A recording nobody can hear scores like one that was never made. How to set up your microphone, room and volume so TOEFL Speaking answers are captured.
8 min readMonthly & Premium
Quick answer
Your TOEFL Speaking score depends on what the microphone captured, not on what you said. Use a wired headset, position the microphone to the side of your mouth rather than in front of it, speak at conversational volume during the check, and record in a small room with soft surfaces.
The one part of Speaking that has nothing to do with English
Everything else in this course is about what to say. This lesson is about whether anyone can hear it. That sounds like a footnote and it is not, because a recording that captured nothing scores exactly the same as an answer you never gave, and a recording that captured you faintly, under a fan, from two feet away, scores like a candidate with poor delivery.
A small but steady share of the Speaking submissions we review arrive with the recordings technically present and effectively empty: the file exists, the timing is right, and there is almost no speech in it. Almost none of those are language failures. They are a microphone that was not the one the browser selected, a headset plugged into the wrong socket, a candidate who whispered because the room was shared, or a mute switch on a headset cable that nobody knew was there.
Fifteen minutes of setup work removes almost all of this risk permanently. That is a better return per minute than anything else in this course, and unlike pronunciation it is finished the moment you have done it once properly.
What to record with
Option
Verdict
Why
Wired headset with a boom microphone
Best
The microphone sits at a fixed distance from your mouth, so your level stays constant even when you move. No battery, no pairing, no dropout.
Wired earbuds with an inline microphone
Good
Perfectly usable. Keep the inline module out from under clothing, since fabric rubbing against it is louder than your voice.
Bluetooth headset
Risky
Pairing can drop, the battery can die mid section, and some systems fall back to the laptop microphone silently when the connection stutters.
Built in laptop microphone
Last resort
It works, but it sits below your hands, picks up keyboard and desk noise, and captures a lot of room reflection along with your voice.
Phone held in your hand
Avoid
Distance changes constantly as you gesture, which makes your volume rise and fall for reasons that have nothing to do with your speech.
Ranked by how reliably they capture normal speech in a normal room.
A setup routine you do once and then repeat before every test
1
Plug in and check what the system actually selected
Plugging in a headset does not always make it the active input. Open your sound settings and look at which input device is selected and whether the level meter moves when you talk. This single check catches the most common cause of an empty recording, and it takes ten seconds.
2
Position the microphone beside your mouth, not in front of it
About two fingers from the corner of your mouth, off to the side. Directly in front means every p and b sound hits the microphone as a burst of air, which distorts exactly the consonants that carry meaning. Off to the side gives you full volume without the blast.
3
Do the check at your real speaking volume
Most people test at a cautious murmur and then speak up when the test starts, or, far worse, test normally and then drop to a murmur once they feel self conscious. Say a full sentence at the volume you intend to use. A check passed at whisper volume proves nothing about the recording that follows.
4
Record thirty seconds and listen back
Not the level meter, the actual audio. Listen for your own volume, for room echo, and for anything in the background you had stopped noticing. This is the only step that verifies the whole chain from your mouth to the file, and it is the step people skip.
5
Kill the background before you start
Fan, air conditioning, notifications, phone on silent rather than vibrate, window closed. Constant noise is worse than occasional noise, because it sits underneath your voice for the entire recording rather than interrupting it once.
The room matters more than you expect
Hard flat surfaces reflect sound. In a room with bare walls, a tiled floor and a large window, your voice reaches the microphone twice: once directly, and again a few milliseconds later off the surfaces. The result is the boxy, distant quality you hear on recordings made in empty rooms, and it makes consonants harder to distinguish. Since consonants are where English carries most of its grammatical information, this quietly damages exactly the thing being scored.
You do not need acoustic treatment. You need soft things. A room with curtains, a rug, a bed or a sofa will sound noticeably better than a kitchen or a hallway. Smaller is better than larger. Sitting closer to a wall covered in books is better than sitting in the middle of an open space. If you only have a hard room, sit close to the microphone, which raises the ratio of your direct voice to the reflected sound.
The other room variable is people. If you are somewhere you might be interrupted, you will speak more quietly and stop more often, and both of those cost you real points. Practising in a room where you feel free to speak at full volume is not a luxury, it is part of training the delivery the rubric is listening for.
Volume, and why almost everyone is too quiet
There is a strong pull toward speaking quietly during a test. You are in a room by yourself, talking to a screen, possibly with other people in the building, and it feels absurd to project. Almost every candidate underestimates how much this affects them.
Quiet speech does more than lower the level in the file. It changes how you speak. Volume and pitch movement are physically linked, so a quiet voice tends toward monotone, which is exactly the delivery pattern that costs points. Quiet speech also encourages you to under-articulate final consonants, which is where grammar lives. So the same nervousness that makes you drop your volume also flattens your intonation and blurs your word endings, and all three of those show up in the same score.
