I was very close to doing this as a podcast or video post, and eventually realized it was far easier to type it out and use illustrations than it was to try and juggle the recording and screen capture processes, so here we have it – I would apologize to those who hate facing a wall of text but unless you have an actual, physical handicap, go fuck yourself and deal with reading. If you do have a physical handicap, contact me and I’ll be glad to record it.

My recording purposes have been primarily for three reasons: podcasts, capturing interesting sounds from nature, and voiceovers to interesting videos from nature, with a couple of game demos in there. It’s rarely scripted even slightly now, the closest I’m coming is to demonstrate a particular technique, like the previous Tip Jar and several before that. I’ve done scripted stuff, and this tends to come off sounding a little pompous and unrealistic to me, so I’ve been avoiding it. But this also means that I flub a lot, change pitch and tempo, and other things I shouldn’t be doing, so some of this is to correct for these (rather than, as I should be doing, start learning how to speak for recording, which I do recommend even when I’m not jumping on that task myself.)

Equipment.

Use a good microphone. Listen, I feel ya, mics are expensive, often for no good reason – I can’t fathom how a microphone could cost more than I paid for my first car, or any camera that I’ve purchased. It’s a ripoff, pure and simple, but – the benefit from a clean mic is significant, so finding that balance point, or that sudden bargain, is useful. I am presently using a Sampson G-Track that I picked up used for just within my stingey budget, and the background hiss and noise is virtually nonexistent; it makes editing significantly easier.

Use a pop filter. Condenser mics are notorious for capturing those little sudden puffs of air when we make “P” and “B” sounds and similar, exactly what the word “pop” produces, and these can come through as louder thumps in the recording. A pop filter is simply a nylon screen in front of the mic that breaks up the force of this pop, and it works wonders. They’re easy enough to find cheap, and easy to make, but having it mounted directly to the mic is easiest.

Use a wind filter. Just like the ‘pop’ sounds we make, wind outdoors (or even under a fan) can make rumbles come up in audio, and the solution to this is a plush, furry cover to the mic, which does the same breaking down of the wind force, just more so. They’re easy enough to make it you (like I do) balk at paying a ridiculous price for a bit of synthetic fur, or (like I do) have mics for which no one ever made a wind filter in that size. Often called, “dead cats” in the biz, just so you know. They help a lot, and I always recommend using one for any outdoor recording – even those gentle, unnoticed breezes can ruin a recording.

Try not to ever use on-camera mics. They’re notorious for being not very good in quality, and almost always prone to wind noise. If you have the external mic port, use it.

Room acoustics. There’s no real need for a ‘studio’ or acoustic baffles or tile, but carpet helps reduce room echo tremendously, as does clutter – just breaking up the path that sound can bounce back from the walls will make an audible difference. Sometimes a quarter-turn in the direction you’re facing can alter the sound quality notably. Also, just hanging a blanket from the ceiling as a backdrop can chop off room acoustics. Experiment.

Find the right position for your mic, and yourself. Do a session or three of just experimenting with where the mic sits, how you should be sitting or standing, chin up or down, direct or indirect, and so on. In my experience, leaning back is bad for multiple reasons, even when I’d rather be comfortable, and having the mic a little above my mouth and facing down produces the best sound – you may not see this much, because it blocks the speaker’s voice, but if you see studio and voiceover recording, it’s used more often there.

I have terrible sinuses and my voice quality (I use that term very loosely) changes significantly on different days. I usually try to pick a ‘good’ day, get some hot tea to clear sinuses and nasal drip, and ensure the air passages are as good as they can get. I have some recordings done during a post-nasal drip and seriously, you don’t want to hear them.

Habits.

Silence. Record 6-8 seconds of total silence at the beginning. You will use this in post below, and it’s extremely helpful. It also lets you pause and mentally prepare to start speaking. If the room acoustics or background noise change suddenly (like the HVAC unit kicking in,) pause again and get a new patch of ‘silence’ for the background noise level with this change.

