Hi Product Hunt. I built Loorq because I recorded myself talking and did not like what I heard. Not the content, the delivery. The ums were bad enough, but the real problem was the "like" I dropped every time I restarted a sentence, and a "you know" I had genuinely never noticed myself saying. So the first version just counted them. That version was useless, and it took me a while to understand why: it was wrong too often. It flagged "I like what I see". It flagged "so much better". Once a tool accuses you of something you did not do, you stop believing every number it shows you, including the correct ones. That turned into the rule the whole app is built around now: never flag a real word. When a phrase is genuinely ambiguous, Loorq undercounts on purpose. Your count is a floor, not a guess. Getting there meant writing an acoustic model for the sounds and a context engine for the words, and then throwing away rules that scored well but felt wrong when I listened back to my own takes. The final test was never the benchmark. It was playing back a recording and agreeing with what it caught. Everything runs on your Mac. No cloud, no account, no audio written to disk or uploaded. Meetings store counts, not words, so your calls are never transcribed. It is dictation first: press, speak, and clean text lands at your cursor in any app. The coaching is what happens underneath. Three day trial with everything on, then a one time purchase. I would love to hear what your own crutch word turns out to be. Mine was "like", and I fought the detector about it for a week before admitting it was right.

Loorq · The speech coach on your Mac

Every um, counted. Every word, yours.

Every um, counted. Every word, yours.