How to stop saying "um"
Almost everyone says "um". The aim is not to reach zero, which can make speech sound stiff, but to bring fillers down to the point where listeners stop noticing them.
By Mathieu Dutour Sikirić · Updated 2026-10-02
Why we say "um"
Filler words are the sound of planning. When the next phrase is not ready yet, the mouth keeps the turn by making a noise: "um", "uh", "so", "like", "you know", "I mean". They are more frequent when the topic is unfamiliar, when you are nervous, and when you feel you must not leave a silence.
That last point is the key. The silence feels much longer to the speaker than to the listener. Most fillers disappear once you accept a short pause instead.
How many is too many?
Voxa counts fillers per minute so that recordings of different lengths compare fairly. As a rough guide:
- Under about 3 per minute: barely noticeable.
- Around 5 to 9 per minute: noticeable, and starting to undercut what you say.
- Above about 10 per minute: distracting. Listeners start to hear the fillers instead of the content.
In the scoring, fewer than about one per minute counts as fully fluent, and fillers are one of the main factors in confidence, clarity and professionalism.
Five techniques that work
- Pause instead. When you feel a filler coming, close your mouth. A closed mouth cannot say "um". The resulting pause sounds deliberate.
- Know your next point. Most fillers come at the start of a new idea. An outline of three points, even in your head, removes many of them.
- Use shorter sentences. Long sentences force you to plan while speaking; short ones give you a natural place to breathe and think.
- Slow down slightly. Speed and fillers often go together: the faster you talk, the sooner you run out of planned words.
- Practise with a recording. You cannot fix what you do not hear. Many people are surprised by which filler they actually use most.
A one-week drill
Pick one topic you know well. Each day, record one minute on it and note your fillers per minute. Keep the topic, the room and the microphone the same. Many people find the count drops once they start to notice the moment just before a filler.
What a filler count gets wrong
Automatic filler counts are indicative, not exact. Words like "like", "so" and "actually" have perfectly legitimate uses ("I like this approach") and are counted anyway, and very quiet hesitation sounds may be missed. Look at the trend across your own recordings rather than treating a single number as precise.
Count your own fillers
Voxa counts your fillers per minute when transcription is turned on. The transcription runs inside your browser, so your recording is never uploaded.