YT YouTube2Text

YouTube2Text → Transcripts → How to make your voice sound better in Audacity 4.0

How to make your voice sound better in Audacity 4.0 — Transcript

by Master Editor · 3,342 words · 505 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00In this video, we are going to look at a
0:02five-step process to make your audio
0:04sound better in Audacity 4. The five
0:06steps I will show now can be used for
0:08professional voice over, audiobook
0:10narration, content creation, podcasts,
0:13or almost any spoken word recording.
0:16This is going to be a detailed video. I
0:18will walk you through all five steps so
0:20you can follow along and understand what
0:22each step is doing. Also, if you want a
0:24written guide with screenshots showing
0:26all five steps, you can get my sound
0:28better in Audacity guide. I will put the
0:31link in the description. Once you
0:32purchase the guide, it will be sent to
0:34your email. Now, let's get back to the
0:37process of making the audio sound
0:39better. The recording you see on the
0:41screen is a raw, unprocessed recording.
0:44That means no audio effects have been
0:45applied to it. Let's listen to a little
0:47bit of the audio. I'll click the play
0:49button.
0:52Some people claim that all air raators
0:54are created equal. Well, we beg to
0:57differ. Ranch works pasture air raators
0:59are engineered with full. The first
1:01thing I notice is that the audio is not
1:03very loud. That is expected because this
1:05is the original recording. Through the
1:08processing, we will make the audio
1:09louder and improve the overall sound.
1:12One tool that can help us determine
1:13whether we are moving in the right
1:15direction is ACX check. If I go to
1:18analyze and then ACX check, I can see
1:21the current measurements of the audio.
1:23ACX check gives us three important
1:25readings. Peak, RMS, and noise floor.
1:30ACX check is an external tool that you
1:32can install in Audacity for free. I will
1:35include a video in the description
1:37showing you how to install ACX check.
1:40For now, think of ACX check as a
1:43measurement tool that gives us different
1:45technical parameters of the audio. As we
1:47process the recording, we can continue
1:49checking these measurements to make sure
1:50we are moving in the right direction.
1:53For now, let's first look at the peak
1:55level. The current peak is -2.78
1:58dB. That is a little low. So, let's go
2:01through the processing steps and see how
2:04these measurements change. The first
2:06step of our audio editing process is
2:08normalize. First select the entire
2:11recording by double clicking inside the
2:13track. Then go to effect volume and
2:15compression and normalize. Inside the
2:18normalize effect, you need to set the
2:20normalize peak amplitude value. So what
2:23exactly does peak amplitude mean? Peak
2:25amplitude means the loudest point of the
2:27audio. The maximum possible value is 0
2:30dB. You can also see 0 dB on the meter.
2:330 dB is the maximum level that the peak
2:36can reach. When we change the peak
2:38level, the other parts of the waveform
2:40are adjusted accordingly. So what should
2:42we use as the ideal peak amplitude?
2:45Normally I recommend minus3 dB. This is
2:48also in accordance with the ACX audio
2:50requirements for audiobooks. If we look
2:52at the ACX audio requirements, you can
2:54see that the peak level needs to be
2:55below minus3 dB. These requirements are
2:58specifically designed for audiobook
3:00narration, but they are also a very
3:02useful technical guide for other types
3:04of spoken word recordings. Even if you
3:07are not producing an audio book, you can
3:09use the ACX requirements as a kind of
3:11north star to help keep your audio
3:13processing moving in the right
3:14direction. That is why we are going to
3:16normalize this recording to minus 3 dB.
3:19Set the value to minus3 dB. Leave the
3:22other settings as they are and click
3:24apply. As soon as I click apply, you can
3:27see that the waveform increases in
3:29height. That tells us the recording has
3:31become louder. Let's listen.
3:33Some people claim that all air raiders
3:36are created equal. Well, we beg to
3:38differ.
3:39>> You can definitely hear that the audio
3:41has become louder. However, increasing
3:43the level also increases other parts of
3:45the recording, including the noise. We
3:47will take care of the noise in the next
3:49step. Before that, let's check the audio
3:51again with ACX check.
