Recording, Editing and Processing Voice-Overs with Free Software (Patreon)

In this video you'll learn how to record, edit, process and export speech suing Ardour 6, Noise Repellent and LSP plug-ins. I work under Manjaro Linux.

Below are my notes, which could act as a script for the video.  You can also download the full Ardour session to play with it (licensed under CC-0 1.0 International license).

Intro

Recently I've been recording some voice overs and thought it's a good opportunity to share my workflow and all the tricks I've learned.

We'll record, edit and process a voice over using Ardour 6 and open-source LV2 plug-ins, all under Manjaro Linux. Let's go!

[ intro ]

Before Recording

1. Make sure your room is as quiet as possible. Close any doors, windows. Turn off equipment you don't need to run, warn people at home that you need them to be quiet for a while etc.

2. Treat your room so it has minimum sound reverbation if you can - sometimes recording with your back against an open wardrobe is a great idea, as the soft fabrics will absorb any reflections - cutting down captured reverb.

3. Use a directional microphone - most mics have a cardioid pattern, that mean they will pick stuff from the sides less, and almost non from the back. Use this knowledge to minimize captured background noise - for example aim the back of the mic at your noisy computer.

4. Get as close to the mic as possible without touching it so your signal to noise ratio is as good as it can be. Another thing it'll do is accentuate the low frequencies in your voice - giving you a deeper tone, warmer sound - this is known as the proximity effect. We can boost lows with an EQ later, but it's always better to get the most done at source.

5. Use a pop filter or a windshield to prevent plosives from hitting the microphone capsule

6. If you don't have any plosives protection, put the microphone to your side so that you don't blow into the mic

7. Warm up your voice and prepare yourself water to sip. There's various warm-up routines you can do to improve your articulation and extend your vocal range. If I don't warm up properly, I'm not able to reach these sweet bassy, tones.

8. If you produce a lot of smacking noises, bite your tongue softly to produce more saliva - it should help mitigate these tiny pops and clicks. We can mitigate them in editing, but it's always better to have to do less work in post-production

9. To record you'll need as low audio monitoring latency as you can get. I switch my jack server to 128 or 256 cycles per buffer. I use the commandline program `jack_bufsize` to do that. So in this case `jack_bufsize 128` will do the trick. Sometimes Ardour will crash when you change the buffer size while it's running, so save your work before doing that. Also - use closed-back headphones so they don't bleed back into the microphone - if needed, turn down the monitoring volume. You need to hear if what you're recording sounds well.

10. Se the mic gain so that your loudest voice will not clip it. You'd usually want to have as high preamp gain as possible to minimize extra noise, but clipping loud sounds is not acceptable, so make a noise as loud an anything you'll ever do plus a little bit and set your mic gain so it's just a little shy of clipping.

Let's record!

Depending on how much text you have I think it's best to first read your text out loud once to get comfortable with it, so that you don't make as many mistakes. Maybe you want to read one paragraph as practice, capture it and once you're sure you have it done, you can move on to the next. Sometiems you can rehearse and record all the text in one go.

If you make a mistake - back off and start again from the same sentence, trying to start with the same intonation as you had before. Once you learn your voice and learn to perform the same exact intonation with it, you can back off a few words or even half a word, and still have enough room to make a clean cut later and no one will hear the edit. But for starters, give yourself a bit more material to work with. I think the best workflow is to separate recording and editing. Focus on the performance and on making sure whatever you're recording is going to work. Then move on to the next part until you're done.

I'm going to record a bit of text now.

Ok, the recording is done, let's edit the recording to remove any mistakes. A very useful thing in Ardour is switching to the Ripple Edit Mode. It saves me a lot of work just going and grabbing regions from later in the timeline.

Procesing

Now that we have our voice over edited, it's time to process it.

First thing we can do is use subtle noise reduction - the open-source plug-in that does that is called Noise Repellent.

Now let's add a noise gate to mitigate the background noise in the pauses, and it'll also help us cut down the room reverb tail.

Another thing I'll do is add an EQ to boost the lows a little, cut the subs (we don't want plosives). I also often boost the highs a bit and then search for some resonances in lower mids.

After that, let's add a multi-band compressor to make the sound fuller - we can also control the sybilants with it, making the high band reduce louder sounds above 4-5 kHz.

After that, let' add a subtle compressor to glue it all together and even out our performance a little bit.

And in the end (optionally, if you still have issues with sybilants) - a De-Esser, to make sure these hisses are not going to hurt anybody.

One final thing we should do is to add a master limiter and set the final output level. As a voice over will usually not be heard on it's own, we will probably want to leave a lot of headroom for mixing this later with some other audio.

Now let's set our export markers and export to 24-bit FLAC.