However, if you are doing a screen-capture to demonstrate something on the computer, start the screen-capture process first. Then begin your recording through Audacity/et al, and what I recommend is be producing some steady sound (hum, whatever) as you click the Record button. What this does is give you both an audible and visual cue to sync the video and audio together, because you’ll have video of the recording starting through the capture software. Later on (after audio editing, to a degree anyway,) you’ll line up the audio and video tracks using this point at the beginning, which can be deleted once the two are linked.

Slow down. Even if you’re trying to fit within a specific time frame, such as having 8.2 seconds of video space to fit in with, trying to talk fast rarely helps, and usually just slurs words together – it often takes training or at least practice to speak both fast and clearly. If you don’t have those time constraints, just ease up and relax a bit, you’ll sound better. Also, long pauses are not anything to worry about, because these can be removed.

Immediately redo a flub. If you stumble or stutter, or mispronounce a word, or what-have-you, just pause, exhale once, and speak your line again. Go back as far as you need to if you need to redo a sentence or two, or even if you realize you’re talking too fast or saying, “um” too much – I have a bad habit (well, one of many) of running “um” directly into the next words so even editing those out doesn’t sound right.

If you’re doing a screen-capture demo and recording audio as well as video, be aware of what was happening onscreen as you pause to repeat a line, since both will have to be edited out together – try to avoid sudden screen jumps and such. If you’re doing voiceover while viewing an already-edited video, as I do fairly frequently, go ahead and back up the video to a ‘clean’ point before you even started that sentence and say it again in sync.

Be aware of volume. I’m notoriously bad about changing volume, either with sudden louder portions or simply getting caught up in what I’m doing and not paying attention to how loud or quiet I’m getting. While these can be corrected in post (covered below,) it’s a hell of a lot easier not to have to – it’s part of the proper speaking habits I’ve never developed, and why I recommend developing them.

All of these make the next part a lot easier, which is why I’ve led off with them.

Editing.

Everything here is using Audacity, which is a free, open source audio editor that does so much you should be donating to them (yes, I am.) Other editing programs are similar in may ways, but I can’t address particulars.

Most of the below has come from this page, so full credit to them, but I’ve added a couple of bits from my own experience. So here’s the overall process, which is easier than it initially sounds.

Save the original recording. You may need it again, even though Audacity automatically saves all edits when you save the project, and can be reversed to the start – you may not want to go back through thirty edits.

By the way, using a ‘lossless’ format like AIFF, FLAC, or WAV to save recordings avoids the problems with truncated dynamic range that MP3 and such produce. These may make very large files, but if you’re doing voiceover for video, this will get reduced in size when you render the video into the final format.

Start with Noise Reduction. Multi-step portion here, but once you’ve done it a couple of times it’s automatic in your head.

1 – Select the silence that you recorded at the beginning. Make sure you have no extraneous clicks or bumps in your selection – it should be the flattest waveform you can find.

2 – Go to Effect/Noise Removal and Repair/Noise Reduction…, and click on the Get Noise Profile button. This saved your silence as a baseline.

3 – Select the entire recording now. (See note below.)

4 – Go back into Effect/Noise Removal and Repair/Noise Reduction…, and this time, we can hit Apply. This may take several seconds, depending on the length of the recording. Initial settings may take some experimentation – I am using Noise Reduction (dB): 15, Sensitivity: 5.51, Frequency smoothing (bands): 5, and make sure Noise: Reduce is checked and not Residue. Audacity will save these settings so they’re the same every time you open it, thus you won’t need to keep notes on this, though you may need to change them with a different mic.

That’s it. Listen to the recording, and the background hiss should be so reduced that you should barely be able to tell when it even starts. What this is doing is subtracting the background frequencies you selected from the entire recording, and it works best with the cleanest recordings.

