{ "model": "nvidia/Nemotron-3-Diarization", "backend": "CPU", "torch": "2.14.0+rocm7.2", "model_load_seconds": 1.5619377790717408, "results": [ { "input": "../test-pack/opening-multispeaker.wav", "duration_seconds": 120.0, "wall_seconds": 2.0936129729961976, "realtime_factor": 0.017446774774968313, "segments": [ { "Start": 4.75, "End": 5.28, "Speaker": 0 }, { "Start": 5.75, "End": 8.74, "Speaker": 0 }, { "Start": 11.21, "End": 13.45, "Speaker": 1 }, { "Start": 13.85, "End": 14.37, "Speaker": 0 }, { "Start": 15.94, "End": 16.44, "Speaker": 2 }, { "Start": 17.34, "End": 21.73, "Speaker": 2 }, { "Start": 22.19, "End": 22.46, "Speaker": 2 }, { "Start": 23.35, "End": 24.17, "Speaker": 0 }, { "Start": 24.65, "End": 26.88, "Speaker": 2 }, { "Start": 27.89, "End": 28.3, "Speaker": 0 }, { "Start": 28.4, "End": 29.3, "Speaker": 3 }, { "Start": 30.44, "End": 31.08, "Speaker": 0 }, { "Start": 31.41, "End": 34.98, "Speaker": 1 }, { "Start": 32.1, "End": 32.32, "Speaker": 0 }, { "Start": 32.12, "End": 33.12, "Speaker": 2 }, { "Start": 35.25, "End": 37.21, "Speaker": 1 }, { "Start": 37.74, "End": 38.29, "Speaker": 1 }, { "Start": 39.45, "End": 43.71, "Speaker": 1 }, { "Start": 44.28, "End": 44.64, "Speaker": 1 }, { "Start": 45.0, "End": 47.53, "Speaker": 1 }, { "Start": 48.3, "End": 49.87, "Speaker": 1 }, { "Start": 50.42, "End": 51.67, "Speaker": 1 }, { "Start": 52.03, "End": 53.07, "Speaker": 1 }, { "Start": 53.88, "End": 59.01, "Speaker": 1 }, { "Start": 59.4, "End": 62.88, "Speaker": 1 }, { "Start": 63.11, "End": 63.46, "Speaker": 1 }, { "Start": 64.4, "End": 64.75, "Speaker": 1 }, { "Start": 64.63, "End": 65.23, "Speaker": 0 }, { "Start": 65.04, "End": 66.79, "Speaker": 1 }, { "Start": 67.09, "End": 68.83, "Speaker": 1 }, { "Start": 68.21, "End": 68.61, "Speaker": 4 }, { "Start": 69.3, "End": 70.52, "Speaker": 4 }, { "Start": 69.4, "End": 70.52, "Speaker": 2 }, { "Start": 71.1, "End": 72.3, "Speaker": 4 }, { "Start": 72.84, "End": 74.61, "Speaker": 4 }, { "Start": 75.09, "End": 75.47, "Speaker": 4 }, { "Start": 75.16, "End": 78.26, "Speaker": 0 }, { "Start": 79.57, "End": 82.32, "Speaker": 4 }, { "Start": 82.91, "End": 83.53, "Speaker": 4 }, { "Start": 84.55, "End": 86.7, "Speaker": 4 }, { "Start": 87.06, "End": 87.84, "Speaker": 4 }, { "Start": 88.52, "End": 88.87, "Speaker": 0 }, { "Start": 89.55, "End": 90.57, "Speaker": 2 }, { "Start": 89.69, "End": 90.25, "Speaker": 4 }, { "Start": 90.88, "End": 94.72, "Speaker": 2 }, { "Start": 94.95, "End": 95.19, "Speaker": 2 }, { "Start": 95.1, "End": 103.3, "Speaker": 0 }, { "Start": 95.53, "End": 96.87, "Speaker": 2 }, { "Start": 104.93, "End": 105.38, "Speaker": 0 }, { "Start": 105.95, "End": 106.29, "Speaker": 1 }, { "Start": 107.42, "End": 110.88, "Speaker": 0 }, { "Start": 111.7, "End": 111.87, "Speaker": 4 }, { "Start": 111.97, "End": 117.65, "Speaker": 5 }, { "Start": 118.03, "End": 118.89, "Speaker": 1 }, { "Start": 118.07, "End": 120.0, "Speaker": 5 } ] }, { "input": "../test-pack/mid-meeting-turns.wav", "duration_seconds": 120.0, "wall_seconds": 2.030187763972208, "realtime_factor": 0.016918231366435065, "segments": [ { "Start": 0.31, "End": 1.23, "Speaker": 0 }, { "Start": 2.78, "End": 3.2, "Speaker": 1 }, { "Start": 4.84, "End": 9.38, "Speaker": 0 }, { "Start": 9.08, "End": 9.33, "Speaker": 1 }, { "Start": 9.82, "End": 11.27, "Speaker": 1 }, { "Start": 10.09, "End": 11.29, "Speaker": 0 }, { "Start": 11.48, "End": 12.53, "Speaker": 0 }, { "Start": 11.75, "End": 13.14, "Speaker": 1 }, { "Start": 14.03, "End": 14.64, "Speaker": 0 }, { "Start": 15.14, "End": 20.37, "Speaker": 0 }, { "Start": 21.01, "End": 26.07, "Speaker": 0 }, { "Start": 26.8, "End": 27.5, "Speaker": 0 }, { "Start": 27.75, "End": 32.35, "Speaker": 