3:55This time, the peak level is at minus 3
3:57dB, which is passing. However, the RMS
4:00level is minus 23.18 dB. According to
4:03the ACX requirements, the RMS should be
4:06between minus23 and minus 18 dB. So, at
4:11the moment, the RMS is failing. We will
4:13fix that later. RMS represents the
4:16average loudness of the audio. Keeping
4:18the RMS within the correct range is
4:20important because it helps provide a
4:22consistent listening experience. For
4:24this particular recording, the noise
4:25floor is already passing even after
4:27normalizing the audio to minus 3 dB.
4:30That will not necessarily happen with
4:32every recording. So, our second step is
4:35noise reduction. Let's see how we can
4:37reduce the noise and how that affects
4:38the overall audio. Noise reduction is a
4:41two-step process. You also need to
4:43understand that Audacity can remove
4:45static noise or continuous hissing
4:47noise. If you have irregular background
4:49noises such as a dog barking, other
4:51people talking, a car horn, a lawn
4:54mower, or similar unpredictable sounds,
4:57those cannot be removed using this type
4:58of noise reduction. This is something
5:01that beginners in audio recording and
5:03editing sometimes misunderstand.
5:05Noise reduction does not mean that every
5:07type of noise can automatically be
5:09removed. It works best with continuous
5:11static or hissing type background noise.
5:14To begin the noise reduction process,
5:16first select a noiseonly section of the
5:18recording. At the beginning of this
5:20recording, I have a small section where
5:21there is no talking. There is only
5:23background noise or room tone. I'll drag
5:25over that section to select it. Then
5:27I'll go to effect, noise removal and
5:29repair and noise reduction.
5:32As I mentioned, noise reduction is a
5:34two-step process. Step one is to give
5:36Audacity a noise profile. You can see
5:39the get noise profile button here. Using
5:41this noise profile, Audacity learns what
5:44kind of noise exists in the recording.
5:46It can then use that sample to identify
5:48and reduce similar noise throughout the
5:49audio. I'll click get noise profile.
5:52Audacity has now analyzed the noise
5:54sample. The next important step is
5:56selecting the entire recording before
5:58applying the actual noise reduction. At
6:02the moment, only the small noise only
6:03section at the beginning is selected. We
6:05don't want to reduce noise only from
6:07that section. We want to reduce the
6:09noise from the entire recording. So I'll
6:12double click inside the track to select
6:14everything. Then I'll go back to effect,
6:16noise removal and repair and noise
6:19reduction. Now we can complete step two.
6:22In step two, you can see three main
6:24settings. Noise reduction, sensitivity,
6:27and frequency smoothing. The noise
6:29reduction setting determines how much
6:31noise we want to reduce. When we checked
6:33this recording earlier, the noise floor
6:35was already around -60 dB. So, we do not
6:39need to remove a huge amount of noise
6:41from this recording. Ideally, if you
6:43have a good recording environment, a
6:44noise reduction value of around 6 dB can
6:47work well. If necessary, you can
6:49sometimes go as far as 12 dB. The
6:52important thing to remember is that
6:53noise reduction can introduce harshness
6:55or artifacts into the audio. The more
6:58noise you try to remove, the greater the
7:00chance that the recording will begin to
7:01sound unnatural or robotic. For example,
7:04Audacity allows you to set the noise
7:06reduction value as high as 48 dB. But if
7:09you try to remove that much noise from a
7:11recording, it can sound very strange and
7:13robotic. So I recommend that you do not
7:15go beyond around 12 dB. Ideally, if 6 dB
7:19is enough, use 6 dB. For demonstration
7:22purposes, I am going to use 12 dB for
7:24this recording. Next, we have
7:26sensitivity and frequency smoothing.
7:28Sensitivity controls how aggressively
7:30the noise reduction algorithm identifies
7:32noise. For voice recordings, I normally
7:34recommend setting sensitivity to six.