Initially, however, you may want to change these settings – this should only have to be done once per mic or room setup. Once you get the initial noise level, instead of selecting the entire recording for the second step, select just a small patch in the middle someplace, that has typical vocals and at least a brief patch of silence. Use the Preview button to test your settings here (it only runs for six seconds,) and once you’re happy, note these settings, back out without applying it and re-select the entire recording, then apply those settings. The more background noise or hiss there is, the worse the results may have warble or truncated effects, so you may want to reduce the sensitivity or the noise reduction level, and have a little hiss rather than sound like you’re talking partially underwater.

The biggest alteration to the above. If the room acoustics changed during recording, as noted above, select the Noise Profile for the first, and then select only the portion of the recording before the background noise changed, and Apply to that. Then, select the new ‘silent’ section from after the background noise changed, and apply that to only the portion that has that background. If you do not do this here, before anything else, you will have a hard time getting rid of it later – voice of experience here.

Moving on.

My mic only records on one stereo channel, leaving the other blank. If this is the case, right-click over in that space at the left of the track and select, Split Stereo Track. All alterations below should be done solely on the track that has recording – having a silent track in there can alter the effects, so this works better.

Select the entire track for these.

Initial Normalize. Go into Effect/Volume and Compression/Normalize, and make sure Remove DC Offset is checked, then set Normalize Peak Amplitude to 0.0 dB. You may see a jump in the waveform. you may not – all this is doing is increasing the volume so the highest points hit the upper limit of the recording – it helps with the next step.

Compress. I need this seriously, because as I said, I tend to wander a bit too much in my volume, and what this does is even things out a bit. Go into Effect/Volume and Compression/Compressor… – you will get a graph and variety of settings. You may be fine with the default, or you may (like me) have to change them to be more aggressive – these are my settings:

What this does is to bring up the lower patches and reduce the higher patches, making the recording more evenly-toned throughout. Again, not a bad place to pick a small patch, one with a varied waveform, and Preview it, but like before, note the settings and deselect that small patch to select the entire recording/track to apply those settings to. This is one you may revisit from time to time to improve it, because the changes are subtle and may only show well in certain sections.

Tweak your vocals. This is very subjective, and what I provide is only a starting point. We’re going to apply some specific pitch changes to make us sound a little better – calm down, I said a little, nothing’s going to make me sound like James Earl Jones no matter what, Audacity isn’t that good. So go into Effect/EQ and Filters/Filter Curve EQ, and you’ll get a graph, or even a flat line. Mine, after much tweaking, looks like this:

This curve cuts out the lowest background noise (the left side at and below 60Hz,) a slight curving lift between 60 and 500Hz to improve the bass tones, then a slight increase up to 8000Hz to not let things get too muddy and low-toned. Your best results, however, may only slightly resemble this, if at all – it depends on your voice, and to a lesser extent your mic and acoustics.

You can see that the average of the curve falls above the centerline – this makes the volume jump a bit higher, and you will see the waveform get broader as you apply it. You can alter the curve here to be more centered, or not worry about it, because the next step will correct things.

Second Normalize. Same as the Initial Normalize, only this time you’ll change Peak Amplitude (dB) to -2.0 or thereabouts – experiment. What this means is, the loudest portion of your recording, the uppermost peaks of that waveform, will fall 2 decibels below the upper limits for recording, which is generally a good range for voice. If you’re a very animated speaker, this might cause the waveform and volume to drop notably because your peaks are so high, in which case you might want more aggressive Compressor settings.

Export this as a lossless format. This helps with the next step, or more specifically, where the next step has some issues. I usually call mine something like, “Voiceover-pregate.wav.”

Apply Noise Gate. I definitely need this, because I’m a breathy speaker and too many of my inhalations come up on the audio, not to mention lip sounds and all that. Noise Gate drops out anything that falls below a very low volume threshold, and can wipe these out entirely – I used to have to manually remove these, and that took ages.