0 }, { "Start": 33.18, "End": 33.59, "Speaker": 1 }, { "Start": 34.17, "End": 37.92, "Speaker": 0 }, { "Start": 38.19, "End": 40.77, "Speaker": 0 }, { "Start": 41.75, "End": 43.64, "Speaker": 0 }, { "Start": 46.55, "End": 48.07, "Speaker": 0 }, { "Start": 50.1, "End": 54.22, "Speaker": 0 }, { "Start": 56.3, "End": 56.81, "Speaker": 0 }, { "Start": 57.19, "End": 58.91, "Speaker": 0 }, { "Start": 59.22, "End": 61.66, "Speaker": 0 }, { "Start": 63.95, "End": 64.85, "Speaker": 0 }, { "Start": 65.27, "End": 67.62, "Speaker": 0 }, { "Start": 69.04, "End": 74.72, "Speaker": 2 }, { "Start": 74.74, "End": 74.75, "Speaker": 2 }, { "Start": 74.75, "End": 75.06, "Speaker": 0 }, { "Start": 75.3, "End": 76.58, "Speaker": 0 }, { "Start": 75.84, "End": 76.77, "Speaker": 2 }, { "Start": 77.18, "End": 78.23, "Speaker": 2 }, { "Start": 77.68, "End": 78.52, "Speaker": 0 }, { "Start": 79.36, "End": 79.5, "Speaker": 2 }, { "Start": 80.22, "End": 81.12, "Speaker": 0 }, { "Start": 82.73, "End": 83.65, "Speaker": 0 }, { "Start": 85.83, "End": 89.59, "Speaker": 3 }, { "Start": 89.83, "End": 91.07, "Speaker": 0 }, { "Start": 89.9, "End": 90.86, "Speaker": 3 }, { "Start": 91.42, "End": 92.27, "Speaker": 0 }, { "Start": 97.17, "End": 98.82, "Speaker": 3 }, { "Start": 99.95, "End": 101.35, "Speaker": 3 }, { "Start": 105.12, "End": 106.85, "Speaker": 4 }, { "Start": 108.79, "End": 109.57, "Speaker": 4 }, { "Start": 109.24, "End": 109.58, "Speaker": 0 }, { "Start": 111.87, "End": 117.14, "Speaker": 4 }, { "Start": 118.52, "End": 120.0, "Speaker": 0 } ] }, { "input": "../test-pack/signoff.wav", "duration_seconds": 120.0, "wall_seconds": 2.0377615810139105, "realtime_factor": 0.016981346508449253, "segments": [ { "Start": 0.0, "End": 0.02, "Speaker": 0 }, { "Start": 0.23, "End": 0.74, "Speaker": 0 }, { "Start": 0.9, "End": 1.4, "Speaker": 1 }, { "Start": 1.29, "End": 1.98, "Speaker": 0 }, { "Start": 2.09, "End": 2.67, "Speaker": 1 }, { "Start": 2.83, "End": 4.59, "Speaker": 0 }, { "Start": 5.74, "End": 6.76, "Speaker": 0 }, { "Start": 8.35, "End": 8.73, "Speaker": 0 }, { "Start": 10.26, "End": 11.08, "Speaker": 0 }, { "Start": 11.55, "End": 12.81, "Speaker": 0 }, { "Start": 18.71, "End": 19.84, "Speaker": 1 }, { "Start": 19.9, "End": 20.89, "Speaker": 0 }, { "Start": 19.95, "End": 20.25, "Speaker": 2 }, { "Start": 21.32, "End": 21.81, "Speaker": 3 }, { "Start": 22.54, "End": 24.41, "Speaker": 2 }, { "Start": 22.7, "End": 22.81, "Speaker": 0 }, { "Start": 23.36, "End": 23.59, "Speaker": 1 }, { "Start": 23.6, "End": 23.65, "Speaker": 1 }, { "Start": 24.46, "End": 24.78, "Speaker": 0 }, { "Start": 25.78, "End": 26.85, "Speaker": 0 }, { "Start": 27.43, "End": 29.61, "Speaker": 0 }, { "Start": 30.78, "End": 35.61, "Speaker": 0 }, { "Start": 35.91, "End": 40.57, "Speaker": 0 }, { "Start": 41.02, "End": 47.83, "Speaker": 0 }, { "Start": 47.99, "End": 49.65, "Speaker": 0 }, { "Start": 50.23, "End": 55.46, "Speaker": 0 }, { "Start": 55.75, "End": 56.05, "Speaker": 0 }, { "Start": 56.99, "End": 57.41, "Speaker": 3 }, { "Start": 57.24, "End": 57.75, "Speaker": 2 }, { "Start": 58.21, "End": 58.5, "Speaker": 4 }, { "Start": 58.5, "End": 58.97, "Speaker": 0 }, { "Start": 58.51, "End": 58.59, "Speaker": 4 }, { "Start": 59.1, "End": 60.18, "Speaker": 5 }, { "Start": 59.23, "End": 60.17, "Speaker": 0 }, { "Start": 60.38, "End": 60.95, "Speaker": 4 }, { "Start": 60.45, "End": 61.66, "Speaker": 2 }, { "Start": 61.85, "End": 62.56, "Speaker": 5 }, { "Start": 62.58, "End": 63.36, "Speaker": 0 }, { "Start": 62.74, "End": 63.59, "Speaker": 3 } ] }, { "input": "../test-pack/pretrim-tail.wav", "duration_seconds": 84.192, "wall_seconds": 1.4631924339337274, "realtime_factor": 0.01737923358435157, "segments": [] } ] }