Why does your mic sound bad on Discord, and how do you fix it?

Your squad is mid-round. You say "he's behind the left crate, I'm one shot, push now", and what reaches the channel is "he's behind the left cr... one shot, push". Someone asks you to repeat it. After the match a friend says, kindly, that you sound like a walkie-talkie in a bathroom. Nothing is broken. Discord is doing exactly what its settings tell it to do with the signal your Mac hands it.

That is the useful way to think about Discord audio quality: your voice passes through three stages before anyone hears it, and each stage fails with its own recognizable sound.

1. Your side, before Discord. The mic, where it sits, the room, the input level in macOS, Bluetooth.
2. Discord's processing on your computer. Input sensitivity, noise suppression, echo cancellation, automatic gain control.
3. The trip to everyone else. The channel's bitrate, which the server sets, and your connection.

You control the first two completely. The third belongs partly to whoever runs the server and partly to your network. Most guides on how to make your mic sound better on Discord pour all three into one list of twenty toggles, so people flip everything and end up sounding worse than before they started.

If your real complaint is that friends hear your keyboard, fan or dog, rather than your voice itself sounding wrong, the page you want is how good Discord noise suppression is, and when it is not enough. This one is about the voice: clipped, thin, robotic, pumping or muffled.

First, name the kind of bad

"My mic sounds bad on Discord" covers at least seven different faults. Each leaves a fingerprint, so start from what people hear and go straight to the matching fix.

One shortcut settles the stage question in two minutes. Join a Zoom, FaceTime or Google Meet call with the same mic and ask the other person how you sound. If you are fine there and bad only in Discord, the cause lives in stage two or three. If you sound bad everywhere, Discord settings will not rescue you, and the fix starts with the mic and the room.

If the main complaint is volume, with people asking you to speak up, read why your mic is so quiet and how to make it louder. Quiet input has its own chain of causes, and it is shorter than this one.

Hear what Discord hears

Discord has a mic test built in. Open User Settings (the cogwheel next to your name), go to Voice & Video, and on the Voice tab press Let's Check. Discord plays your own voice back to you.

Two limits apply. The test runs on your computer, so it shows the effect of sensitivity, noise suppression and gain control, but it says nothing about the channel bitrate or what your connection does to the packets. And you hear yourself through your own headphones at a volume you are used to, which flatters everything. For a test that includes the trip over the network, have a friend record you, or follow the steps in how to test your mic and hear what others hear on a call.

Run the test with headphones on. Through speakers the playback leaks back into the mic, and you end up judging a feedback loop instead of your voice.

Discord mic sensitivity: why the ends of your sentences vanish

Discord has two input modes. Push to Talk sends audio only while you hold a key. Voice Activity opens the mic when the incoming level crosses a threshold and closes it when the level falls back. That threshold is the Input Sensitivity setting, and it produces the most common Discord complaint of all: clipped words.

Speech does not arrive at a steady level. "He's" starts on a breathy h, and the end of a sentence trails off as your breath runs out. If the threshold sits above those quiet parts, Discord opens the gate late and shuts it early. Your friends get the loud middle of each phrase and lose the edges.

By default Discord can Automatically determine input sensitivity. Discord's own support article says this generally works but may misbehave with significant background noise or large swings in how loudly you speak, and it recommends switching it off if your voice jumps between too loud and too soft or gets cut off at the end of sentences. That is the scene from the top of this page, almost word for word.

To set it by hand:

1. In Voice & Video, under Input Mode, select Voice Activity and switch off Automatically determine input sensitivity.
2. Sit the way you normally do, stay silent, and watch the input bar. Note where the room noise peaks: fans, keys, the hum of a desktop PC.
3. Now speak softly, the way you trail off at the end of a sentence, and note where that peaks.
4. Drag the marker between the two, a little nearer the room noise than your soft voice, so the quiet ends of words still clear it. Discord's own guide asks for the same balance: high enough that nothing transmits while you are silent, low enough that soft speech still gets through.

Discord also notes that Voice Activity carries an inherent delay of about 200 milliseconds, plus the distance to the server. In a fast callout, that is a fifth of a second your team hears you late.

If your room is loud enough that no marker position separates your voice from the noise, Voice Activity cannot work well, and Push to Talk is the honest fix. Set the Push to Talk Release Delay long enough that the tail of your last word still goes out after you let go of the key. According to Discord, Push to Talk has practically no delay when you press the key, and with no threshold to cross, a soft first syllable goes out whole too.

