How to record yourself speaking and actually learn from it

Recording yourself is the most-recommended speaking tip there is, and probably the least-used. The advice nearly always stops at "record yourself and watch it back", as if pressing play were the whole method. In practice, people watch about ten seconds, decide they hate their voice, and close the file. Nothing changes because nothing got measured.

This guide covers the part that usually gets skipped: how to set the recording up so takes are comparable, how to review it without spiralling, what to count, and how to use a second take to prove a fix worked. It is the detailed version of the "record it and play it back" step in our guide on how to practice public speaking, and it feeds the practice loop in how to improve public speaking.

Here is the sequence:

1. Choose real material: the talk, answer, or update you actually have to give.
2. Set the phone at eye level, about two metres away, in landscape.
3. Record one full take without stopping.
4. Before you press play, write down what you expect to see.
5. Review in three passes: audio only, picture only, then full.
6. Count fillers and time the talk on a one-line scorecard.
7. Pick the single loudest problem and fix only that.
8. Record a second take and compare the numbers.

Why does recording yourself speaking work when practising in your head doesn't?

When you rehearse, you are both the speaker and the only judge, and you are a bad judge of yourself in real time. Your attention goes on what to say next, not on how the last sentence came out. You don't hear the "um" you just said, you don't notice the pace creeping up, and you don't see your hand going back to your pocket for the fifth time.

A recording splits those two jobs. You give the talk once as the speaker. Later you review it as a listener, with nothing to plan and all of your attention free to notice things. That is why the first playback is so often a surprise. The talk you remember giving and the talk on the file are different talks, and the file is the one your audience heard.

It also gives you something your memory can't: a fixed record. Next week you can play the old take beside the new one and see whether the thing you worked on moved. Without that, "I think I'm getting better" is a feeling, not a measurement.

How do you set up your phone to record yourself speaking?

You don't need a camera, microphone, or ring light. A phone is enough. The setup matters because you want every take to look and sound the same, so that the only thing changing between takes is you.

Setting	What to do	Why
Height	Lens at eye level, on a shelf, stack of books, or small tripod	A low angle looking up changes how you hold your head and makes posture hard to judge
Distance	About 2 metres for standing talks; arm's length for a video-call or interview answer	Far enough to see hands and stance, close enough for the phone mic to pick up your voice clearly
Orientation	Landscape	Shows your gestures and the space around you; portrait crops your hands out
Framing	Head to hips for a standing talk; head and shoulders for a seated answer	Matches what a real audience or interviewer sees
Light	Face a window or lamp, don't sit with it behind you	Backlighting turns your face into a silhouette and hides expressions
Room	Quiet, soft furnishings if possible, no music or TV	Echo and background noise make it hard to judge clarity
Phone	Do Not Disturb on, enough storage for 10 minutes of video	An alert halfway through ruins a full take

Use the phone's own microphone and keep the distance the same each time. A clip-on lavalier mic sounds better, but it also changes your recordings, so take 1 and take 6 are no longer comparable. If you do switch, start a new baseline.

Stand where you would stand to give the talk, and put the phone where your audience would be. If your real talk is in a meeting room, record seated at a table. If it's on a video call, record through the call software itself, so you see what the other people see. Our guide to being confident on video calls covers camera framing for that case.

Should you record audio only or video?

Use both, for different jobs. An audio-only voice memo is quick, easy to do daily, and enough for pace, fillers, clarity, and vocal variety. Video is necessary for body language: eye contact, hands, posture, fidgeting. A good default is audio-only for quick daily reps and one video take per session on the day you plan to review properly.

What should you record?

Record the real thing, not a generic exercise. Practise the actual talk, interview answer, toast, or meeting update you have coming up. Specific material gives you specific feedback, and you will want to know how that talk lands, not how a random paragraph does.

If you have nothing scheduled, use one of these:

• A two-minute opinion. Argue for something you believe. It forces you to structure a point without a script.
• An impromptu prompt. Set a timer, pick a question ("What's a decision you'd make differently?") and answer it with no preparation. This is the closest practice to a real conversation, and our guide to impromptu speaking has frameworks to try.
• A "tell me about yourself" answer, if you have interviews coming up.

Keep takes short: two to three minutes is enough to show your habits and short enough to review properly. A twenty-minute recording is one you will never finish watching.

Record the whole thing in one go. Don't stop and restart when you stumble. The stumble and what you do next are exactly the data you want. Real audiences don't get a restart, so your practice take shouldn't either.

How do you watch yourself on video without cringing?

Almost everyone dislikes their first recording, and part of that is not about your speaking at all.

Your voice sounds strange because it is missing a part you normally hear. When you speak, sound reaches your inner ear through the air and through the bones of your head. A recording only has the air part. Researchers who measured both paths found the bone-conducted part dominated across the 1–2 kHz range for most speech sounds they tested (Reinfeldt et al., Journal of the Acoustical Society of America, 2010). So the recording is the voice everyone else hears, minus something you have heard your whole life. That's why it sounds wrong to you and normal to everyone else.

