
These steps are for use in Audacity; you can adjust them as necessary for other editors.
Original Recording Quality & Format
First, make sure the original is recorded in as high a quality as possible. Follow the easy to find best practices: quiet location, quality microphone, wind screen and pop filter on the mic, etc.
Make it a high priority to capture the voice (or voices) at a moderate level, but never drive the recorder into clipping. Make sure there is as little hum as possible on the audio. If necessary, use battery power for the mic and recorder. If you have problems with power line hum, see the tutorial for reducing power mains hum in your audio chain
As to format, you cannot go wrong with 24 bit / 96 kHz sampling, if your equipment can do it. Otherwise, use 16 bits if you cannot record at 24 and 48 kHz sampling if you cannot run at 96 kHz.
I know some older, legacy artists may still use CD sampling at 16 bits / 44.1 kHz. Unless you are making CDs, don't bother with that. The future is sampling in multiples of 48 kHz.
1: Sync if Multiple Tracks
When I get the raw recordings, the first thing I do is open a new project in Audacity and - if there are multiple participants - get all of the tracks synced in time. Syncing isn't necessary if everyone's mic is recorded on one multitrack device.
Line up the tracks on some kind of identifiable sound as early as possible. Use a hand clap, count-in, or something you can see on an audio envelope or spectrogram.
If there are multiple recorders in use, you'll notice that the other tracks eventually drift out of sync. Make the best track the "main" or "master" and then stretch or shrink the other tracks to match.
All of the text below refers to one track, but should be done for all tracks it you are working on a podcast with multiple tracks.
Leave a few seconds of dead air / room tone, containing the ambient background sounds, before and after the program.
Normalize the sound level of all of tracks to -3 dB relative to full scale (-3 dBFS).
Cut out sections with extended silences. Make the same cut across all tracks, as it is essential to keep the whole project in sync.
If there are sneezes, coughs, or odd background noises on any track, replace them with silence.
Apply noise reduction to tracks individually. Audacity is great for identifying and removing rhythmic sounds, such as fans, motors, or anything which whines or buzzes.
To get a profile for the noise, find a few seconds with none of the desired content - some dead air between sentences, for example. Highlight it and run the noise reduction "get profile" function.
After profiling the noise, highlight the whole track and apply noise reduction. Use the maximum number of frequency bands. Don't attack the noise too aggressively, as it makes the recording sound hollow and unnatural. Just apply 6 dB of noise reduction.
If the noise is still there, take a new profile then apply another 6 dB of noise reduction.
5: Re-Normalize the Levels
Normalize the sound level of all of tracks to -3 dB relative to full scale (-3 dBFS).
The noise gate lowers the background level when there is no strong signal, as between sentences. Work with the settings. I like to set the attenuation to about - 20 dB, with only a few milliseconds (ms) of attack time, about a second (1000 ms) of hold time, and a half second (500 ms) of decay. But work with what you have and go with what sounds best for your situation.
7: Re-Normalize the Levels
Normalize the sound level of all of tracks to -3 dB relative to full scale (-3 dBFS).
8: Apply Amplitude Compression
To bring the peak and average voice levels closer, which is to make the sound more even, use the compressor. Depending on the plugins you have, there may be several choices available. There might be more settings or fewer settings to tweak. In a general sense, these settings work:
Ratio: between 3:1 and 5:1
Adjust the threshold, compression ratio, and timings as necessary for best sound of your content.
9: Apply Limiting to Catch the Strongest Peaks
The compressor will let some of the loudest peaks get through, although they will be moderated. Use the limiter to catch those peaks.
Threshold: Set to match level of other peaks.
10: Re-Normalize the Levels
Normalize the sound level of all of tracks to -3 dB relative to full scale (-3 dBFS).
11: Amplify / Attenuate for Platform Compliance
Amplify the whole project + / - a few dB as necessary to meet the requirements of the distribution platform. Whether Spotify, YouTube, SoundCloud, or other platforms, there is some kind of standard for peak and average loudness levels. Also, there is often a requirement that noise be below a certain level.
12: Set the Pre and Post Empty Space
Trim off the dead air before and after the program content. I find that one second of room tone before the start and four seconds after the end is sufficient most of the time.
