How do you record your voice in GarageBand so it sounds good?
Input device, a mono audio track, Mic Mode, recording level, voice patches and the Noise Gate in GarageBand on a Mac, in the order the sound travels, plus what to do when the mic itself is the problem.
You open GarageBand, make a track, press R and read the first paragraph of your podcast intro. Playback is a letdown. The voice sounds thin and a little far away, a soft hiss sits under every pause, and somewhere in the background the fridge hums along. GarageBand did nothing wrong. It recorded exactly what reached it, and most of what decides how a GarageBand vocal sounds happens in the half minute before you press R.
So this guide runs in the order the sound travels: which microphone GarageBand listens to, what kind of track receives it, how loud it arrives, and only then the patches and the Noise Gate that people usually reach for first. If you are still choosing an app and are not sure GarageBand is the right one, the comparison of voice recording software for the Mac is the better page for you. If you only need a quick voice memo and not a project, recording audio on a Mac covers QuickTime and Voice Memos, which take less setup.
What GarageBand is, and what it is not
GarageBand is Apple's free music and recording app. As of October 2026 the Mac version on the App Store is 10.4.14, and it requires macOS 15.6 or later. There are versions for Mac, iPhone and iPad. There is no GarageBand for Windows, and the copies that turn up on download sites under that name are not from Apple. On a Windows PC the closest free equivalent for voice work is Audacity, which runs on Windows, macOS and Linux.
GarageBand was built for music, which matters for voice in two ways. The good part: it has proper tracks, takes, plug-ins and a mixer, so a podcast or voice-over gets the same tools a song does. The awkward part: a lot of its defaults assume a band. The control bar carries a Count-in button and a Metronome button, and the ruler counts bars and beats. For speech you switch both buttons off and ignore the bars, and nothing is lost.
Step 1: choose the GarageBand microphone input
Choose GarageBand > Settings (Command-Comma), then Audio/MIDI. Two menus matter here:
- Input Device: the microphone GarageBand records from. Any Core Audio device shows up: the MacBook's built-in mic, a USB mic, a headset, AirPods, an audio interface.
- Output Device: where you listen. Pick your headphones.
Check this every time you plug in a different mic. Plenty of "GarageBand recorded nothing" and "why does my expensive mic sound like a laptop" stories end in this menu, with the built-in microphone still selected while the USB mic sat unused on the desk.
AirPods and other Bluetooth headsets will appear here too. They work, but Bluetooth limits how much detail a mic can send, so for anything you plan to publish a wired or USB mic gives you more to work with.
Step 2: record your voice on a mono audio track
Start a new project, choose Empty Project, and in the track type dialog pick the one with the microphone icon, labelled Record using a microphone or line input. That is an audio track, the only kind that records a real voice. Software Instrument tracks record notes from a keyboard, not sound.
Now press B to open Smart Controls and look at Recording Settings:
- The Input button switches between a mono input and a stereo pair. One voice into one microphone is a mono source. Recording it as stereo doubles the file and adds nothing; on a two-channel interface it can also mean one side is your mic and the other side is an empty input.
- The Input pop-up picks the channel. For a USB mic or a single-input interface that is usually Input 1.
If you are recording two people, give each mic its own track. Two USB mics at once need an extra step: macOS has to combine them into an Aggregate Device, and Apple describes the setup for GarageBand on that page.
Step 3: set Mic Mode to Standard
This step is easy to miss on recent macOS versions. With an audio app open, a Mic Modes icon can appear in the menu bar. Apple's own note on Mic Mode when recording in Logic Pro or GarageBand says the settings other than Standard are designed for background noise on calls and change the sound of the input, and recommends Standard when recording for creative purposes. If the icon is not there, Mic Modes does not apply to that app and you can skip this step.
Step 4: set the recording level, the step that decides everything
In the same Recording Settings area sits the Record Level slider. If it is dimmed, GarageBand cannot change the input volume of that device, and you set the level on the device itself: the gain knob on a USB mic, or the gain control on an interface.
