Guide

Prompts for cleaning up raw, unscripted footage

Turn a rambly 20-minute take into something watchable.

Raw talking-head or vlog footage is usually the messiest starting point — lots of "um", restarts, and long pauses. This is exactly what transcript-driven cutting is built for.

  • "Transcribe this and cut every filler word and awkward pause, but keep the actual content."
  • "I restarted that sentence three times around the two-minute mark — cut the false starts and keep only the clean version."
  • "This has a long dead silence in the middle where I was checking my notes — cut it."
  • "Keep the charming mistakes, just tighten the pacing overall."

Because cuts are transcript-driven (not just silence-detection), it holds up on real-world audio with background noise — the words tell the agent what to keep, not just the waveform.