{
  "description": "Five identical fixed-test texts for all 12 plotted acoustic models, plus original LJSpeech ground truth",
  "sample_rate": 22050,
  "audio_encoding": "mono PCM-16 WAV",
  "model_point_count": 12,
  "reference_count": 1,
  "sample_count_per_model": 5,
  "vocoder": "Shared EfficientSpeech LJSpeech HiFi-GAN v2 for all model outputs; ground truth bypasses the vocoder",
  "samples": [
    {
      "display_order": 1,
      "sample_order": 0,
      "sample_id": "LJ015-0096",
      "text": "whether representing real or fictitious shares does not appear; but they were certificates connected in some way with robson's long practiced frauds"
    },
    {
      "display_order": 2,
      "sample_order": 1,
      "sample_id": "LJ015-0187",
      "text": "redpath passed away into the outer darkness of a penal colony, where he was still living a year or two back"
    },
    {
      "display_order": 3,
      "sample_order": 2,
      "sample_id": "LJ002-0217",
      "text": "the warden of the fleet at the commencement of the eighteenth century, are too well known to need more than a passing reference."
    },
    {
      "display_order": 4,
      "sample_order": 3,
      "sample_id": "LJ037-0157",
      "text": "taken from oswald."
    },
    {
      "display_order": 5,
      "sample_order": 4,
      "sample_id": "LJ050-0062",
      "text": "(c) who express or have expressed strong or violent anti-u.s. sentiments"
    }
  ],
  "models": [
    {
      "group": "ours",
      "model_id": "es117_l1",
      "display_name": "ES117-L1",
      "plot_label": "L1",
      "parameters": 264881,
      "mean_utmos_128": 3.8034581653773785,
      "files": [
        {
          "model_id": "es117_l1",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_es117_l1_01_LJ015-0096.wav",
          "utmos": 3.413954734802246,
          "sha256": "3bb2dc82071051ab483f672fb345fb205789c17ee2a6bcb3501e0d5a998a4895",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 189696,
          "duration_seconds": 8.602993197278911
        },
        {
          "model_id": "es117_l1",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_es117_l1_02_LJ015-0187.wav",
          "utmos": 4.133847713470459,
          "sha256": "998f94bdbdfdfb8c83075660edbbb6afe98f53b4662be4ac2a36284a3341aa98",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 149248,
          "duration_seconds": 6.768616780045352
        },
        {
          "model_id": "es117_l1",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_es117_l1_03_LJ002-0217.wav",
          "utmos": 4.284744739532471,
          "sha256": "2455cd346ff137521eaa4d495d8cb26a71bbac5ca3b04a90326d594e52f2ea9f",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 155136,
          "duration_seconds": 7.035646258503402
        },
        {
          "model_id": "es117_l1",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_es117_l1_04_LJ037-0157.wav",
          "utmos": 4.026908874511719,
          "sha256": "69cbb2e6147ed1932e5675ffea349bf481a8ce939e98b27468d06b22f92f7de9",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 25344,
          "duration_seconds": 1.1493877551020408
        },
        {
          "model_id": "es117_l1",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_es117_l1_05_LJ050-0062.wav",
          "utmos": 3.243276834487915,
          "sha256": "1da81d05b393e50bd732b3cbb4a5694d58ebca962296338d4df4366421084884",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 102912,
          "duration_seconds": 4.667210884353741
        }
      ]
    },
    {
      "group": "ours",
      "model_id": "es117_l1_ssim_gvar",
      "display_name": "ES117-L1+SSIM+GVar",
      "plot_label": "L1+SSIM+GVar",
      "parameters": 264881,
      "mean_utmos_128": 4.086451526731253,
      "files": [
        {
          "model_id": "es117_l1_ssim_gvar",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_es117_l1_ssim_gvar_01_LJ015-0096.wav",
          "utmos": 4.366945743560791,
          "sha256": "a156b946590ec7ae03b5c8e1585e4ec0ad9227c79609db0c021429077b9db248",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 190976,
          "duration_seconds": 8.661043083900227
        },
        {
          "model_id": "es117_l1_ssim_gvar",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_es117_l1_ssim_gvar_02_LJ015-0187.wav",