The cringe is mostly about recognition. In one study, 80 people rated a set of voice recordings that secretly included their own. They rated their own voice as more attractive than other listeners did (Hughes & Harrison, Perception, 2013). When they didn't know it was them, they liked it. The discomfort comes from knowing it's you, not from the voice itself.

How you watch changes what you take from it. In research on speech anxiety, people who prepared before watching their recording judged their performance more accurately and less harshly. The preparation had three steps: predict in detail what you expect to see, picture yourself giving the talk, then watch the recording as if it were a stranger (Harvey, Clark, Ehlers & Rapee, Behaviour Research and Therapy, 2000). A replication with 40 highly socially anxious people found the benefit carried into a second speech 30 minutes later (Kim, Lundh & Harvey, 2002). In a study of 95 speech-anxious students, the same preparation had a strong effect on how people saw their performance, though a weak effect on anxiety itself (Rodebaugh, 2004).

In practice, before you press play:

1. Write down three things you expect to see ("I'll look nervous", "I'll say um constantly", "I'll rush the start").
2. Watch it as if you were coaching a friend: what would you tell them, kindly and specifically?
3. Check your three predictions against what's actually on screen.

Most people find at least one prediction was worse than reality. That gap is useful information on its own: the audience saw a calmer speaker than the one you felt like.

What to look for when you record yourself speaking: the three-pass review

Watching everything at once is overwhelming, and you end up with a general feeling ("that was bad") instead of a fix. Split the review into three short passes, each with one job. This is close to the method speaking coach JD Schramm teaches, which separates audio-only and picture-only viewing (JD Schramm).

Pass 1: audio only

Cover the screen, or play the file with the phone face down. Without the picture, you hear the voice the way someone on a phone call would. Keep a pen in your hand and:

• Tally every filler word: um, uh, like, so, you know, kind of. Put a mark for each one. Divide by the length of the talk in minutes to get fillers per minute.
• Note the total time. Compare it with your target. If a "five-minute" talk ran 3:40, you rushed it or skipped material.
• Mark the moments that sound rushed, mumbled, or flat with the timestamp. You're not fixing them yet, just locating them.

This pass usually tells you the most. If filler words are the loudest problem, how to stop using filler words has the replacement drill. If the whole thing is fast, start with how to stop talking too fast.

Pass 2: picture only

Mute the sound and watch again. Without your voice to distract you, body language becomes obvious:

• Where are your eyes? On the lens (your audience), on notes, or on the ceiling while you think?
• What are your hands doing? Count any repeated movement: touching your face, adjusting glasses, clicking a pen, hands in and out of pockets.
• Are you swaying, shifting weight, or stepping back and forth with no reason?
• Does your face match what you are saying, or is it blank throughout?

Watching on fast-forward at 2× speed makes repeated habits even more obvious. A gesture that seems natural at normal speed looks like a tic when it happens eight times in a minute.

Pass 3: full playback, as a listener

Now watch it once with everything on, as if you were the audience. Don't count anything. Answer two questions at the end:

1. What was the point? Can you say it in one sentence? If you can't, the audience couldn't either, and the problem is structure, not delivery.
2. Did the ending land, or did the talk trail off with "so, yeah, that's it"?

Keep a scorecard so takes are comparable

Write one line per take in a notes app or notebook. It takes thirty seconds and turns "I think I'm better" into evidence.

Take	Date	Length	Fillers/min	Rushed spots	Point clear?	Ending landed?	Fix for next take
1	Mon	3:10	9	3	No	Trailed off	Write a one-sentence point first
2	Mon	2:55	8	2	Yes	Trailed off	Memorise last sentence
3	Wed	3:05	6	1	Yes	Yes	Pause instead of "so" at transitions

Two rules make this work:

• Same material, same setup. Compare take 1 and take 3 of the same talk, recorded the same way. Different talks have different difficulty, so their numbers aren't directly comparable.
• One fix per take. If you change your opening, your pace, and your hands at once, you won't know which change helped. Pick the single loudest problem from the scorecard and work only on that.

Keep your first recording. Even if it's painful, it's your baseline. A month from now, playing take 1 next to your latest take is the clearest proof you'll get that the practice is working.

How do you record yourself giving a speech or presentation?

For a full speech or slide presentation, the setup is the same but a few things change:

• Record the whole run, including the transitions. The seams between sections are where most speakers lose the thread, and they're invisible if you only record sections. Our guide to speech transitions covers what to listen for at each seam.
• Capture the slides too. If you present from a laptop, start a screen recording alongside the phone video, or record the presentation software's own rehearsal mode. That way you can check whether you're reading slides or talking to the room.
• Set up the phone where the back row would be. It shows whether you're loud enough and whether gestures read from a distance.
• Rehearse your recovery on camera. At least once, deliberately lose your place, pause, and pick the thread back up. Watching yourself recover calmly is reassuring evidence for the day.

