lhm.plugin.json

{
  "author": "audiojs",
  "authorUrl": "https://github.com/audiojs",
  "description": "Edit, analyze and convert audio files and the sound of videos locally, no ffmpeg: loudness and true peak, checks against ACX, podcast, Spotify and EBU R 128 specs, denoise, EQ, trim, fades, BPM, key.",
  "homepage": "https://github.com/audiojs/audio",
  "identifier": "audiojs-audio",
  "name": "audio",
  "prompts": [],
  "resources": [],
  "tags": [
    "audio",
    "podcast",
    "loudness",
    "audio-editing",
    "dsp"
  ],
  "tools": [
    {
      "name": "audio",
      "title": "Audio",
      "description": "Shell: `npx -y audio ARGS`. MCP tool `audio`: same ARGS.\n\nARGS = SOURCE [OP ARGS]… [SINK]\n- SOURCE: path (audio/video), URL, `'*.wav'` (batch), `a.mp3 + b.wav` (concat), `record 10s` (mic)\n- SINK: `stat [NAMES]` (the default) · `check SPEC` · `save PATH [192k] [24bit]` (format by extension, `-f` overwrites; .mp4 keeps a video's picture; .edl/.fcpxml/.otio: cut list) · `play`\n- Units: `1.5s` `500ms` `1:30` · `-3db` · `80hz` `2khz` · band `1khz..4khz`. Range `10s..20s`, `-5s..` (last 5s): after an op: that op; alone: the output. Options: `name:value`.\n\nOps: trim [DB] · shrink [GAP] · normalize podcast|spotify|broadcast|DB [lufs|rms] · gain DB · fade IN -OUT · crop RANGE · remove RANGE [XFADE] · pad S · speed R · stretch F (2 = twice as long) · pitch SEMI · intonation F (0 monotone, 2 twice as wide) · formant SEMI (pitch kept) · highpass HZ [ORDER] · lowpass HZ · eq HZ DB [Q] · lowshelf/highshelf HZ DB · notch HZ · mix FILE [AT] [DB] · crossfade FILE S · remix 1|2 · resample HZ · dither 16 · match REF · master REF · roomtone (fill silence) · spectral BAND [DB] RANGE · repair RANGE · split T… (save `part-{i}.wav`). Plugins by name: compressor, limiter, deesser, dehum, declick, dereverb, ducker key:VOICE… `--help`: all ops; `OP --help`: params.\n\nStats: db peak rms noisefloor loudness momentary shortterm dialog truepeak lra dc clipping silence crest correlation bpm key chords notes; `spectrum 32`: 32 bins.\n\nMeasure, edit, verify:\n1. `in.wav`: levels, BPM, clipping, DC\n2. `in.wav trim normalize podcast check podcast`: preview, nothing written\n3. `in.wav trim normalize podcast save out.mp3`: writes a new file\n4. `out.mp3 check podcast`: ✓/✗ per rule; report a fail, don't hide it\n\nSpecs: podcast (Apple, -16 LUFS) · spotify (-14) · broadcast (EBU R 128, -23) · netflix (dialog -27, ≤ -2 dBTP) · acx (audiobook). normalize holds ≤ -1 dBTP (`ceiling:-2`). ACX: `omlsa compressor -24 2.5 normalize -20 rms ceiling:-3.5 trim pad 1.5s 2s roomtone resample 44100 save book.mp3 192k`; `'ch*.wav' check acx`: a book.\n",
      "inputSchema": {
        "type": "object",
        "properties": {
          "args": {
            "type": "string",
            "description": "Arguments after `audio`, shell-quoted, absolute paths: \"/Users/me/voice.m4a trim normalize podcast save /Users/me/voice.mp3\""
          }
        },
        "required": [
          "args"
        ],
        "additionalProperties": false
      }
    }
  ],
  "version": "2.10.0"
}