          "utmos": 3.259036064147949,
          "sha256": "7640978d563092b2ef82a717169943c19aac46b6a11e6c24fef3d436e8f0e911",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 142080,
          "duration_seconds": 6.443537414965986
        },
        {
          "model_id": "es117_l1_ssim_gvar",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_es117_l1_ssim_gvar_03_LJ002-0217.wav",
          "utmos": 4.219937324523926,
          "sha256": "1be74779c6d13598a3accde4ce5d0f25dd0a454ea5a793b7468f4cfd990fd39f",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 148480,
          "duration_seconds": 6.733786848072563
        },
        {
          "model_id": "es117_l1_ssim_gvar",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_es117_l1_ssim_gvar_04_LJ037-0157.wav",
          "utmos": 4.167572975158691,
          "sha256": "a754ac3bf4941089a2bbd791f2b06e9922411b6d61d199ad2e5437e7229534c9",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 28160,
          "duration_seconds": 1.2770975056689342
        },
        {
          "model_id": "es117_l1_ssim_gvar",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_es117_l1_ssim_gvar_05_LJ050-0062.wav",
          "utmos": 3.6421713829040527,
          "sha256": "27f31c3eb5789dfdf809d3a04adceb270af4bb658be62d7484a579e6366fbf1e",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 101376,
          "duration_seconds": 4.597551020408163
        }
      ]
    },
    {
      "group": "es-series",
      "model_id": "efficientspeech-tiny",
      "display_name": "EfficientSpeech-Tiny",
      "plot_label": "ES-Tiny",
      "parameters": 266417,
      "mean_utmos_128": 3.5909248664975166,
      "files": [
        {
          "model_id": "efficientspeech-tiny",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_efficientspeech-tiny_01_LJ015-0096.wav",
          "utmos": 3.578371524810791,
          "sha256": "6b7dd3115d244febe51f769a0004bd484abeb5b5063df8e46b8b2d28a648eb5b",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 188160,
          "duration_seconds": 8.533333333333333
        },
        {
          "model_id": "efficientspeech-tiny",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_efficientspeech-tiny_02_LJ015-0187.wav",
          "utmos": 3.264836072921753,
          "sha256": "b4c6539e7d14178061ba81118deff32edb44364bb6bb4ca436aadc12d0a94b63",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 146688,
          "duration_seconds": 6.6525170068027215
        },
        {
          "model_id": "efficientspeech-tiny",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_efficientspeech-tiny_03_LJ002-0217.wav",
          "utmos": 4.121789455413818,
          "sha256": "ce9ac3027e4f0b08109c2c38d86376fcdc06873102658fae5015a7705c5c1c4e",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 144640,
          "duration_seconds": 6.559637188208617
        },
        {
          "model_id": "efficientspeech-tiny",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_efficientspeech-tiny_04_LJ037-0157.wav",
          "utmos": 4.0054545402526855,
          "sha256": "33b3cf1a2ae2465d8ba7fa53518cf4637311970f180deb7826b789f580e1a1cf",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 27136,
          "duration_seconds": 1.230657596371882
        },
        {
          "model_id": "efficientspeech-tiny",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_efficientspeech-tiny_05_LJ050-0062.wav",
          "utmos": 3.1130383014678955,
          "sha256": "a082d45792578139e22c52a181e160e041440c89bcfbdba7a2998e027f242fad",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 92160,
          "duration_seconds": 4.179591836734694
        }
      ]
    },
    {
      "group": "es-series",
      "model_id": "es0_l1_ssim_gvar",
      "display_name": "ES0-L1+SSIM+GVar",
      "plot_label": "ES0+L1+SSIM+GVar",
      "parameters": 266417,
      "mean_utmos_128": 3.9454641174525023,
      "files": [
        {
          "model_id": "es0_l1_ssim_gvar",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_es0_l1_ssim_gvar_01_LJ015-0096.wav",
          "utmos": 3.9319639205932617,
          "sha256": "4df3c3eebf7880c4ece19112a479fb9dad553db2dbb117585963d9f68826d629",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 189440,
          "duration_seconds": 8.591383219954649
        },
        {
