This is part II of my guide on podcast production, if you haven’t read part I please do. You can find it here. I’m afraid this one is unavoidably more technical than the recording guide, so bear with me and I’ll explain the best I can without bamboozling you too much. The idea here is that you can understand what it is you’re trying to achieve and then apply that to any software that you might happen to use.
So you have your recorded material, preferably in a high quality format. Now we need to clean it up, cut bits out if necessary and use a data compression algorithm (eg. MP3) to make it suitable for downloading quickly on the internet.
First step is get the file into an audio editing program. If you are looking for a free one use Audacity. I like to do everything in Cockos Reaper because it is more versatile and is capable of real-time processing. You can try the evaluation version for free.
This is the order in which I usually do things:
1) Import files
2) Trim/crop and edit and arrange
3) Noise reduction/removal
4) Equalisation (cutting out any unwanted or annoying frequencies)
5) Audio compression and limiting (reducing the range between loud and quiet and increasing the overall volume)
6) Finalising (converting to a media format suitable for fast download from the web)
Step 1 – Import Files
Import your audio file/s into your editing program of choice. Usually this is an option in the file menu. In Reaper it’s under ‘Insert’. If you are using more than one audio file you should put them on separate tracks/channels. So now you should have all your audio files loaded up.
Step 2 – Trimming/Cropping and Editing
Now to get choppy. This is where you are really putting together your show, making all the parts fit together smoothly.
Are there any unwanted sections of audio? – You may want to use just one part of the recording or get rid of some mistakes. Cut these out and trim and crop the start and end points. Piece it together.
Tip: In Reaper you can slice by positioning the play head over where you want to chop and use the ‘S‘ key to cut)
Have you got more than one vocal track? – You may have recorded a number of different participants each on their own microphone. These will each have an audio track that will need to be treated separately. You may want to fade them in or out where your participants are getting excited and talking over each other so as to avoid a big mess.
Do you have music? – Put your music on a separate track and fade in or out respectively.
Step 3 – Noise Reduction
Is there noise on the recording? – If your original recording has a lot of background noise (maybe it has been recorded at a very low level or was recorded in a noisy environment), you can employ a noise reduction processor to filter it out. This area is fairly complex and is a whole other post so instead I’ve provided you with some links if you want to take the leap into this area.
Step 4 – Equalisation
Are there any nasty sounds? “Boomyness” or “scratchyness”? These can be caused by a combination of the microphone, the acoustics, the position of the participant and vocal characteristics of the voice. Some audio frequencies will build up and cause the unpleasant afore-mentioned sounds. An equaliser helps to locate and reduce these frequencies, therefore equalising the frequency spectrum.
These “nasty sounds” can be found at certain frequencies along the frequency spectrum, so what you have to do is try and find it. This takes a bit of practice and can be tricky to know what you’re looking for but give it go and you’ll get better.
Finding the offending frequency
Here’s how I do it. So assuming you’ve got your equaliser open in front of you (use one that looks similar to the one in the picture below, which is ReaEQ in Reaper) and are listening to the recording. Increase the gain (volume) of one of the middle bands (band 2 in this case). Drag with your mouse slowly left and right, sweeping across the frequency spectrum until you hear the sound you want to get rid of. Congratulate yourself you’ve now found the offending frequency.
Note: You might need to reduce the bandwidth a bit to create a more prominent peak, it makes it easier to pin point the offending frequency band. (a bit like increasing the focus on a lens)
Examples of an ‘offending’ frequency may be ‘boxiness’, ‘boomyness’, ‘muffledness’ all subjective terms I know but these are usually associated with frequencies around the 200Hz-500Hz range. Other nastiness may be perceived as harshness, sibilance etc around the 5Khz-8Khz range. The idea is to reduce the level of a particular band of frequencies to make the overall sound, more pleasant. The reduction in gain can range greatly depending on your material but my advice would be to err on the side of caution when doing any processing.
Make the offending frequency go away
So you’ve found your nasty frequency that you want to remove by “sweeping” across the frequency spectrum until you hit your accentuated bit of nastiness? Now we just turn it down (reduce the gain) of the frequency band by about -3dB at first then if needed further increments until it sounds less prominent. The idea here is to equalise the loudness of all the frequencies we are hearing.
Cleaning up the low end
Standard practice is to apply a High-pass filter on spoken word at about 80Hz or 100Hz so as to omit any rumbling or low frequency traffic noise from the recording. High-pass lets through any sound above the threshold that you set it at. Anything below that threshold is filtered out and inaudible.
Step 5 – Compression and Limiting
So now you should have a better sounding voice recording, but what about those quiet bits that might get drowned out by your noisy neighbour hoovering up? Or those loud bits that might blow your ears off if you’ve got the volume turned way up just to hear the quiet bits? You see where I’m going with this? This is where compression comes in, it reduces the difference between the loud bits and the quiet bits. It makes the difference between them smaller (see pic below)
So ask yourself, are there any loud and quiet bits? – The speaker may have moved away or too close to the microphone causing large differences in volume. We can reduce these differences by compressing (turning down the loud bits) and then raising the overall volume again so that everything fits into a more consistent range that is more comfortable for the listener.
