How long should an interview answer be?
Interviewers rarely time you with a stopwatch, but they feel the shape. A behavioral story that ends in forty seconds sounds like a headline. The same story at three minutes sounds like you cannot tell what mattered. The ranges below are for the spoken answer, not for silent thinking before you start and not for the coding you do after you have explained the approach.
Sixty to ninety seconds is enough for “tell me about yourself,” a weakness, a values question, or “why this job.” You have time for one present-tense sentence, the relevant path, and why you are in this conversation. You do not have time for every role since school. If you hear yourself narrating a job description, you are past the point where a new fact is helping.
Behavioral stories, including STAR, fit in about one to two minutes. Spend roughly fifteen seconds on the situation, most of the minute on the actions you took, and the last fifteen or twenty seconds on the result. The failure mode is a long situation and a thin action. The STAR guide on the long action is the written version of that split. Practice it here with the timer set to Behavioral / STAR.
A coding explanation is the talk before and during the solution, not the whole hour. Ninety seconds to three minutes is enough to restate the input, name a slow correct approach, name the one you will build, and say the cost. System design needs more room: three to seven minutes for a first pass that an interviewer can interrupt. If you talk for twelve minutes before they can ask about the cache, you have buried the decision.
| Question | Spoken length | What has to fit |
|---|---|---|
| Tell me about yourself | 60–90 seconds | Now, the path that got you here, and why this role |
| Behavioral / STAR | 60–120 seconds | A short situation, a long middle on what you did, a result |
| Weakness | 45–90 seconds | A real gap, what you practice, and the boundary |
| Why this role | 45–75 seconds | The work, a reason you looked, and a proof point |
| Coding explanation | 90–180 seconds | Restate, naive approach, better one, cost |
| System design, first pass | 3–7 minutes | Users, API, data, and the bottleneck you accept |
| Culture or values | 45–90 seconds | One decision with a cost, not a definition |
These are practice targets, not a rule a company publishes. A hiring manager who says “keep going” wants the extra minute. A recruiter screen often wants the short end. Use the range as the default, then follow the person in the room. The timer is how you learn the default with your own voice before you need it.
What the timer measures
Duration is the clock from record to stop. You can also type the seconds if you practiced away from the mic and want the transcript scored against a length you already know. Words per minute is the word count divided by the minutes you spoke. A calm interview answer usually lands between 120 and 160 words per minute. Faster than that, names and numbers blur. Slower than about 110, it starts to sound like you are searching for the next sentence instead of telling it.
The longest pause is the longest stretch of quiet between bits of speech, measured from the microphone on this device. A pause under about two seconds reads as thinking. A pause past four seconds needs a bridge, something like “The decision was,” so the interviewer knows you are still in the answer. The tool only counts a pause that ends when you speak again. The silence after you are finished, while your hand moves to Stop, does not count.
Repeated words are two different things. A stutter is the same word twice in a row, as in “the the” or “I I.” A lean is a content word used three or more times, such as “team” or “impact,” when a specific name would be clearer. Both show up in the result so you can hear the habit instead of guessing.
How to cut filler words
The counter looks for um, uh, uhh, umm, uhm, er, ah, you know, I mean, kind of, sort of, basically, actually, literally, like, so, well, and right. The first group is almost always filler. like, so, well, right, actually, kind of, sort of, and literally are often real words, so they count at half weight in the score and they are marked in the list. Read the line. Keep “so” when it means therefore. Cut it when it is only a runway into the sentence.
Fillers show up when your mouth starts before the next phrase is chosen. The fix is not a personality transplant. It is a smaller gap. Finish the word you are on, close your mouth, and start the next word when you have it. A silent half-second is easier to listen to than “um, like, basically.” People who try to ban “um” by talking faster usually add more of them. Slow the join between sentences, not the sentence itself.
A useful target in practice is under two weighted fillers a minute, and no “um” or “uh” in the first sentence. The opening is where interviewers decide whether you know the story. If the first line is clean, later fillers bother them less. Record the same answer twice. On the second take, you are not allowed to start a sentence with so, well, or yeah. That single rule removes a lot of the count without making you sound scripted.