          "model_id": "es0_l1_ssim_gvar",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_es0_l1_ssim_gvar_02_LJ015-0187.wav",
          "utmos": 4.270212650299072,
          "sha256": "4eba3548cecd980f1685b0933670b62221f305be6e07d631fa1f45f8a4fc2fba",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 146432,
          "duration_seconds": 6.640907029478458
        },
        {
          "model_id": "es0_l1_ssim_gvar",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_es0_l1_ssim_gvar_03_LJ002-0217.wav",
          "utmos": 4.160552978515625,
          "sha256": "0c4b5d449ba8b982b09779ffafa9dbcf0bf106747d2cc74480f6f4db40f78939",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 147968,
          "duration_seconds": 6.710566893424036
        },
        {
          "model_id": "es0_l1_ssim_gvar",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_es0_l1_ssim_gvar_04_LJ037-0157.wav",
          "utmos": 3.990647315979004,
          "sha256": "d2ae92b3d116cb00bfe7f64a04d6f600e3c6de7fb8f018540b0861b60ac66511",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 25856,
          "duration_seconds": 1.172607709750567
        },
        {
          "model_id": "es0_l1_ssim_gvar",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_es0_l1_ssim_gvar_05_LJ050-0062.wav",
          "utmos": 3.1748409271240234,
          "sha256": "d8b2d86ec75adc4cbf2b4a073afe04ac5e16532c59bc14a2cdbeb6b16111c9f6",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 93952,
          "duration_seconds": 4.260861678004535
        }
      ]
    },
    {
      "group": "es-series",
      "model_id": "efficientspeech-small",
      "display_name": "EfficientSpeech-Small",
      "plot_label": "ES-Small",
      "parameters": 952465,
      "mean_utmos_128": 3.8849045857787132,
      "files": [
        {
          "model_id": "efficientspeech-small",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_efficientspeech-small_01_LJ015-0096.wav",
          "utmos": 3.754232406616211,
          "sha256": "90e99dcfb641dbe14489bb00ed2f7324c22fa825a2b6e10851157d3177185ebc",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 184320,
          "duration_seconds": 8.359183673469389
        },
        {
          "model_id": "efficientspeech-small",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_efficientspeech-small_02_LJ015-0187.wav",
          "utmos": 3.4004082679748535,
          "sha256": "dc192e1e2f7d01fdeced4bcaf87899b052e4ab4894ddbe328b21725da9ed0688",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 147712,
          "duration_seconds": 6.698956916099773
        },
        {
          "model_id": "efficientspeech-small",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_efficientspeech-small_03_LJ002-0217.wav",
          "utmos": 4.10672664642334,
          "sha256": "735c976474f29984dd3c6abc506ef0574b49165355cc4a251c6093635c728339",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 145408,
          "duration_seconds": 6.594467120181406
        },
        {
          "model_id": "efficientspeech-small",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_efficientspeech-small_04_LJ037-0157.wav",
          "utmos": 3.418642520904541,
          "sha256": "0f730fa519c8cf81bd3f5e761ec4cb86d7fbfce3445cf00e69578151b0fde8bc",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 26368,
          "duration_seconds": 1.195827664399093
        },
        {
          "model_id": "efficientspeech-small",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_efficientspeech-small_05_LJ050-0062.wav",
          "utmos": 2.93107533454895,
          "sha256": "4c79f1f8338e4e65ac3c87d0493af6debf345d0ac6b7912c4e56bd0a0ebd6b04",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 96256,
          "duration_seconds": 4.365351473922902
        }
      ]
    },
    {
      "group": "es-series",
      "model_id": "efficientspeech-base",
      "display_name": "EfficientSpeech-Base",
      "plot_label": "ES-Base",
      "parameters": 3953489,
      "mean_utmos_128": 3.9001119192689657,
      "files": [
        {
          "model_id": "efficientspeech-base",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_efficientspeech-base_01_LJ015-0096.wav",
          "utmos": 4.18983793258667,
          "sha256": "3f80a162584dd2352b240a8b3b6e01458743c594dac993e2cdfb550fe6e09152",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 189184,
