{
  "amiSingleMeetingDiagnostic": {
    "audioSeconds": 1049.3546875,
    "endToEnd": {
      "peakRssMB": 944.375,
      "throughputXRT": 23.988282918384776
    },
    "exactMatchToUpstream": true,
    "exclusive": {
      "collar025": {
        "derPercent": 22.58107201156437,
        "jerPercent": 24.12583550793453
      },
      "strict": {
        "derPercent": 34.977370101045615,
        "jerPercent": 34.678204358087
      }
    },
    "oracleSpeakerCount": 4,
    "recording": "ami-ES2004a",
    "regular": {
      "collar025": {
        "derPercent": 23.00216585375451,
        "jerPercent": 23.619490786739856
      },
      "strict": {
        "derPercent": 35.22383954393369,
        "jerPercent": 33.206537162384095
      }
    }
  },
  "method": {
    "collarsSeconds": [
      0.0,
      0.25
    ],
    "hostProcessing": "official pyannote.audio Community-1 powerset, speaker counting, VBx, and reconstruction",
    "networkDuringInference": false,
    "neuralStages": "compiled Core ML",
    "overlapScored": true,
    "telemetry": false
  },
  "runtime": {
    "computeUnits": "cpu-and-neural-engine",
    "coremltoolsVersion": "8.3.0",
    "cpu": "Apple M5 Pro",
    "machine": "arm64",
    "platform": "macOS-26.5.2-arm64-arm-64bit",
    "pyannoteAudioVersion": "4.0.7"
  },
  "schemaVersion": 1,
  "sourceModel": "pyannote/speaker-diarization-community-1",
  "sourceRevision": "3533c8cf8e369892e6b79ff1bf80f7b0286a54ee",
  "voxconverseFiveFileSubset": {
    "createdAt": "2026-07-14T14:50:36.914392+00:00",
    "endToEnd": {
      "peakRssMB": 863.421875,
      "throughputXRT": 25.001608024686572
    },
    "exactMatchToUpstream": true,
    "exclusive": {
      "collar025": {
        "derPercent": 7.8330999608667895,
        "jerPercent": 21.825168113918547
      },
      "strict": {
        "derPercent": 10.309137316831276,
        "jerPercent": 23.61927726234164
      }
    },
    "files": [
      {
        "audioSeconds": 148.32,
        "detectedSpeakers": 3,
        "exclusive": {
          "collar025": {
            "confusionSeconds": 0.6615937500000002,
            "derPercent": 3.2672644439454674,
            "falseAlarmSeconds": 0.7983437499999628,
            "jerPercent": 3.5324457444260835,
            "missSeconds": 2.9057812499999702,
            "referenceSpeakerSeconds": 133.62
          },
          "strict": {
            "confusionSeconds": 0.7865937500000002,
            "derPercent": 4.908588280338341,
            "falseAlarmSeconds": 2.1820312499999526,
            "jerPercent": 5.841663331740058,
            "missSeconds": 3.76203124999998,
            "referenceSpeakerSeconds": 137.12
          }
        },
        "id": "vox-ampme",
        "referenceSpeakers": 3,
        "regular": {
          "collar025": {
            "confusionSeconds": 0.6615937500000002,
            "derPercent": 4.163930362221178,
            "falseAlarmSeconds": 1.9964687499999674,
            "jerPercent": 9.504901132865212,
            "missSeconds": 2.9057812499999702,
            "referenceSpeakerSeconds": 133.62
          },
          "strict": {
            "confusionSeconds": 0.7865937500000002,
            "derPercent": 5.782366722578717,
            "falseAlarmSeconds": 3.380156249999957,
            "jerPercent": 11.412048028876494,
            "missSeconds": 3.76203124999998,
            "referenceSpeakerSeconds": 137.12
          }
        }
      },
      {
        "audioSeconds": 194.136,
        "detectedSpeakers": 3,
        "exclusive": {
          "collar025": {
            "confusionSeconds": 2.09715625000001,
            "derPercent": 13.524134267851926,
            "falseAlarmSeconds": 0.0,
            "jerPercent": 33.47446402551495,
            "missSeconds": 25.592156250000023,
            "referenceSpeakerSeconds": 204.74000000000004
          },
          "strict": {
            "confusionSeconds": 3.131187500000001,
            "derPercent": 15.926859919571058,
            "falseAlarmSeconds": 0.23187500000000938,
            "jerPercent": 35.194641726815576,
            "missSeconds": 32.281250000000014,
            "referenceSpeakerSeconds": 223.8
          }
        },
        "id": "vox-azisu",
        "referenceSpeakers": 4,
        "regular": {
          "collar025": {
            "confusionSeconds": 5.680687500000005,
            "derPercent": 7.149021319722563,
            "falseAlarmSeconds": 1.4464687499999922,
            "jerPercent": 30.93591229987478,
            "missSeconds": 7.509749999999983,
            "referenceSpeakerSeconds": 204.74000000000004
          },
          "strict": {
            "confusionSeconds": 6.850281249999995,
            "derPercent": 9.529923480786389,
            "falseAlarmSeconds": 3.7775937499999817,
            "jerPercent": 32.515982955223386,
            "missSeconds": 10.700093749999965,
            "referenceSpeakerSeconds": 223.80000000000004
          }
        }
      },
      {
        "audioSeconds": 259.160875,
        "detectedSpeakers": 2,
        "exclusive": {
          "collar025": {
            "confusionSeconds": 0.14215624999999932,
            "derPercent": 0.7351801529790701,
            "falseAlarmSeconds": 0.4240312500000061,
