Filler word counter and remover

There are two kinds. Pure fillers are sounds that carry no meaning at all: um, uh, er, hmm. Discourse fillers are real words used as padding: like, you know, basically, I mean. This tool counts them separately, because the pure sounds are always safe to cut while the discourse words sometimes genuinely belong in the sentence.

Ordinary conversation runs about 2 to 5 per 100 words and nobody notices. Polished speakers land under 1 per 100. Above roughly 8 per 100 it starts to distract listeners. Context matters: a casual podcast tolerates far more than a keynote.

Replace the filler with a pause. Fillers appear when your mouth outruns your thinking, and silence does the same job without the noise. Record yourself, check your rate here weekly, and slow down slightly. Measuring your speaking speed helps too, since fillers cluster when you rush.

Yes. OpenFlow has an optional on-device cleanup pass that strips fillers, stutters, and false starts as you dictate, so the text that lands at your cursor is already clean. It runs locally with Apple Intelligence, so nothing is uploaded.