          "duration_seconds": 8.579773242630386
        },
        {
          "model_id": "efficientspeech-base",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_efficientspeech-base_02_LJ015-0187.wav",
          "utmos": 3.6385998725891113,
          "sha256": "32bc83e0f95a3d6676c090c7c05458b4a677a541d2784c3d88bce446c1831188",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 152064,
          "duration_seconds": 6.896326530612245
        },
        {
          "model_id": "efficientspeech-base",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_efficientspeech-base_03_LJ002-0217.wav",
          "utmos": 4.0536580085754395,
          "sha256": "21efa9c3fdd9d4886665d9b19b20a11a44ed3bb1c240b869867f9f01f12b25ec",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 154112,
          "duration_seconds": 6.989206349206349
        },
        {
          "model_id": "efficientspeech-base",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_efficientspeech-base_04_LJ037-0157.wav",
          "utmos": 4.047499179840088,
          "sha256": "09d5bdaa46c6231e6cce318655a0c11ddeb8a98fe7ed02139988270559080fdb",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 26112,
          "duration_seconds": 1.18421768707483
        },
        {
          "model_id": "efficientspeech-base",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_efficientspeech-base_05_LJ050-0062.wav",
          "utmos": 3.1074509620666504,
          "sha256": "40ca4726ef6029afa5f247daa5bcf95e42cb25ead0c5dd568bf28fa7b646655e",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 97536,
          "duration_seconds": 4.423401360544218
        }
      ]
    },
    {
      "group": "other-models",
      "model_id": "lightspeech",
      "display_name": "LightSpeech",
      "plot_label": "LightSpeech",
      "parameters": 1771349,
      "mean_utmos_128": 3.689273364841938,
      "files": [
        {
          "model_id": "lightspeech",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_lightspeech_01_LJ015-0096.wav",
          "utmos": 3.249260902404785,
          "sha256": "83c897501d1eed0a7ce898f5248d14ca224fe457c0f1278760e7d8732d20e96a",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 188672,
          "duration_seconds": 8.55655328798186
        },
        {
          "model_id": "lightspeech",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_lightspeech_02_LJ015-0187.wav",
          "utmos": 3.4036355018615723,
          "sha256": "d5d0e4e5a05417ad24a6df4c7f38e113d40e7f9b5cb1e70182d69ebf9ab12382",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 146688,
          "duration_seconds": 6.6525170068027215
        },
        {
          "model_id": "lightspeech",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_lightspeech_03_LJ002-0217.wav",
          "utmos": 3.7187299728393555,
          "sha256": "06350cf2dedf04af39e40469d4c806a6657bcc6bed6435f4e3aea87f2251582a",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 150016,
          "duration_seconds": 6.8034467120181406
        },
        {
          "model_id": "lightspeech",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_lightspeech_04_LJ037-0157.wav",
          "utmos": 3.7734367847442627,
          "sha256": "71af93f0446e0d7f4ee10b53aada9540a57dd08600c044962f638405753a065e",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 29184,
          "duration_seconds": 1.3235374149659864
        },
        {
          "model_id": "lightspeech",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_lightspeech_05_LJ050-0062.wav",
          "utmos": 3.537675619125366,
          "sha256": "1131b34250dea00f1ed3a2e214d943b5c7daa205c864fa8f9c0df384adad1bdb",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 106496,
          "duration_seconds": 4.829750566893424
        }
      ]
    },
    {
      "group": "other-models",
      "model_id": "speedyspeech",
      "display_name": "SpeedySpeech",
      "plot_label": "SpeedySpeech",
      "parameters": 4306001,
      "mean_utmos_128": 3.716241255402565,
      "files": [
        {
          "model_id": "speedyspeech",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_speedyspeech_01_LJ015-0096.wav",
          "utmos": 3.8448548316955566,