7:36Making it much more aggressive by
7:38increasing the value does not
7:39necessarily produce a better result.
7:42Reducing it too much may not give you a
7:43good result either. So, a sensitivity
7:46setting of six is a good starting point.
7:48I also recommend setting frequency
7:50smoothing bands to six for spoken word
7:52recordings. For music recordings, you
7:54may want to use a lower frequency
7:56smoothing value, but for spoken word
7:58audio, six usually works well. Once
8:01these settings are configured, you can
8:02click apply. You can also use the
8:05preview button before applying the
8:07effect. This allows you to hear how the
8:09audio will sound with the selected
8:10settings. Let's preview it.
8:15>> Some people claim that all air raiders
8:17are created equal. Well, we beg to
8:20differ. Ranch works past. If you are
8:22listening in a quiet environment, you
8:24should notice that the hissing noise has
8:26been reduced. If you are listening in a
8:28noisy environment, this small amount of
8:30background noise may blend in with the
8:32noise around you, so the difference may
8:34be harder to hear. Now I'll apply the
8:36noise reduction. Let's measure the audio
8:39again using ACX check. After reducing
8:43the noise, which was our second step,
8:44you can see that the noise floor is now
8:46around -72 dB. Previously it was around
8:50minus60 dB. It has dropped by
8:53approximately 12 dB because we applied
8:5512 dB of noise reduction. You may also
8:58notice that the peak and RMS readings
9:00have changed slightly. That is normal.
9:02Reducing noise changes the audio data
9:04slightly. So the peak and RMS values may
9:07also move a little. That is not a
9:09problem at this stage. Now let's move to
9:11step number three which is EQ or
9:14equalization. Equalization is a very
9:16large topic. For this video, I'm going
9:18to keep the EQ process basic. To open
9:21the EQ, go to effect, EQ, and filters,
9:24and then filter curve EQ.
9:27If I choose the default preset, you can
9:29see a flat line at 0 dB. At the moment,
9:32this EQ is doing nothing to the audio.
9:34At the bottom, you can see the different
9:36frequencies of the audio, covering
9:38approximately 20 hertz up to 20 kHz.
9:41Using filter curve EQ, you can draw a
9:43curve to change different parts of the
9:45frequency spectrum. So what exactly is
9:48EQ? EQ basically means adjusting the
9:51volume of different frequencies by
9:53different amounts. For example, if I
9:55drag this entire line upward, I am
9:57increasing all of the frequencies by
9:59approximately the same amount. If I drag
10:01the entire line downward, I am reducing
10:04all of the frequencies by approximately
10:06the same amount. But if we adjust every
10:08frequency by exactly the same amount,
10:10that is basically just changing the
10:12overall volume. The real purpose of EQ
10:14is to identify which frequency areas
10:16need to be increased or decreased and
10:18then adjust those frequencies
10:20accordingly. As I said, EQ is a very
10:23large topic. So instead of going into
10:25advanced EQ in this video, I am going to
10:28show you one basic EQ adjustment that
10:30can be useful for almost every voice
10:32recording. That is a highpass filter. To
10:36create this, go to factory presets and
10:38choose low rolloff for speech. That
10:40gives us a useful starting point. When
10:43you select low rolloff for speech, you
10:45can see that the curve starts reducing
10:47frequencies in the lower frequency
10:48range. The preset begins rolling off
10:51frequencies around 100 hertz. For this
10:54recording, I am going to adjust it so
10:56that the cutff is around 80 Hz. I'll
10:58drag this point so the rolloff starts
11:00around 80 Hz. Why are we using 80 Hz?
11:03Usually, the human voice does not
11:05generate useful vocal frequencies below
11:07approximately 80 Hz. So if there are
11:10frequencies below that area in your
11:12recording, they may mainly consist of
11:14unwanted low-frequency noise. By using a
11:16highpass filter, we can reduce those
11:18unnecessary low frequencies. Why is it
11:21called a highpass filter? Because it
11:23reduces or blocks the lower frequencies
11:25while allowing the higher frequencies
11:27above the cutff point to pass through.
