Why isn’t every "like" flagged as a filler?

Question

Accepted Answer

Because "like" is off until you switch it on. Hesitations — "um", "uh", "mm" — are flagged everywhere by default; real words that double as filler, like "like", "so" and "you know", wait in the dictionary until you enable them. One click there flags every instance in the transcript.

The dictionary sorts fillers into two kinds, and they behave differently on purpose. Hesitations are noises — "uh", "um", "erm", "hmm" — that carry no meaning, so Cast flags every one of them out of the box. Discourse markers are real words: "like", "so", "right", "actually", "you know". A transcript where every "so" is flagged is a transcript you stop trusting, so those stay off until you say otherwise.

Enabling a word is all-or-nothing in the good sense: click "like" in the dictionary and every "like" in the episode is flagged — the whole transcript is re-scanned on the spot, not just the spots the initial analysis happened to notice. The Fillers count updates immediately, and so does the transcript highlighting.

The dictionary matches the word, not the grammar. "I like your show" gets flagged along with "it was, like, weird" — no detector can hear the difference reliably, so Cast does not pretend to. The escape hatch is per-instance: click a flagged word you want to keep and choose Keep. That clears the one flag without touching the dictionary.

Your dictionary travels with the project. Enable "like", add your own words, strike out a built-in — reopen the project on another machine and the same rules apply. Each speaker can also carry their own word list, so one host’s "basically" habit does not flag the other’s vocabulary.