The fix is to deliberately speak as if explaining something to a person sitting across a table from you. Not shouting, not projecting to a room, just the volume of ordinary conversation with someone you are not trying to be quiet around. Record yourself at that volume once, listen back, and notice that it does not sound loud at all in the recording. It sounds normal, which is the point.
The pre-test hardware check
Headset plugged in fully, and the wire routed so it will not tug during the section.
Sound settings confirm the headset is the selected input, and the level meter moves when I speak.
No mute switch engaged on the cable or the headset itself.
Microphone positioned two fingers from the corner of my mouth, off to the side.
A thirty second test recording made and listened back to, not just watched on a meter.
Fan, air conditioning and notifications off, phone silent, door closed.
Room has soft surfaces, or I am sitting close to the microphone to compensate.
I know I will not be interrupted for the next fifteen minutes.
I did the check at the volume I actually intend to speak at.
Setup judgement
Three practical decisions that change what ends up in the file.
Where should the microphone sit relative to your mouth?
Off to the side keeps you close enough for a strong, consistent level while keeping the air from plosive consonants out of the capsule. It is why boom microphones on headsets are angled that way by design.Directly in front means every p, b and t sends a burst of air into the microphone, distorting precisely the consonants that carry grammatical information.Near the collar picks up clothing movement and breath noise, and it sits far enough off axis that your level drops whenever you turn your head.A desk microphone is much further away, so it captures more room reflection relative to your voice, and it picks up every contact with the desk surface.
What is the most reliable way to confirm your setup is working?
Listening back is the only check that tests the whole chain, from your voice through the selected device to the file. It catches low volume, room echo, background noise and wrong-device problems in one step.A moving meter tells you some input is live, but not which device it is or how the result actually sounds. A laptop microphone picking up your voice from three feet away moves the meter perfectly well.Physical connection is necessary and not sufficient. Systems regularly keep a different device selected after a headset is plugged in.Yesterday is not evidence about today. Device selection changes after restarts, updates and any other audio application being opened in between.
Why does speaking too quietly cost more than just a low recording level?
Volume and pitch movement are physically connected, so a quiet voice flattens, and quiet speakers under-articulate final consonants. Delivery and language both take the hit, on top of the raw audibility problem.The recording length is fixed regardless of how loudly you speak. What changes is how much of what you said is usable.Microphones do not cut out below a volume threshold in normal use. The audio is captured, it is simply too quiet and too flat to score well.The window is fixed by the test, at 44 seconds for interview answers. Nothing you do with your voice changes it.
What to do when something goes wrong mid section
If you realise during item 3 that your headset has slipped, fix it between items, not during one. The seconds inside a recording window are worth far more than the improvement from a slightly better microphone position, and reaching for your head mid answer produces handling noise that is louder than your voice.
If you suspect the microphone has failed completely, keep speaking anyway. You cannot verify it, you cannot repair it, and every item you abandon on suspicion is an item that definitely scores nothing. A failure you imagined costs you nothing if you kept going, and everything if you stopped.
After a practice test, always listen to at least one of your own recordings rather than only reading the feedback. It takes a minute and it is the only way to catch a slowly degrading setup: a headset cable starting to fail, a new background noise you have habituated to, a volume drift that crept in over a few weeks. Feedback tells you about your English. Only the audio tells you about your setup.
Monthly & Premium
8 more sections in this lesson
You have read the opening. The rest covers the method in full, with worked examples and practice questions that explain why each wrong answer is wrong.
All 45 lessons: Reading, Listening, Writing and Speaking
Real exam audio and record-yourself speaking drills
A wired headset with a boom microphone is the most reliable, because the distance from your mouth stays constant and there is no battery or pairing to fail. Wired earbuds with an inline microphone are fine too. Bluetooth is the riskiest option because a dropout can silently switch the input to your laptop microphone.
Almost always a device selection problem, a mute switch on the cable, or speaking too quietly for the microphone that was actually active. Confirming the selected input in your sound settings and making a short test recording that you listen back to removes nearly all of this risk.
At the volume you would use explaining something to someone across a table. Most candidates are far too quiet, and quiet speech also flattens intonation and blurs final consonants, so it costs delivery and language points on top of the audibility problem.
Yes. Hard surfaces reflect your voice back into the microphone slightly delayed, which smears consonants and makes you harder to understand. A room with curtains, a rug or soft furniture is much better than a kitchen or a hallway. If you only have a hard room, sit closer to the microphone.
Always. Your speaking volume, your distance and how much of your own voice you hear all adapt to the setup. Practising on a phone and testing on a laptop headset means your habits were calibrated for the wrong conditions, and you find out during the first scored item.
Keep speaking. You cannot verify or repair it mid section, and an item you abandon scores nothing for certain, while an item you complete on a working microphone scores normally. Fix anything physical between items rather than inside a recording window.