11:29That is the basic idea of a highpass
11:31filter. Again, EQ can become much more
11:33detailed than this, but for this
11:35five-step process, we are keeping the EQ
11:38simple. For most spoken word recordings,
11:40using this type of highpass filter is a
11:42good starting point. Once again, you can
11:45use the preview button to hear how the
11:46recording sounds before applying the
11:48effect. I'll apply it now. The tonal
11:50change is not necessarily going to sound
11:52dramatic. Let's listen.
11:56Ranch works pasture air raators are
11:58engineered with forged ranchte tech
12:00blades designed to penet you may not
12:02hear a major difference. However, when
12:04we check the recording using ACX check,
12:07we can see what happened to the
12:08technical measurements. You can see that
12:10the noise floor has been reduced
12:12slightly. As I mentioned earlier, there
12:14is usually no useful voice information
12:16in those very low frequencies, but there
12:18can be noise. By reducing those
12:20frequencies, we have also reduced a
12:22little more noise. The peak level is
12:24passing, but the RMS level is still
12:26failing. So now our main concern is
12:29fixing the RMS. The next major effect in
12:31our five-step process is the compressor.
12:34A compressor can help us increase the
12:35RMS level. At this point, our peak is
12:38around minus 3.32 dB. So technically, we
12:41still have a little bit of room to
12:42increase the overall level. I could
12:44normalize the recording to -3 dB again,
12:47but simply normalizing it will not
12:48properly fix the RMS problem. Let me
12:51show you. I'll go to effect volume and
12:53compression and normalize. I'll
12:55normalize the recording to minus 3 dB
12:57again. Now let's go back to ACX check.
13:03You can see that the audio is still
13:05failing on RMS. So we need to fix the
13:08RMS using the compressor effect. As I
13:11mentioned earlier, RMS represents the
13:13average loudness of the recording. A
13:15compressor can help control the dynamics
13:17and increase the average loudness of the
13:19audio. Let's see how to do that. Go to
13:22effect volume and compression and
13:24compressor. There are many different
13:26compressor presets and settings. First,
13:29I'll go back to the default settings.
13:31You can see that there are several
13:32controls available. However, an easier
13:35way to get started is to use the lead
13:37vocals factory preset. Choose lead
13:39vocals from the factory presets. Lead
13:42vocals is a useful starting point
13:44because singing vocals and spoken voice
13:45share some similar characteristics.
13:48Using this preset allows us to avoid
13:50manually configuring every single
13:52parameter. For this demonstration, we
13:54are not going to worry about attack,
13:55release, and look ahead. The preset
13:58already gives us useful starting values
14:00for those controls. We will mainly focus
14:02on threshold and ratio. First, let's
14:05talk about threshold. You can see a
14:06curve here. The threshold determines the
14:09point at which compression begins. When
14:11the audio level crosses the threshold,
14:13the compressor becomes active. You can
14:16see that the line begins to bend at this
14:17point. If I change the threshold, you
14:19can see that the bending point moves.
14:22The ratio determines how strongly the
14:23audio above the threshold is compressed.
14:26If I change the ratio, the threshold
14:28point remains in the same place, but the
14:30slope of the line changes. That is the
14:32basic relationship between threshold and
14:34ratio. Threshold determines when
14:37compression begins. Ratio determines how
14:39much compression is applied after the
14:41signal crosses that threshold. For this
14:43type of recording, I normally use a
14:45threshold somewhere between -12 and -8
14:48dB as a starting point. The exact
14:50threshold depends on how much RMS
14:53increase you need. For this recording,
14:55we saw an ACX check that the RMS is
14:58already very close to passing. So, we do
15:01not need a huge increase. Because of
15:03that, we can keep the threshold
15:04relatively close to -12 dB. For example,
15:08we can use around -14 dB. I'll set the
15:11threshold to approximately minus14 dB.