Practice drills
Drill one is the window. Pick a behavioral question from the hub, or press Random question. Set the type so the ideal range is visible. Answer once without stopping the clock early to “fix” yourself. Score it. If you are short, the missing piece is almost always the result or the action, not another sentence of context. If you are long, delete the first two sentences and start on the day the problem showed up.
Drill two is the filler pass. Read the filler list out loud from the result, not from memory. Say the answer again and replace each um with silence. Do not replace it with “like.” Then score the second take. You are done with the drill when the clear fillers (um, uh, you know, basically, I mean) are at zero, even if a few real uses of “so” remain.
Drill three is five answers, one sitting. Use one “tell me about yourself,” two behavioral stories, one weakness, and one “why this role.” Stop at ninety seconds even if you have more. The point is the stop, not the perfect ending. Write down only the tip the score gives you, then do the next question. A pile of notes is how people avoid the next recording.
Drill four is the pause bridge. Answer a design or coding question and, on purpose, pause for one second before the tradeoff. Then pause for four seconds and notice how different it feels. The score treats a completed pause around one second as a strength and a pause past four seconds as something to cover with a short bridge. You want the first feeling to be the one you can repeat on a call.
Use a real question, then your own
Random question draws from the same interview questions hub as the company pages: behavioral, coding, system design, role, and culture prompts for fifty companies. “Tell me about yourself,” weakness, and “why this role” use a short classic list, because those prompts are not stored as company questions. You can also type or paste any question an interviewer actually asked you. The transcript box is editable, which is the fallback when a browser has no speech recognition.
Chrome and Edge can transcribe with the Web Speech API while you talk. Firefox and many in-browser previews cannot. The page says so, and the timer, the seconds field, and the paste box still work. Nothing about that path sends audio anywhere. Speech recognition, when the browser has it, also runs locally in the browser’s own speech engine. This site does not receive the recording.
How to read the score
The score is out of 100 and is only a practice signal. About thirty points are the length against the range for the question type. Twenty-five are the filler rate. Twenty are pace. Fifteen are the longest pause, with partial credit when the microphone was not allowed. Ten are stutters and repeated words. A 70 with a specific tip is more useful than a 95 you cannot explain. If the transcript is empty, the score stays low on purpose.
Do not chase 100 by speaking in a monotone or by cutting every natural “so.” Chase the tip. Then record once more. Two takes of one story will change a real interview more than ten unscored reads in your head. Silent practice lies about length. The sentence that felt like twenty seconds is often fifty, and the result you were sure you said was still in your head.
When the same story is inside the window, with the action longer than the setup, and the filler list is mostly empty, stop polishing it. Put the next story on the clock. A loop of five solid answers beats one answer you have sanded for an hour. If you want a suggestion while a real interview is happening, Luna listens on the call and offers a line you can say. This timer is the practice before that, on your own machine, with no login.
FAQ
Does this upload my interview audio?
No. The timer uses the microphone only inside your browser, to measure pauses and, where the browser supports it, to transcribe. Audio is not sent to Luna and there is no account. When you close the tab, the recording is gone.
How long should a behavioral interview answer be?
About 60 to 120 seconds. Keep the situation short, spend most of the time on what you did, and end with a result. Under a minute usually skips the action. Past two minutes, cut context before you cut the result.
Which words count as filler?
um, uh, and the other hesitation sounds, plus you know, I mean, and basically, always count. like, so, well, right, actually, kind of, sort of, and literally count too, at half weight, because they are sometimes real words. The list shows each one so you can keep the ones that mean something.
What if my browser cannot transcribe?
The page tells you. Press record to run the clock, then type or paste what you said and press Score answer. Chrome and Edge support the Web Speech API. The typed path is the same score.
Do I need an account?
No. The tool is public and free. Nothing is saved on a server. If you refresh, the transcript is gone, which is the point of keeping the audio local.
What is a good words-per-minute pace?
Roughly 120 to 160 words per minute for most answers. Coding and system design can sit a little slower, closer to 110 to 150, because you are naming tradeoffs. The score uses the range for the question type you picked.