But it’s slightly tricky to get set, initially. Once you do, you should need only very minor tweaks for any application, if at all. So here’s the rundown, with an illustration:

1 – Select a portion of the audio with the worst low-threshold noise you can find. In the case of the illustration, it’s the part selected in medium-blue, an intentional loud inhalation.

2 – Right click on the decibel range bar immediately left of the waveform, and select ‘Logarithmic (dB)’. You can grab the bottom of the whole track window to expand it for more detail, as I have here. The peak of the selected portion of the waveform reaches roughly -27 dB (red line) – this is our guide. You can see also that the speech outside of the selection is much louder, reaching about -10 db (green line.)

3 – Open Effect/Noise Removal and Repair/Noise Gate… – you’ll get a window like that shown here. The key is the Gate Threshold, here at -19 dB – a little louder than the peak of my noisy inhalation, but lower than the speech. This means that everything in the wave form below -19 dB (greater negative number) will be wiped. Level Reduction is actually greater than the threshold, so it should produce silence.

Attack, Hold, and Decay all relate to how soon and how long this reduction will take place – mine could probably stand some tweaking because of the next bit, but this is where it stands now.

4 – Preview the result, and if happy with it, apply these settings to the entire track.

Again, these settings will remain, so from here on out you can just apply the Noise Gate without changing anything, as long as your recording is reasonably consistent.

Now, your audio should be crisp and clean overall, but still might need a few cleaned up parts here and there, especially deleting out the “ums” and flurbles. Still leave the stereo channels separated, if you had to in the first place.

If syncing with video or anything else, it’s important not to delete any portion of the track until it’s linked to the video and can be deleted together – otherwise you’ll spend a ridiculous amount of time trying to sync up separate sections of audio and video. Instead, select the offending section and either go into Effect/Volume and Compression/Amplify… and choose an adequate negative number (I usually use -14 dB,) or choose Generate/Silence… instead – it will automatically be set to the length of your selection. If you have any background noise for either of these, the dropout will be noticeable though, so hopefully you don’t.

Also, if you have any hard peaks in your speech that you’d like to reduce, select those peaks and Effect/Volume and Compression/Amplify… to a setting of -2 to -3 dB, enough to reduce the harshness but not go too quiet. The same can be done in reverse if you start mumbling too much.

The Noise Gate may have a detrimental effect, though, in that it drops off “S” and sometimes “TH” sounds at the ends of sentences, where my settings may need further tweaking. This, however, is where that “-pregate” file you exported right before doing the Noise Gate step now comes in, in that you can go into that file and find the matching portion with the esses intact, and simply cut-and-paste it in. To me, it happens perhaps three or four times a session, so not a big deal, but it does sound so much better than abruptly truncated words.

Re-recording. On occasion, you may find that you completely screwed up something: used the wrong name or term, mispronounced a word consistently, totally forgot to tell listeners not to pull the pin just yet – whatever. And so, for clarity’s sake, you decide to re-record that section to patch in seamlessly. Hopefully, you’re better than I am, because my voice changes tone to a ridiculous amount depending on far too many factors, so even an hour later when I start to re-record something, I don’t sound the same at all and patching that in has a distinct audible jump – it’s not ‘seamless,’ in other words. Even when I’m listening to exactly what I sounded like, I have trouble reproducing this.

You can open as many Audacity windows as you like, so you can re-record while still having your original open. Then pick the whole sentence that you need to correct (or a significant portion thereof,) start recording with that silence at the beginning, and re-record that sentence several times over, making whatever subtle changes you like. Because your original was buried in the ‘flow’ of what you were saying, even an entire sentence re-recorded may have a different cadence or timbre to it, and I know I’m bad about exhaling right at the end of the re-recorded sentence and this changes the tone at the end.

Unfortunately, now you’ll have to redo all of those compression and equalizing things above for this new recording just to have it sounding almost exactly alike, but seriously, once you’re used to it the whole process takes a minute or so. Then pick the best version of what you re-recorded, and simply cut and paste it over the offending section of the original. It’s not uncommon for volume to be different between the two, but this is easy enough to correct. No problem.

