Choosing a Microphone for Speech and Transcription

Distance matters more than the microphone

Sound weakens quickly with distance. Each time you double the distance between mouth and microphone, the direct voice drops by about 6 dB, while room echo and background noise stay roughly the same. Moving a microphone from 15 centimeters to 60 centimeters is two doublings: the voice falls about 12 dB relative to the room, which you hear as a hollow, distant sound and a speech recognizer experiences as more noise.

That is why a modest microphone used close beats an expensive one used far away, and why every recommendation below is really about how close each type lets you get. The general recording advice in the transcription accuracy guide mentions microphone distance in passing; this article goes into choosing the microphone itself.

Microphone types for speech

• Lavalier (lapel) microphones clip to clothing 15 to 20 centimeters below the chin. They stay at a constant distance as the speaker moves, which makes them strong for interviews, lectures and video. Wired versions are inexpensive; wireless systems add freedom and batteries to manage.
• Dynamic microphones are rugged, need no power, and are relatively insensitive, so they pick up the voice close in front and much less of the room. They suit untreated rooms, podcasts and voice-overs, as long as the speaker stays within a hand's width or two.
• Condenser microphones are more sensitive and capture more detail, including more of the room. In a quiet, furnished space they sound excellent; in a bare, echoey room they reveal every reflection. Most need power, often 48-volt phantom power from an interface or recorder.
• USB microphones are dynamics or condensers with a built-in interface, so they plug straight into a computer. They are simple and good for desk work, but you can't easily use two on one computer at once, and you are tied to a computer.
• Shotgun microphones are highly directional tubes designed to reach a little further, for example on a boom pole just out of frame. They work well outdoors and on video sets, and less well in small reflective rooms, where they pick up reflections from behind and around the speaker.
• Headset microphones sit at a fixed distance from the mouth, which makes them consistent for presenters, instructors and long video calls, at the cost of being visible.
• Boundary microphones lie flat on a table and pick up voices around it. They're a practical compromise for meetings, though each person is farther away than with individual microphones.

Polar patterns and what they reject

A polar pattern describes which directions a microphone hears best. It decides how much of the room ends up in the recording.

Omnidirectional

hears equally in all directions; natural sound, forgiving placement, more room noise

Cardioid

hears the front, rejects the back; the default choice for speech

Supercardioid and hypercardioid

narrower front pickup, some sensitivity directly behind

Figure-eight

hears front and back equally, rejects the sides; useful for two people facing each other

Shotgun (lobar)

very narrow front pickup with some side and rear sensitivity

Most lavaliers are omnidirectional, which works because they are so close to the mouth. Most desk dynamics and condensers for speech are cardioid. Directional microphones also show the proximity effect: as you get very close, bass increases, which can make a voice sound boomy. A little distance or a low-cut filter fixes it.

Placement for each type

• Lavalier: clip it on the center of the chest, 15 to 20 centimeters below the chin, on the side the speaker turns toward if they look at an interviewer. Keep it away from jewelry, scarves and collars that rustle, and loop the cable to absorb tugs.
• Desk dynamic: 5 to 10 centimeters from the mouth, slightly off to the side so breath doesn't hit it directly, on a stand or arm rather than in a hand.
• Condenser: 15 to 30 centimeters away, with a pop filter in front, and the back of the microphone facing the quietest part of the room.
• Shotgun: aimed at the mouth from above or below, as close as the frame allows, rarely more than an arm's length away.
• Headset: at the corner of the mouth rather than directly in front, so breathing doesn't pop.

If the voice sounds muffled despite a good microphone, placement is usually the cause; muffled audio in recordings covers the common reasons.