Set it like this. Read the loudest line of your script at the volume you will really use, not a polite test mumble, and watch the track meter. You want the peaks well into the meter but clear of red. Red means the signal ran out of room and the top of the waveform got cut flat, and that kind of distortion stays in the file: no patch removes it later. Too low is the opposite trap: you will raise it afterwards, and the hiss of the mic and the room comes up by exactly the same amount.
There is also an Automatic Level Control checkbox. Apple describes it as lowering the level to prevent feedback and raising it when the input is too low. For a song that can be handy. For speech, leave it off: a level that rises when you go quiet also lifts the room tone in every pause, and you hear the background breathe in and out between sentences.
A starting chain for one voice that works on most setups:
Input Device your mic (not "MacBook Pro Microphone" unless you mean it)
Track type Audio, "Record using a microphone or line input"
Input mono, Input 1
Mic Mode Standard
Record Level loudest line peaks around -12 dB, never red
Auto Level Control off for speech
Monitoring on only with headphones
Mouth to mic a hand-span, roughly 10 to 15 cm, slightly off to the side
Bit depth 24-bit (Settings > Advanced), set before the first take
The distance line does more than any plug-in in this article. Every time you double the distance to the mic, the room gets a bigger share of what the mic hears. A hand-span away, your voice dominates; at arm's length, the walls and the fridge get a vote. Aiming slightly past the side of the mic keeps the bursts of air from "p" and "b" out of the capsule. If the level is the thing you keep fighting, the article on gain staging goes deeper into where the gain should live.
About the 24-bit line: GarageBand's Advanced settings let you choose the bit depth for new recordings. Apple calls 24-bit the most commonly used, notes that 32-bit float makes larger files, and warns not to change the setting after you have already recorded audio in the project. Set it once, before the first take.
Step 5: hear yourself without feedback
Each audio track has a Monitoring button in its header, and the same button sits in Recording Settings. With it on, GarageBand plays your mic back to you as you speak. Apple's guide is blunt about the risk: with monitoring on, the mic can pick up your speakers and you get feedback. Use headphones, and turn monitoring off when you are not recording.
For speech you often do not need monitoring at all. When you record vocals in GarageBand for a song, it is different: singers need it to hear pitch against a backing track; a narrator does fine without it and avoids the slight, distracting lag that any processing chain adds to what you hear.
Step 6: pick a voice patch, then open it and look
The Library on the left shows patches for the selected track. For an audio track, patches can include default effects settings, and the Voice category holds vocal patches such as Narration Vocal, Natural Vocal and Bright Vocal. A patch can carry up to four built-in effect plug-ins, and you see them listed in the Plug-ins area of Smart Controls.
Do not pick a patch by name and move on. Pick one, then open the Plug-ins area and see what it added: compression, EQ, reverb, something else. For spoken word, start from Narration Vocal or Natural Vocal, and if there is any reverb in the chain, switch it off. Your room already adds reverb, usually more than you want, and a second layer on top makes a voice sound further away, not richer.
One useful property of how GarageBand handles this: the plug-ins sit on the track, not inside the recorded file. Record a take with one patch, then click a different patch, and the same take plays back through different processing. That means you can record first and decide on the sound later, and for spoken word that is the better order.
Step 7: the GarageBand noise gate, and why it comes last
Selecting the Noise Gate checkbox on an audio track loads GarageBand's Noise Gate plug-in, and a slider sets its threshold. The mechanism is simple. Anything quieter than the threshold is muted; anything louder passes through untouched.
That is also its limit, and the reason behind many "the GarageBand noise gate doesn't work" complaints:
| What you hear | What the gate does |
|---|---|
| Hiss or fan noise in pauses | Mutes it, if the threshold sits above the noise |
| The same hiss under your words | Nothing, it passes with your voice |
| Quiet word endings ("-ts", "-th", soft consonants) | Can cut them off if the threshold is too high |
| A breath before a sentence | Usually muted, which can make speech sound clipped |
| Background that appears with every word and vanishes between them | Causes it: on a noisy recording the gate opens on every word and lets the noise through with it |
So use it gently. Start with the threshold low, play back a stretch with pauses, and raise it until the pauses go quiet. Then listen to the ends of words. The moment the last consonant of a sentence starts sounding chopped, back off a notch. On a quiet recording you may not need the gate at all, and that is the best case. The same mechanism appears in streaming software, which is why the guide to noise gate settings in OBS reads almost the same.