The voice processing toggles, one at a time

Discord runs its own processing on your voice before sending it. In Voice & Video you will find Noise Suppression, Echo Cancellation and Automatic Gain Control. Each solves a real problem. Each also changes your voice, and stacking all three on a signal that did not need them is a common source of that over-processed sound.

Noise Suppression offers several modes, and one of them is Krisp, a third-party machine-learning filter built into Discord. Discord says Krisp runs on your device, costs nothing inside Discord, and on a Mac needs an Apple processor and macOS 12 or later. The same FAQ contains a line most guides skip: in a quiet room with a good mic you may notice a decline in voice quality with Krisp on, and Discord encourages going without it there. Krisp also switches itself off when CPU use gets very high (Discord shows a notice when it does), which can explain background noise "suddenly coming back" in the middle of a heavy game.

Echo Cancellation exists for people on speakers. It removes the voices of others that come out of your speakers and back into your mic. On headphones there is little echo to cancel.

Automatic Gain Control raises a quiet signal and lowers a loud one. Paired with a mic set too low in macOS, it keeps hauling the level up, and in the pauses between words it hauls the room noise up with it. People describe the result as breathing or pumping.

Our recommendation for most people on headphones in a reasonably quiet room: switch all three off, run Let's Check, then turn them back on one at a time and keep only the ones that make an audible improvement. With a fan, a desktop PC in the room or people talking next door, Krisp earns its place. On laptop speakers, leave Echo Cancellation on.

Discord bitrate: what it does and who controls it

Discord encodes voice with Opus, an open audio codec. Its developer documentation specifies Opus at 48 kHz with two channels. Bitrate is how many kilobits per second the codec may spend on the sound, and it is set per voice channel by the server, not in your personal settings.

To see it, hover over the voice channel, click the Edit Channel cog, and find the Bitrate slider on the Overview page. Changing it takes permission to edit the channel, so on someone else's server it is a request to an admin, not a setting you own. On an unboosted server the slider runs from 8 to 96 kbps. Server boosts raise the ceiling. As of October 2026, Discord lists these levels:

Now the part that saves money. The people who maintain Opus publish their own guidance: for VoIP, 24 kbps should already give fullband speech, and podcast-style voice is fullband from about 24 kbps in mono or 32 kbps in stereo. Music streaming sits at 64 to 96 kbps. Discord's help pages do not say how its client spends a channel's bitrate, so read this as a direction rather than a promise. The direction is still clear: for one person talking, a channel at 64 kbps or more is not your bottleneck. Boosting to 256 or 384 kbps helps music, instruments and stream audio. It will not turn a thin MacBook mic into a full one.

The slider is still worth a look, because channels do get set low. If it sits near the far left, at 8 or 16 kbps, that alone explains a muffled, telephone-like sound, and moving it up costs nothing. Discord's own caveat: a higher bitrate needs more bandwidth and can affect people on poor connections, a trade-off the server owner makes for everyone in the channel.

When it sounds robotic: the network

A voice that turns metallic, warbles or drops syllables in bursts, then sounds normal again, is usually not a mic problem. That is the sound of packets arriving late or not at all, with the codec trying to paper over the gaps.

Discord's troubleshooting guide lists the steps for this: check the speed and stability of your connection, close heavy apps running in the background, and try the Quality of Service High Packet Priority toggle on the Voice tab. The same guide suggests turning it on for connection issues in one section and turning it off for choppy audio in another, so try both positions and keep whichever behaves better on your network. A cable instead of Wi-Fi removes one more variable.

The network is also the one place where your friends' setup matters. If only one person hears you as a robot, the trouble is more likely on their connection than on yours.

Your side: the Mac and the mic

Every Discord setting works on the signal it receives. If that signal is muffled, clipped or full of room, no toggle restores what the mic never captured.

Pick the mic by name. In Discord's Input Device menu, choose the actual mic rather than Default. On Default, Discord follows whatever macOS currently treats as the input, and AirPods connecting halfway through the evening can quietly switch you to their mic. On the Mac side, open System Settings > Sound > Input and check which device is selected there as well.

Set the level once, in macOS. The same Input pane has an input volume slider. Apple's guide says to lower it for loud sources so the sound does not distort and raise it for someone speaking softly. Aim for your loudest laugh to stay clear of the top of the meter. If you clip here, before Discord ever sees the signal, Discord's Input Volume slider can make the result quieter but cannot take the distortion out. The longer version is in how to change microphone settings on a Mac.