A note about audio/video syncing. If you’re doing this on a track where it’s crucial to sync precisely to the video, you’ll likely notice that the re-recorded section is trivially (or not) different in length than the original, which will make syncing less precise. If you watch and see the waveform jump in either direction just as you paste, you can potentially cut it or add silence to it to bring things close – usually, a fraction of a second isn’t noticeable. The worst part is if you have to significantly increase the length of what you’re correcting, in which case you may be playing around with lengthening the video with a freeze frame or something to fill in the gap.

Once you have the audio track down to where you want it, if needed, copy this over to the separate, blank stereo track by selecting the entirety of both – they should then match up perfectly. Test it of course, then right-click in that blank portion to the left to select Make Stereo Track.

You can now export the audio in the way you need it – an MP3 or whatever for podcasts or uploads, or a lossless format to add to the video edit, whatever. Unless you want…

Music. I usually do a quick opening and closing music clip to video, at least, and a lot of people recommend/insist upon background music throughout the entire recording, though I consider this extraneous and usually unnecessary, and very often annoying – consider that your listeners may not like the piece you chose and don’t want to keep hearing it. Either way, this is easy to accomplish.

In Audacity, click on Tracks/Add New/Stereo Track – you’ll see it pop in underneath your current one. Then in a separate window, select your music tracks and copy them, then paste these into the new blank stereo track below your original recording – you will likely have to play a bit to line things up the way you want them, by either cutting or Generating Silence as needed on either track. If desired, do your fading in and/or out by selecting the end portions (however long you want the fade to take) and using Effect/Fading/Fade In (or Out – you’ll figure it out.) I like having the fade out cross over into the beginning of the vocals, as shown below.

Now, you’ll notice that the original, top tracks start with something in there, a waveform right off the bat, and the inserted music intro comes in after that in the gap of silence. That beginning is the hum I produced as I started the recording, to sync to the video track I was recording at the same time, and it remains in place now so the audio and video can be synced, whereupon both will be cut out of the final project and the audio will start with the music (and the video with my title card fading in simultaneously.)

Also, it’s usually not a bad idea to keep the music down in volume so you don’t assault your listeners, and you can easily adjust that now.

Syncing audio and video. I won’t go into too much detail here since your video editor may be different and this is long enough. With most non-linear video editing suites, however, you can line up things together fairly easily. For simple voiceovers, I’ll cut the audio tracks as needed and simply slide them along the timeline to line them up – these have been recorded after I already edited the video tracks to what I wanted. Volume levels may require adjustments depending on what was captured with the video and whether you want it or not – I have changed them in specific places at times to highlight the sounds a species might make. While I endeavor not to have too much “dead air,” the silences where I’m not talking and little else is coming through the video, sometimes this works as ambience because it really was this quiet, and you don’t need to hear me blathering nonstop.

For syncing to a screen capture, however, that opening noise is used to sync to the visual start of the recording, and then both audio and video are ‘grouped together,’ so that all edits are done to both simultaneously – this is where your flubs and “ums” can get cut, while everything before and after still matches together. This may produce jumps in the video, which is why I say to be aware of what you were doing onscreen when you misspoke, so you can back up and hopefully have a near-perfect match in video.

In this example, actually from the previous Tip Jar, you can see that I instead imported the opening musical introduction separately – that’s the A2 track at bottom, synced with the faded in/out still of the title card, track V2. Then the voiceover track (A1) is aligned with the screen capture track (V1,) where both can be edited together and still maintain perfect (ahem) synchronization.

There you have it. This is by no means ‘complete,’ but it holds most of the tips that I can provide towards getting better audio recordings for your purposes, and you can now see why I balked at tackling this in video. Hopefully, this provides something (preferably a lot) that you can use, and will improve your own recordings notably. Good luck!