Skip to content
  1. Developers
  2. Models
  3. Media analysis
Built by VidmoatAvailableFree

Speech activity and dead air

A voice activity model marks every stretch of speech, dead-air analysis finds the pauses worth cutting, and audio kind tells speech from music from ambience, so an agent can tighten a talking-head edit without listening to it.

id analyze_voice_activity, analyze_dead_air, analyze_audio_kind

What it does well

  • Pauses ranked by how much time cutting them saves.
  • Speech versus music versus noise, per stretch.

Limits

  • Measurement only; cuts are applied with edit commands.

Where to call it

Try it

MCP tools/call
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "analyze_dead_air",
    "arguments": {
      "src": "https://example.com/talk.mp4"
    }
  }
}

A test key answers with a sample and spends nothing.

Read more