"Um" is not a vocabulary problem, and it is mostly not nerves. It is the sound of planning. Speaking and composing happen at the same time, and when composition falls behind, the mouth holds the floor with noise until the next clause arrives. In conversation the noise earns its keep: an "um" tells the other person you have not finished, so they do not start. That is why fluent, confident people say it constantly — it is a turn-keeping signal, not a flaw. On camera, the logic collapses. There is nobody to hold the floor against; a recording cannot interrupt you. The sound that does useful work across a table does none in a monologue, and the viewer just has to sit through it. This matters because it changes the fix. You do not need to become a fluent speaker. You already are one. You need to replace the placeholder with the thing it stands for.
What it stands for is a pause. When the next sentence is not ready, the honest move is to stop — silently — and let it finish forming before you say it. This feels terrible the first few times: from inside, two seconds of silence feels like dead air, like a failing take. From outside it reads as thought. A viewer does not experience a speaker who pauses as someone who is lost; they experience someone deciding what to say, which is exactly what is happening. So the practice is small and specific. Finish the sentence. Close your mouth. Make no sound until the next sentence exists. Slowing down helps, because fillers cluster wherever speech outruns planning; knowing the script helps for the same reason; and pausing at full stops instead of rolling through them gives the habit somewhere legitimate to go. But most of the work is done by permission — deciding that silence is allowed.
The habit is genuinely hard to hear from inside, which is why counting beats willpower. Talk2Camera’s delivery metrics count filler words per take, measured on the device from the recording itself, alongside your pace and how much of the script you covered. The first number is usually a surprise; people guess three and read eleven. After that it becomes a dial you can watch move: record, look, record again. The count falls not because you are suppressing fillers in the moment — trying to do that while presenting splits your attention and makes the delivery worse — but because the number makes the habit audible. Once you can hear an "um" coming, you can pause instead. Most people cut the count sharply within a handful of takes, and the residue stops mattering.
The trade you have made — silence instead of noise — has one cost left, and the edit removes it. A filler has to be lived with; a pause is just dead air, and dead air cuts cleanly. Talk2Camera finds the long silences in a take and proposes each one as a cut you review one at a time, with undo — nothing is removed without your agreement. Take out the gaps where you were composing the next sentence; the seam is invisible, because nothing was interrupted. But do not take all of them. A pause before the sentence that matters is doing the same work on camera that it does in a room, and a video with every silence removed is exhausting to watch. The finished video ends up tighter than a fluent-sounding take could have been — because "um" never cuts cleanly, and silence always does.