15:14If we needed a much bigger increase in
15:16RMS, we might use a lower threshold such
15:18as -8 or even -20 dB. But for this
15:22recording, we do not need that much
15:24compression. So around -14 dB should
15:27work. Sometimes you will need to do a
15:29little trial and error to find the best
15:31threshold for a particular recording.
15:33Now let's look at the ratio. A useful
15:36ratio is normally somewhere between 2:1
15:38and 4:1. If you need stronger
15:40compression and a greater increase in
15:42average loudness, you might use a ratio
15:44closer to 4:1. If you only need a small
15:46amount of compression, you can use a
15:48ratio closer to 2:1. For this recording,
15:50we only need a slight increase in RMS.
15:54So, I'll use a ratio of around 2.5 to 1.
15:57Once again, you can click preview to
15:59hear how the compressor is affecting the
16:01recording. If you are happy with the
16:03settings, click apply. I'll click apply.
16:08The compressor has now been applied.
16:10Let's listen to the audio.
16:13>> Some people claim that all air raators
16:15are created equal. Well, we beg to
16:17differ.
16:18>> You may notice that the audio has
16:20actually become quieter. That may sound
16:22strange because I just told you that the
16:24compressor can help increase the RMS or
16:26average loudness. So, let's understand
16:28why this happens.
16:31If I go to ACX check now, you can see
16:33that the RMS has not yet increased the
16:36way we want. You can also see that the
16:38peak has dropped significantly. Before
16:40compression, the peak was around minus3
16:42dB. After compression, the peak is much
16:45lower. This happens because the
16:47compressor reduces the louder peaks of
16:48the recording. So, after applying
16:51compression, we need to normalize the
16:52recording again to minus 3 dB. That
16:55final normalization allows us to take
16:57advantage of the reduced dynamic range
16:59created by the compressor. Then we can
17:01see the true effect on the RMS.
17:04So let's do our final step. Go to effect
17:07volume and compression and normalize.
17:10Set the peak value to minus 3 dB again
17:12and apply it. You can immediately see
17:15that the waveform has become louder
17:16again. Now let's go back to ACX check.
17:20This time you can see that the peak RMS
17:22and noise floor are all passing. The
17:25peak is around minus3 dB and the RMS has
17:29increased to around minus20 dB. So this
17:32demonstrates one of the main reasons we
17:34use a compressor. If you no longer have
17:36enough peak headroom to simply increase
17:38the overall volume, but you still need
17:40to increase the RMS or average loudness,
17:43compression can help you do that. The
17:45compressor reduces the louder peaks and
17:47controls the dynamic range. Then after
17:50normalizing the recording again, the
17:52average loudness can become higher
17:54without pushing the peak above the
17:56target level. And that completes our
17:58five-step process for making spoken word
18:00audio sound better. The five steps are
18:02normalize, noise reduction, EQ,
18:06compression, and normalize. Again, keep
18:08in mind that audio settings sometimes
18:10need to be adjusted depending on the
18:12recording. This is especially true for
18:14the compressor settings. The ideal
18:16threshold and ratio can change depending
18:19on the original recording and how much
18:21RMS adjustment you need, so you may need
18:24to experiment a little. However, the
18:25five-step process and the settings I
18:27have demonstrated in this video give you
18:29a very good starting point. You can
18:32apply this process to your recording,
18:34check the results, and make adjustments
18:36as necessary. If you want this complete
18:38five-step process as a written Audacity
18:41guide, check the link in the description
18:43and the pinned comment. And if you are
18:45looking for a complete step-by-step
18:47Audacity course for beginners, you can
18:49also check out my Audacity course. This
18:52course is designed for people who want
18:53to work professionally with voice over,
18:55audiobook narration, podcasts, or spoken
18:58word audio and want to produce better
19:00quality recordings in Audacity. You will
19:02find all the links in the description
19:04and the pinned comment. Click the links
19:06to learn more.

About this transcript

This page contains the full transcript of How to make your voice sound better in Audacity 4.0 by Master Editor, generated from the public captions YouTube serves with the video. The transcript has 3,342 words across 505 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.