If the noise under your words is the problem, a gate cannot help. That is a job for noise reduction after the fact, covered in removing background noise from a recording, or better, for fixing the input before it reaches GarageBand.
How to record a podcast in GarageBand
Everything above applies, plus a few habits that make editing much faster:
- One track per voice. Two people on one mic means you can never fix one without the other. If you have two mics, record two tracks.
- Takes instead of restarts. Drag across the top of the ruler to make a cycle area (a yellow strip). Record, and from the second pass GarageBand collects each pass into a take folder. You pick the best take afterwards instead of starting over.
- No click. Turn off the metronome and the count-in before the first take. A click in your headphones and a bar or two of lead-in before every take only get in the way of speech.
- Ten seconds of silence. Before you start talking, record ten seconds of the room. It tells you honestly how much hiss and hum you are dealing with, and it gives a noise reduction tool a clean sample later.
- Export. Choose Share > Export Song to Disk and pick AAC, MP3, AIFF or WAVE. Keep a WAVE or AIFF master and make the MP3 from it, so a later re-edit does not start from an already compressed file.
GarageBand has no built-in way to bring in a remote guest from a call. For interviews over the internet, and for tools built around multitrack remote recording, see podcast recording software for the Mac.
Why a clean input beats a clever patch
Look at the order of the steps again. Every tool GarageBand gives you works on what already arrived: a patch shapes it, a compressor evens it out, a gate mutes the quiet parts. None of them can put back detail a laptop mic never captured, and none of them can separate your voice from the room once both are mixed into the same waveform. That is why the cheapest real improvements come before the app: the right input, the right distance, a softer room, a level that stays out of the red.
Some setups still sound wrong after all that. A built-in mic or a cheap headset in an untreated room, which is what most people record with, leaves too little for patches to work with. This is where a processed input can help, and TunedMic is one way to get it on a Mac.
TunedMic adds a virtual microphone called TunedMic to the Mac. Your real mic stays as it is; TunedMic processes its sound on the Mac in real time, and any app that lets you pick a mic can pick it, GarageBand's Input Device menu included. Setup is a one-minute Sound Check: you read a short passage aloud while TunedMic takes more than 30 measurements of your voice, your mic and your room, and it builds nine profiles from that. The profiles come in three families: Creator is meant for podcasts, streams and videos, Narrator for calls and lessons, Vocal for singing and practice. Noise, room echo, clarity, loudness and a 12-band EQ are there if you want to adjust them, and you never have to.
Its limits, stated plainly, because they decide whether it fits:
- The processing is baked in. GarageBand records what TunedMic sends. The plug-in trick from Step 6 still works on top, but you no longer have a raw take underneath. If you mix carefully in GarageBand and want full control later, record the real mic.
- There is a delay, less than a tenth of a second. On speech that does not matter. If you sing or play against a backing track, record a few claps on a test take and check that they land on the beat before a real session.
- It is Mac only: Apple silicon (M1 or newer) and macOS 14 or later. No Intel Macs, no Windows. Note that current GarageBand asks for macOS 15.6 anyway.
- It does not turn a cheap mic into a specific expensive one, and it does not remove every noise.
Where it fits well: you record spoken word on the mic you already have, in a normal room, and you also want the same voice in Zoom, Telegram or Discord without setting up processing separately in each app.
Your next ten minutes
Open GarageBand and go straight to Settings > Audio/MIDI: confirm the Input Device is the mic you think it is. Make an Empty Project with a mono audio track, set Mic Mode to Standard, and record two things: ten seconds of silence, then your loudest paragraph at a hand-span from the mic, peaks clear of red. Play it back in headphones. If the silence is quiet and the voice is close, try Narration Vocal and you are done. If the silence hisses and the voice sounds far away even at that distance, the input is the bottleneck: watch the 38-second video on the TunedMic home page to see what setup asks of you, one minute of reading aloud, then use the free 7-day trial (no card, no account) to try it in your own room before deciding. The price, $19 once, is on the pricing page; payments are final, which is what the trial is for.