            "jerPercent": 0.9684476301170961,
            "missSeconds": 1.2600000000000051,
            "referenceSpeakerSeconds": 248.40000000000003
          },
          "strict": {
            "confusionSeconds": 0.6398749999999751,
            "derPercent": 2.156736760124583,
            "falseAlarmSeconds": 1.8436874999999742,
            "jerPercent": 2.7212320527959633,
            "missSeconds": 3.05493749999998,
            "referenceSpeakerSeconds": 256.8
          }
        },
        "id": "vox-blwmj",
        "referenceSpeakers": 2,
        "regular": {
          "collar025": {
            "confusionSeconds": 0.14215624999999932,
            "derPercent": 0.472473832528178,
            "falseAlarmSeconds": 0.9199374999999963,
            "jerPercent": 0.6409231193258367,
            "missSeconds": 0.11153124999999875,
            "referenceSpeakerSeconds": 248.40000000000003
          },
          "strict": {
            "confusionSeconds": 0.5524374999999964,
            "derPercent": 1.8084355529594591,
            "falseAlarmSeconds": 2.9842499999999186,
            "jerPercent": 2.2384443742178157,
            "missSeconds": 1.1073749999999762,
            "referenceSpeakerSeconds": 256.8
          }
        }
      },
      {
        "audioSeconds": 300.264,
        "detectedSpeakers": 2,
        "exclusive": {
          "collar025": {
            "confusionSeconds": 2.691531249999919,
            "derPercent": 10.18159170067545,
            "falseAlarmSeconds": 1.601812500000004,
            "jerPercent": 10.631003936851133,
            "missSeconds": 24.646812499999974,
            "referenceSpeakerSeconds": 284.24
          },
          "strict": {
            "confusionSeconds": 5.173843749999847,
            "derPercent": 13.722947492313498,
            "falseAlarmSeconds": 5.943906249999916,
            "jerPercent": 14.60187647291351,
            "missSeconds": 31.730781249999907,
            "referenceSpeakerSeconds": 312.24
          }
        },
        "id": "vox-crixb",
        "referenceSpeakers": 2,
        "regular": {
          "collar025": {
            "confusionSeconds": 2.5665937499999885,
            "derPercent": 3.1307829650998653,
            "falseAlarmSeconds": 2.35656249999992,
            "jerPercent": 4.0015791666816805,
            "missSeconds": 3.9757812499999474,
            "referenceSpeakerSeconds": 284.24
          },
          "strict": {
            "confusionSeconds": 4.162781249999981,
            "derPercent": 6.574139684217199,
            "falseAlarmSeconds": 8.856218749999943,
            "jerPercent": 7.647274233385495,
            "missSeconds": 7.508093749999858,
            "referenceSpeakerSeconds": 312.24
          }
        }
      },
      {
        "audioSeconds": 155.6114375,
        "detectedSpeakers": 3,
        "exclusive": {
          "collar025": {
            "confusionSeconds": 6.980000000000001,
            "derPercent": 11.408968739662592,
            "falseAlarmSeconds": 0.25465625000000386,
            "jerPercent": 39.92085630987593,
            "missSeconds": 10.010000000000005,
            "referenceSpeakerSeconds": 151.15
          },
          "strict": {
            "confusionSeconds": 7.859874999999977,
            "derPercent": 13.51821608040199,
            "falseAlarmSeconds": 0.9405624999999989,
            "jerPercent": 40.3348462453058,
            "missSeconds": 12.720562499999994,
            "referenceSpeakerSeconds": 159.20000000000002
          }
        },
        "id": "vox-cwryz",
        "referenceSpeakers": 4,
        "regular": {
          "collar025": {
            "confusionSeconds": 6.980000000000001,
            "derPercent": 11.408968739662592,
            "falseAlarmSeconds": 0.25465625000000386,
            "jerPercent": 39.92085630987593,
            "missSeconds": 10.010000000000005,
            "referenceSpeakerSeconds": 151.15
          },
          "strict": {
            "confusionSeconds": 7.859874999999977,
            "derPercent": 13.613614949748722,
            "falseAlarmSeconds": 1.0924374999999957,
            "jerPercent": 40.50417077688739,
            "missSeconds": 12.720562499999994,
            "referenceSpeakerSeconds": 159.20000000000002
          }
        }
      }
    ],
    "regular": {
      "collar025": {
        "derPercent": 4.648825392554887,
        "jerPercent": 21.415118827307566
      },
      "strict": {
        "derPercent": 6.987107610452046,
        "jerPercent": 23.072546415351944
      }
    },
    "speakerCount": {
      "accuracy": 0.6,
      "exactFiles": 3,
      "scoredFiles": 5
    },
    "stageTiming": {
      "embedding": {
        "calls": 1016,
        "medianMsPerWindow": 33.23672899932717,
        "p95MsPerWindow": 35.55754149965651,
        "throughputXRT": 297.19853527611525,
        "totalSeconds": 34.18590199497703,
        "windows": 1016
      },
      "segmentation": {
        "calls": 1016,
        "medianMsPerWindow": 7.43545799923595,
        "p95MsPerWindow": 7.705229250859702,
        "throughputXRT": 1338.3739525810397,
        "totalSeconds": 7.59130135520536,
        "windows": 1016
      }
    },
    "totalAudioSeconds": 1057.4923125
  }
}