Distance beats settings. A MacBook's built-in mic sits inside the chassis, usually an arm's length from your mouth, and it hears a lot of room. A USB mic on the desk a hand or two from your face captures more voice and less wall at the same input level. If you sound boomy or far away, read why your mic sounds muffled or far away before buying anything.

Bluetooth has a ceiling. AirPods and Bluetooth headsets work in Discord, but Bluetooth limits how much detail a mic can send. If you play on AirPods and sound thin, try the Mac's own mic or a wired one for a session and compare. More on that in why your AirPods mic sounds bad on a Mac.

Best mic settings for Discord on a Mac: the whole chain

Put together, here is a starting point that settles most "Discord mic sounds bad" cases on a Mac with headphones. Adjust from it rather than treating it as law.

macOS   System Settings > Sound > Input
  Device ................ your mic, by name
  Input volume .......... loudest laugh stays below the top of the meter

Discord   User Settings > Voice & Video > Voice
  Input Device .......... the same mic by name, not Default
  Input Volume .......... leave it; the level is set in macOS
  Input Mode ............ Voice Activity
    Auto sensitivity .... OFF
    Marker .............. between room noise and your softest word
    Loud room? .......... Push to Talk with a release delay instead
  Noise Suppression ..... off in a quiet room, Krisp with fans or a PC nearby
  Echo Cancellation ..... OFF on headphones, ON on speakers
  Automatic Gain Control  OFF once the macOS level is right
  Test .................. Let's Check, headphones on

Server   voice channel cog > Overview > Bitrate (needs edit rights)
  Speech ................ 64 kbps or more is plenty
  Music, streams ........ higher, if the server level allows it

Change one line at a time and test after each change. Five changes at once is how people end up with a setup nobody can explain, including them.

When the settings are right and you still sound bad

Sometimes every line above is set correctly and the voice is still thin, boomy or swimming in room. The settings only decide what Discord does with the sound it receives. They cannot add the body a small mic misses, and they cannot take the echo of a bare room out in a way that sounds natural.

At that point there are three honest ways to improve mic quality on Discord. Move the mic closer or soften the room; a rug and a heavy curtain do more than most people expect. Buy a better mic: the best USB microphone for calls, streams and podcasts compares the sensible choices. Or process your voice before it reaches Discord.

On a Mac, processing before Discord means a virtual microphone: an app takes your real mic, processes it, and presents the result as a new input device that Discord lists next to the others. The idea is explained in what a virtual microphone is and why your Mac would need one. Streamers who already run OBS sometimes do this with OBS filters, and which OBS mic filters to use and in what order covers that route, though it ties your Discord voice to OBS being open.

TunedMic is our app for this job. It adds a mic called TunedMic to your Mac. You read a short passage aloud for about a minute (the Sound Check), the app takes more than 30 measurements of your voice, your mic and your room, and builds nine profiles from them. In Discord you pick TunedMic as the Input Device. Since your voice arrives already processed, it makes sense to switch Discord's own Noise Suppression and Automatic Gain Control off, so two processors are not working on the same voice. Switching profiles changes the sound in every app at once, and the processing adds less than a tenth of a second of delay.

What it does not do matters just as much. TunedMic cannot fix packet loss, a low channel bitrate or a Voice Activity marker set too high: those live in Discord and in your network. It needs a Mac with Apple silicon (M1 or newer) and macOS 14 or later, there is no Windows version, and once the free 3-day trial is over, there are no refunds. So listen before you buy. The TunedMic home page has before and after recordings from three mics: a MacBook's built-in mic, a cheap wired headset and a Hollyland Lark M2 (once with a Dyson hair dryer running nearby). Start with the one closest to your setup, and if it sounds like the fix you need, the price is on the pricing page.

What to do before your next session

Open Discord, go to Voice & Video, and switch off Automatically determine input sensitivity. Put your headphones on, press Let's Check, and say a full sentence that fades out at the end: "okay, I think that's it for tonight". If "tonight" comes back whole, leave the marker where it is, and the most common fault is gone. If it gets cut, drag the marker down until it survives, then switch off Automatic Gain Control and listen once more. Fifteen minutes with one change at a time will do more than an evening of flipping every switch in the menu.