Best Microphone for Dictation: What Actually Matters

For dictation, the microphone that matters most is the one that sits at a steady, short distance from your mouth. A close headset or a handheld speech microphone gives the recognizer a stronger voice signal than a laptop's built-in microphone, and that matters more than what the microphone costs. In a quiet room, the built-in microphone is often enough for short notes. In a noisy room, or for long sessions, a wired or USB headset is the most consistent choice. Bluetooth earbuds are the least predictable option, because their microphone can move the connection into a narrower call mode.

This guide covers the four microphone types, the research on distance and audio bandwidth, and a short test you can run on your own setup. Most of the evidence comes from speech recognition research rather than product reviews. Where no controlled study exists, we say so. For the wider subject, see what dictation is and how to dictate. The dictation overview lists every guide in the series.

Distance to your mouth matters more than the microphone's price

Distance decides how much of your voice reaches the microphone compared with the room. Under ideal conditions, a sound level drops by 6 dB each time the distance from the source doubles ¹. Under the same ideal conditions, a microphone a foot from your mouth, instead of an inch away, receives a signal about 21.6 dB weaker ². Room noise and echo reach the microphone from every direction, so as your voice gets weaker relative to them, recognition gets harder. This is physics, and no price tier changes it.

The Dragon guidance from Nuance gives the most specific placement advice we found. For a headset, keep the element about an inch from the side of your mouth, but not touching it ³. For a handheld microphone, keep it 1 to 3 inches from your mouth and below or to the side, not directly in front ⁴. These are instructions from one software maker, not a universal standard, but they show how little room there is for error.

The research points the same way. Prasad and Sim treated distance as the central variable in their work, so they recorded Wall Street Journal audio with far-field microphones at eight distances from the source ⁵. Their aim was to help a recognizer cope with distance, which is the same problem a laptop microphone faces on a desk.

A second study shows the size of the gap. Ravanelli and colleagues recorded read Wall Street Journal sentences in a studio with a close-talking microphone and got a word error rate of 3.7% ⁶. With a single ceiling microphone in an apartment, the DNN recognizer's word error rate on real recordings was 12.0% ⁷. The authors say that for distant speech, speech recognition performance is dramatically reduced compared with the close-talking case ⁸. That microphone was far from the speakers, and it is a preprint with a 2017 recognizer, so treat the numbers as the size of the effect, not a forecast for your laptop.

The first pair in the chart is about distance, and the second is about bandwidth, which the next sections explain. The two pairs come from different experiments, so the bars should not be compared across pairs.

Four kinds of microphone suit dictation, each with a different trade-off

There are four microphone types that people use for dictation. They differ mainly in how far the microphone sits from your mouth, and whether that distance stays the same from one session to the next. Distance and consistency are the two properties that the research and the vendor guidance both point to.

A handheld speech microphone asks more of you than a headset. Nuance says to hold it below or to the side of the mouth to avoid breathing noise, not to use a stand or holder, and to be careful not to move the microphone or your head suddenly ( Nuance Dragon microphone placement). Each of those movements changes the distance, which is the variable the research above tracks. A handheld microphone suits a desk where you dictate in one place and want to put the microphone down between sessions. If your hand moves while you talk, a headset is the better choice.

The rest of this guide takes each type in turn, starting with the one most people should choose for long sessions.

A wired or USB headset is the most consistent choice for long sessions

A wired or USB headset with a boom microphone is the most consistent choice, because it holds the microphone at the same distance and angle every session. Consistency matters because you can only fix the position you use. Nuance's guidance is to position the headset the same way each time you use it, and to point the front of the microphone toward your mouth ( Nuance Dragon microphone placement). A headset that slides or drifts gives the recognizer a different signal each time.

Bandwidth is the other reason. Bauer and colleagues state that dictation and other large-vocabulary applications principally use a sampling rate of at least 16 kHz, especially in noisy environments ⁹. Their summary of the Aurora tests reports an average 14% relative gain for noisy speech sampled at 16 kHz instead of 8 kHz, across weak, medium and high mismatch conditions ¹⁰. A wired headset has no radio link that could narrow the audio, so the codec question below does not arise for it.

Wireless headsets are a harder case. The question is whether the link carries your voice at wideband quality, and that depends on the design of the headset and its codec. We did not find a controlled test that compares a wireless headset with a wired one for dictation accuracy, so the choice cannot rest on a measured result. Read the question on wireless headsets for dictation and check the headset's specification for its voice codec before you buy.

Built-in laptop microphones work in a quiet room and struggle in a noisy one

A built-in laptop microphone is fine for dictating in a quiet room, where there is little for it to pick up apart from your voice. In a noisy room, or one with echo, it picks up everything, and Apple's guidance is direct: if you use Dictation in a noisy environment or a room with excessive echo, a headset microphone may help ¹¹.