Setting the compressor
The two key controls in setting a compressor are the threshold (The level at which the signal starts to be compressed) and the ratio (amount of compression). The other controls like attack and decay can be left alone or put on auto if available.
Start by turning the ratio to maximum (infinity), then bring down the threshold from 0dB until you hear or see the sound/signal being compressed (look at the red bars). Don’t over-do it, your aim here should be to tighten up the difference between the loud and quiet sounds not make everything the same volume. Listen to what happens when you set the threshold too low, It’s bad, trust me.
Tip: You want the red bars to be ‘dancing’ around, not completely red or vice-versa.
Next set the amount of compression. Go back to the ratio setting and reduce it. For spoken word I would use a ratio between 2:1 and 4:1 then listen/look at how much compression (red bars) is happening. I would aim for around -6dB of compression. You may need to go back to fine tune the threshold to achieve your ‘dancing’ red bars. As this is pretty subjective to the material you’re working with, you’re just going to have to trust your ears on this one. You don’t really want to be hearing the compressor at work so as I have said before err on the side of caution.
Is it all loud enough?
The volume of the recording may be low when compared to other podcasts. If you’re listening to a series of podcasts, the listener doesn’t want to be turning the volume level up and down for each recording. Especially when listening on headphones with an MP3 player buried deep within their pocket.
You’ll need to turn the overall volume level up to the right kind of level but when you do this it may mean occasional sounds in the recording may go above the limitations of your recording medium. This results in an audible click or crackle (distortion) meaning that some peaks in the recording have breached (clipped) the overload point (0dB).
You can use a limiter to, primarily catch any occasional peaks in the recording from causing clipping (clicks and crackles) and secondly it means you can squeeze a bit more volume out of your recording so that the loudness is similar to other professionally produced podcasts.
A limiter starts limiting the sound level at a certain threshold. The idea is to catch sudden peaks in the recording that may sneak past the compressor. You can increase the level of your audio by about another 3.0dB without the affects of the limiter becoming too noticeable. If you push it too hard you will harm the sound quality and will introduce distortion.
Essentially a limiter doing the same thing as a compressor with but to a greater extreme and faster. ie. a high ratio (eg. anything above 8:1) and a fast attack time.
Setting the Limiter
The limiter is the final processor in the chain. Setting the limiter is pretty similar to the compressor only your aim is to catch those occasional peaks in volume. In Reaper you can setup the compressor plugin (ReaComp) as a limiter , Set a threshold to -0.1dB. Set the attack to quickest to catch those peaks. Tick ‘auto release‘ for convenience and max out the ratio to infinity. We can now increase the master volume of our audio until you see the limiter is just getting tickled by only the occasional loudest peaks.
Tip: A Limiter shouldn’t be pushed hard, no more than -3dB of limiting.
Well done for getting this far, nearly there.
Step 6 – Rendering/Exporting to Master
Hopefully your sound file is now how you want it. What we are going to do now is save it in a form suitable for the web. So we need to Render, Bounce, Mixdown, Export or whatever your software calls it. For example they all essentially give you the option to turn your recording into an Mp3.
You are rendering your audio file for the end user so the file will most likely be reduced in quality for faster download times. Your choice of bitrate is totally dependent on balancing sound quality over fast downloads as I touched on in part I. Quality is related to bitrate (kbps), so the higher the bitrate the better it will sound.
For reference here are the BBC’s format recommendations for podcasts:
- MP3 Mono Speech: 64 kbps, 44.1 kHz
- MP3 Stereo Music: 128 kbps, 44.1 kHz
Also check out this post from My Music Thing and see how they grade bitrate against sound quality (bear in mind they are talking about music rather than voice).
Make up your own mind, I personally would want to go higher than the Beeb, especially now that most people have fast enough internet speeds to deal with larger file size. I think it’s less of a problem now unless your audience are still on dial-up at the end of some copper wire, I guess it really comes down to your audience and what connection they are likely to have.
So a few things to check when doing your final render/bounce:
• Are you getting nice healthy levels without clipping? – Do red lights appear on your master levels? If so, reduce master volume. It’s quicker to check for clipping and make adjustment to your master fader around just the loudest parts of the recording. You can adjust, test, adjust some more, test, until you no longer get any clipping.
• Choose your file format eg. Mp3 CBR (Constant bit rate) 128Kbps Stere
• Engage ‘Dither’ (if available) if you’re converting from one bit depth to another, which you most likely will be.
• You’re ready to Render/Export
Well done you just learnt about producing a podcast. I know there’s a lot more to making a podcast than getting the sound right but hopefully this guide has given you a couple of extra strings to your bow. Now go make yourself a brew and listen back to it.
Of course if you don’t want to do any of this, can’t find the time or simply want it to be done by a professional. Check out Our Services page now, we are more than happy take on the task.