          "sha256": "b9d85f5424a5d9495bdbc1fd576cfdc93cfdf5d08acf4a7cea4a71da24c4c428",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 180224,
          "duration_seconds": 8.17342403628118
        },
        {
          "model_id": "speedyspeech",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_speedyspeech_02_LJ015-0187.wav",
          "utmos": 3.567488193511963,
          "sha256": "d61173662e2e66d9c2fb86ed389795fd4ef3b843f321752e87b7ce37e926d83f",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 129792,
          "duration_seconds": 5.88625850340136
        },
        {
          "model_id": "speedyspeech",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_speedyspeech_03_LJ002-0217.wav",
          "utmos": 3.80731463432312,
          "sha256": "328610151d516c05876f2fce5f597dc13de6eeea83d2c7c680120866b527c098",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 143104,
          "duration_seconds": 6.489977324263038
        },
        {
          "model_id": "speedyspeech",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_speedyspeech_04_LJ037-0157.wav",
          "utmos": 3.7903504371643066,
          "sha256": "cb9d98694dbb2228c95a1bbe68e980f265e33eee30fd60f21a460dd77645fcbf",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 26368,
          "duration_seconds": 1.195827664399093
        },
        {
          "model_id": "speedyspeech",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_speedyspeech_05_LJ050-0062.wav",
          "utmos": 3.471611976623535,
          "sha256": "6ffcc2d2cfb9dfcd905d1214563d1ae11811e0e41f74d316923fa01309fcf0de",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 104960,
          "duration_seconds": 4.760090702947846
        }
      ]
    },
    {
      "group": "other-models",
      "model_id": "grad-tts",
      "display_name": "Grad-TTS",
      "plot_label": "Grad-TTS",
      "parameters": 14835032,
      "mean_utmos_128": 4.2010154239833355,
      "files": [
        {
          "model_id": "grad-tts",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_grad-tts_01_LJ015-0096.wav",
          "utmos": 4.33389949798584,
          "sha256": "d296fc3f29d023c0c1454103e07865c55ba55e2f595e5bb6ea55d4c653e57bc1",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 184832,
          "duration_seconds": 8.382403628117913
        },
        {
          "model_id": "grad-tts",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_grad-tts_02_LJ015-0187.wav",
          "utmos": 3.921321392059326,
          "sha256": "f5f044218afaecf49d14122d6431e52533be413ff25d64ad4f41f469a8b957bd",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 160256,
          "duration_seconds": 7.267845804988662
        },
        {
          "model_id": "grad-tts",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_grad-tts_03_LJ002-0217.wav",
          "utmos": 4.451724052429199,
          "sha256": "9058ec53c158326981b94d985e5916726c745c8a62c9ebe87f1c207c0ae98329",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 150784,
          "duration_seconds": 6.8382766439909295
        },
        {
          "model_id": "grad-tts",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_grad-tts_04_LJ037-0157.wav",
          "utmos": 4.2370524406433105,
          "sha256": "6414f69b40ac07b3da90e6d6a568b2546c659a4d34f29c4607fc70ec1d3cf0ca",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 29184,
          "duration_seconds": 1.3235374149659864
        },
        {
          "model_id": "grad-tts",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_grad-tts_05_LJ050-0062.wav",
          "utmos": 4.02716064453125,
          "sha256": "3044ca99a07699981898c850383b84f9904971487ed22eac7f62e43a9382e030",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 126208,
          "duration_seconds": 5.723718820861678
        }
      ]
    },
    {
      "group": "other-models",
      "model_id": "matcha-tts",
      "display_name": "Matcha-TTS",
      "plot_label": "Matcha-TTS",
      "parameters": 18204193,
      "mean_utmos_128": 4.263990143314004,
      "files": [
        {
          "model_id": "matcha-tts",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_matcha-tts_01_LJ015-0096.wav",
          "utmos": 4.439105033874512,
          "sha256": "cf4b1affb326ac170ae6a64975426ca1f466a8e5da5b756096ca5e1fa29ff3c2",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 205568,