For long talks, don't review every minute in detail. Review the first two minutes, one transition, and the last minute closely, and skim the rest at 2× speed for timing and obvious habits.

How often should you record yourself?

Little and often beats a long weekly session. A sustainable rhythm for most people:

• Daily or most days: one two-minute audio-only take, reviewed with Pass 1 only. Five minutes total.
• Two or three times a week: one video take of real material, reviewed with all three passes, and a second take to test the fix.
• Once a week: play your oldest take next to your newest and compare scorecard lines.

Before a specific talk, record a full run-through at least twice: once about a week out to find the big problems, and once two or three days out to check the fixes held. Don't do a heavy review the night before. At that point, a recording is for confidence, not for redesign. Our guide on what to do the night before a presentation has the rest of that routine.

Common mistakes when recording yourself

• Restarting every time you stumble. You end up with a perfect thirty seconds and no information about how you recover.
• Watching only once, all at once. You get one big feeling and no specific fix. Use the three passes.
• Judging your voice instead of your speaking. The "my voice sounds weird" reaction is mostly the missing bone conduction. Your audience has always heard the recorded version. Focus on what you can change: pace, fillers, clarity, ending.
• Changing the setup between takes. New room, new angle, new microphone, and the comparison stops meaning anything.
• Fixing everything at once. One fix per take. The scorecard will tell you what's next.
• Never comparing. A recording you don't compare against an earlier one is just a reminder that you dislike watching yourself.

Key takeaways

• Prop the phone at eye level, about two metres away, in landscape, and keep that setup identical across takes.
• Record real material in one unbroken take of two to three minutes; the stumbles are part of the data.
• Your recorded voice sounds strange mainly because it lacks the bone-conducted sound you hear when you speak, not because it's a bad voice.
• Before you watch, write down what you expect to see, then watch as if it were a stranger; prepared viewers judge themselves more accurately.
• Review in three passes: audio only (fillers, time, rushed spots), picture only (eyes, hands, posture), then full (was the point clear, did the ending land).
• Log one scorecard line per take and change one thing at a time.
• Keep your first recording as a baseline and compare it with your latest take every week.

Frequently asked questions

How do I record myself speaking on my phone?

Use the built-in camera app for video or a voice memo app for audio only. Put the phone at eye level about two metres away in landscape, turn on Do Not Disturb, face a window or lamp, and record one full take of real material without stopping. The phone's own microphone is fine as long as the room is quiet and you keep the same distance every time.

Why do I hate the sound of my voice on a recording?

When you speak, you hear your voice through the air and through the bones of your skull. A recording only captures the air part, so it's missing a component you have always heard. It sounds unfamiliar to you and normal to everyone else. Research also suggests the discomfort is about recognition: in one study, people rated their own voice as more attractive than other listeners did when they didn't know it was theirs.

How do I watch myself on video without cringing?

Prepare before you press play. Write down three things you expect to see, picture yourself giving the talk, then watch the recording as if you were coaching a friend. Studies of speech anxiety found that this preparation led to more accurate, less harsh judgements. Reviewing against a short checklist (filler count, time, eye contact) also helps, because it gives you a task instead of a general impression to react to.

What should I look for when I record myself speaking?

Use three passes. With audio only, count filler words, time the talk, and mark rushed or mumbled moments. With the sound muted, watch eye contact, hands, posture, and repeated fidgets. With full playback, check whether the main point was clear in one sentence and whether the ending landed. Then pick the single loudest problem to fix in the next take.

Should I record audio or video when practising a presentation?

Both, for different jobs. Audio-only is quick and enough for pace, filler words, clarity, and vocal variety, so it suits short daily practice. Video is necessary for body language: eye contact, gestures, posture. A good routine is daily audio-only takes plus one or two full video takes a week of the real presentation.

How long should a practice recording be?

Two to three minutes is enough to show your habits and short enough to review properly in three passes. For a full speech, record the whole run, then review the first two minutes, one transition, and the final minute in detail and skim the rest at double speed.

Does recording yourself actually improve public speaking?

It does when you review it with a purpose and record again. A recording on its own just shows you what happened. The improvement comes from the loop: record, count specific things, change one thing, record again, and compare. Simply watching yourself without a plan can make some anxious speakers feel worse, so use a checklist and prepare before viewing.

Conclusion

Recording yourself speaking only helps if you can compare one take with the next: the same setup, a three-pass review, one scorecard line, and one fix per take. Prop your phone up now, record two minutes of something you actually have to say, and write down your filler count. That number is your baseline.