Matching a microphone to the situation

Solo voice-over or podcast at a desk

cardioid dynamic in an untreated room, condenser in a quiet treated one

Two people seated for an interview

a lavalier each, or a dynamic each on stands

Lecturer or presenter who moves

wireless lavalier or headset

Video shoot with a camera

lavalier hidden on the speaker, or shotgun on a boom

Remote calls and webinars

headset, or a USB dynamic with headphones

Meeting around a table

boundary microphone, or individual microphones if accuracy matters

Interviews in the field

wired or wireless lavalier into a phone or recorder

When several microphones record in the same room, they pick up each other's voices, which creates echo-like smearing when mixed. That problem has its own article: recording with multiple microphones.

Spending sensibly

There's no need to rank brands to spend well. A sensible order of spending for speech is:

1. Something that gets the microphone close: a stand, a desk arm, a clip-on lavalier or a boom.
2. Accessories that prevent common problems: a pop filter or foam cover, a windscreen for outdoors, headphones for monitoring.
3. The microphone itself, chosen by type and pattern for your situation.
4. An interface or recorder with clean preamps, if you use XLR microphones; low-output dynamics in particular need plenty of clean gain.

Entry-level microphones of the right type, placed well, produce recordings that transcribe accurately. Spending more mainly buys a nicer tone, lower self-noise and better build, which matter for published audio more than for transcripts.

Example: two podcasters in an echoey room

They switch to two inexpensive cardioid dynamic microphones on short desk stands, each about 8 centimeters from one host's mouth and angled away from the other host, recorded to separate inputs. The kitchen hasn't changed, but each voice is now much louder than the room, so the echo practically disappears from the recording.

The next episode's transcript, at 40 credits (4¢), needs far fewer corrections, and the two channels let them balance a quiet host in editing before exporting a single mixed file.

Steps to choose and test a microphone

1. Write down where you record, who speaks, and whether the speaker moves or appears on camera.
2. Pick the type that keeps the microphone closest in that situation, using the table above.
3. Choose the polar pattern: omnidirectional for lavaliers, cardioid for most desk microphones, shotgun for distance on a set.
4. Check the connection: USB for one person at a computer, XLR into an interface or recorder for several microphones, a phone-compatible plug for mobile recording.
5. Place the microphone, set levels as described in the speech recording settings article, and record a two-minute test in the real room.
6. Listen on headphones and transcribe the test; adjust distance and angle before buying anything else.

Trade-offs and common mistakes

• Buying a sensitive condenser for an untreated room. It records the room faithfully, which is the problem.
• Using a shotgun indoors at a distance, expecting it to work like a telescope. It narrows pickup; it doesn't cancel distance.
• Placing a desk microphone beside the keyboard or laptop fan.
• Relying on a single microphone for a group when accuracy matters; the farthest speaker will suffer.
• Using Bluetooth earbuds as a recording microphone; their call mode is tuned for calls, not clean recordings.
• Forgetting phantom power for a condenser, then turning gain all the way up to compensate.
• Swapping microphones to fix echo or background noise that is really a room or distance problem; the background noise article explains which noises hurt recognition most.

How microphone choice affects mydubly results

mydubly doesn't record audio or recommend hardware. It works from whatever file you give it, audio or video, up to 2 hours long, and transcribes speech with Whisper after the browser converts the audio to 16 kHz mono. A close, clear microphone gives the recognizer a strong voice against a quiet background, which shows up as fewer missed words and fewer wrong names in the transcript, the subtitles and any translated transcript.

For dubbing, the translated voice comes from one of 8 stock voices, so your microphone doesn't change how the new voice sounds. It changes how accurate the words are, because the voice reads the translation of what was recognized. A 10-minute dub costs 500 credits (50¢), so it's worth getting the source recording right first. If several people speak, note that mydubly uses one voice for the whole video and doesn't label speakers, whatever microphones you used. Once you have a recording, try AI dubbing or a transcript from it.

For group discussions where several people share the recording, focus group transcription discusses setups that keep every voice usable.

Next step

Pick the type that fits your situation, get it close, and test in the real room before your first important session. Then upload a test recording to audio to text and read the transcript: it's the most honest review your microphone will get. For a full interview setup over a video call, see recording a remote interview.