Podcast transcript cleanup

A raw machine transcript is verbatim: every um, you know, restarted sentence and stutter is preserved. That's correct for legal records — and wrong for everything you actually publish. A clean podcast transcript format keeps the speaker's meaning and rhythm while cutting:

• Filler words — um, uh, like (as a tic), sort of, I mean, you know
• False starts — “the the”, “we can we can”, abandoned sentence openings
• Punctuation debris — doubled commas and orphaned punctuation left by the cuts

This tool is conservative by design: words like “like” or “actually” are only removed when they sit alone between punctuation — the discourse-marker position — never when they carry meaning. Toggle “show what was removed” to audit every cut before you copy the result.