The reason is distance. A built-in microphone sits in the lid or the bezel, usually farther from your mouth than a headset. By the distance rule in the first section, which holds only approximately in a room ¹, that puts your voice at a lower level relative to the room, and the room's noise has a larger share of what the microphone records. A quiet room lowers the room's share, which is why the built-in microphone works there.

Two things are not settled here. First, we searched for a peer-reviewed comparison of word error rate for built-in and headset microphones in noise, and did not find one. Vendor blog posts report large differences, but we could not verify their data, so this guide does not repeat their numbers. Second, the built-in microphone is not always the one the computer uses. On a Mac, check the input in Sound settings and raise the input volume if the Mac does not respond to your voice ¹². The question on laptop microphones goes through the quiet-room case in more detail, and the background noise guide covers choosing a room.

Bluetooth earbuds can drop to a low-quality call microphone while you dictate

When the microphone in Bluetooth earbuds is active, the connection has to carry your voice as well as the sound you hear, and the voice path can use a narrower codec. The Headset Profile carries voice either at 8 kHz with the CVSD codec or at 16 kHz with the mSBC wideband codec ¹³. The source describes that profile rather than the Hands-Free Profile most earbuds use, but the codec families are the same, and the specification should be checked for any particular device.

Bandwidth is the link to accuracy. In a preliminary Wall Street Journal task, the same recognizer gave a word error rate of 8.67% on narrowband speech and 8.12% on wideband speech ^{14 15}. That is a clean read-speech result, and the gap is small. Bauer's statement that dictation expects 16 kHz, especially in noise, suggests the gap grows in a noisy room, but we did not find a controlled dictation test that measured how much.

The AirPods case is the one most people ask about. The question on AirPods and dictation covers it in full. Community reports describe the same switch, where playback drops to a lower-quality call profile while the microphone is in use. We did not find an official Apple statement that confirms a dictation accuracy effect, so treat the pattern as reported rather than measured. If you dictate with earbuds anyway, choose a wired or built-in microphone as the input, so the earbuds' link does not have to carry your voice. For iPhone-specific steps, see the iPhone dictation guide.

Cardioid microphones are least sensitive to sound from directly behind

A cardioid microphone, the pattern behind many noise-canceling microphones, rejects sound from behind it. The Shure explanation of the pattern says a cardioid is least sensitive to sound arriving from directly behind, at 180 degrees off-axis ¹⁶. On a headset, the boom microphone points at the mouth, so the rejection works against sound from behind it, which is usually the room. A pattern rejects sound by direction, so it does not separate your voice from a person speaking in front of you. For the definition of the term, see the noise-canceling microphone entry.

Two points are easy to miss. First, the rear null is a design property measured on axis, so reflections from the room still reach the microphone. Second, placement still matters. Nuance advises placing the headset element to the side of the mouth to avoid breathing noise ( Nuance Dragon microphone placement). A cardioid headset in the wrong position can still pick up breath and clicks. We did not test software noise suppression, which is a separate feature from a microphone pattern, and we make no claim about it here.

A short test of your own setup tells you whether your microphone is good enough

The best test is your own. Dictate the same passage with each microphone, in the room you usually use, and compare the errors. Word error rate is the number of wrong, missing and extra words divided by the number of words in the passage, and it is the measure that the research above uses.

1. Place the microphone correctly. A headset goes about an inch from the side of your mouth ³, with the front toward your mouth. A handheld microphone goes 1 to 3 inches away ⁴, below or to the side.
2. Choose the input. On a Mac, choose Apple menu > System Settings, click Sound in the sidebar, and select the microphone you want. Raise its input volume if the Mac does not respond ¹².
3. Read a fixed passage of about 100 words at a normal volume. Dictate it in the app you use every day. Then repeat with the second microphone, with the same room, the same fan or window, and the same pace.
4. Count the errors for each microphone: wrong words, missing words and extra words. Divide by the passage length to get the word error rate.
5. Test the worst case. Repeat once with the noise you actually have, such as a fan, traffic, or another person talking.
6. Keep the one with fewer errors. If a built-in microphone scores about as well as a headset in your quiet room, the cheaper option is fine. If it falls behind in your noisy room, the headset is the better buy.

The evidence in this guide has limits, and you should weigh it that way. The research numbers in this guide come from four studies, and most use read sentences or telephone corpora rather than natural dictation. None compares the microphones used by the apps people buy today. Recognizers have improved since then, which probably narrows the gaps shown above, but we could not measure how much from sources we could verify. That is the reason for the test in this section: your room, your voice and your app are the variables that the published studies leave out.

Once you know which microphone works, the other settings matter more. The dictation mistakes guide covers accuracy, and the dictation not working guide covers setup faults such as a muted input. If you want a free check before you change anything, the free browser dictation tool lets you test a microphone without installing software. Yap, which we make, runs its dictation on the device, so the placement and input advice above applies to it as well. The short answer is that distance to your mouth matters more than the price of the microphone. A headset held in one position is the most dependable choice, a built-in microphone is fine in a quiet room, and Bluetooth earbuds are the option to test before you rely on them.