cutroom-webmcp.vercel.app

findEnergyMoments

Find where the speaker's voice LIFTS — the passages carrying the most energy relative to their own baseline. This reads the actual audio, not the transcript, so it surfaces things no text search can: someone getting animated, a laugh, a moment of real feeling. The best clip in an hour is usually not the smartest sentence, it's the one with the most life in it. Use this alongside searchTranscript, not instead of it: search finds the topic, this finds the delivery. `lift` is how far above their normal level that passage sits.

Input schema

{
  "type": "object",
  "properties": {
    "limit": {
      "type": "number",
      "description": "How many moments to return (default 12)."
    },
    "minSec": {
      "type": "number",
      "description": "Shortest moment worth returning."
    },
    "maxSec": {
      "type": "number",
      "description": "Longest moment worth returning."
    }
  },
  "additionalProperties": false
}