          "duration_seconds": 9.32281179138322
        },
        {
          "model_id": "matcha-tts",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_matcha-tts_02_LJ015-0187.wav",
          "utmos": 4.024913311004639,
          "sha256": "d7a803925ef405a4b81aa54cae77a80c8251d5adac5131d5dc6219e8cc417f3c",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 169728,
          "duration_seconds": 7.697414965986394
        },
        {
          "model_id": "matcha-tts",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_matcha-tts_03_LJ002-0217.wav",
          "utmos": 4.212703704833984,
          "sha256": "9efe456c74652ab8ead0727e7c13fb470b8da59bc68920266b02a8557893bffa",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 162560,
          "duration_seconds": 7.3723356009070296
        },
        {
          "model_id": "matcha-tts",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_matcha-tts_04_LJ037-0157.wav",
          "utmos": 4.13460111618042,
          "sha256": "0591525e2ba263f70253c1264df3b44756f8913f9996d217c290040a70f5d08d",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 29952,
          "duration_seconds": 1.3583673469387756
        },
        {
          "model_id": "matcha-tts",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_matcha-tts_05_LJ050-0062.wav",
          "utmos": 4.026965141296387,
          "sha256": "5deec7b7100e10c7762d8fcc26978fa0fee6cf41949096b36bc7fdd61eafef54",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 136192,
          "duration_seconds": 6.176507936507937
        }
      ]
    },
    {
      "group": "other-models",
      "model_id": "mixer-tts",
      "display_name": "NeMo MixerTTS",
      "plot_label": "MixerTTS",
      "parameters": 20060002,
      "mean_utmos_128": 4.087328100576997,
      "files": [
        {
          "model_id": "mixer-tts",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_mixer-tts_01_LJ015-0096.wav",
          "utmos": 3.9868881702423096,
          "sha256": "66731f492a52bda3def25a0778ec32175e7f1b8e924ffe1032b22bb4bb9e579b",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 193280,
          "duration_seconds": 8.765532879818593
        },
        {
          "model_id": "mixer-tts",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_mixer-tts_02_LJ015-0187.wav",
          "utmos": 3.3677072525024414,
          "sha256": "2faffcccafbc3711df64beab509b0e236142500b789529380705d19db1e7d455",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 153856,
          "duration_seconds": 6.977596371882086
        },
        {
          "model_id": "mixer-tts",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_mixer-tts_03_LJ002-0217.wav",
          "utmos": 4.44347620010376,
          "sha256": "8cc06937c0e7a1c134b83b46fdcc8a07316405bb97e42f2707cdb93107e6a5a5",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 156672,
          "duration_seconds": 7.1053061224489795
        },
        {
          "model_id": "mixer-tts",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_mixer-tts_04_LJ037-0157.wav",
          "utmos": 4.121338844299316,
          "sha256": "f3c21b0411df80a2d49b3d97cbdda833795de317a5be87571f3e116dd36352c6",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 29440,
          "duration_seconds": 1.3351473922902495
        },
        {
          "model_id": "mixer-tts",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_mixer-tts_05_LJ050-0062.wav",
          "utmos": 3.2659873962402344,
          "sha256": "f5df120dbb0bd515ccfb4c4a7d41b81cdf3d39fcc08a7654abcc95b8b627aed5",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 129536,
          "duration_seconds": 5.874648526077097
        }
      ]
    },
    {
      "group": "other-models",
      "model_id": "fastspeech2",
      "display_name": "FastSpeech 2",
      "plot_label": "FastSpeech2",
      "parameters": 35159361,
      "mean_utmos_128": 3.952998150140047,
      "files": [
        {
          "model_id": "fastspeech2",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_fastspeech2_01_LJ015-0096.wav",
          "utmos": 4.188133239746094,
          "sha256": "a47acd6cd131b1066c599af98fa133fc9ebe7912e119874448f5ed8e23f3cc9c",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 208384,
          "duration_seconds": 9.450521541950113
        },
        {
          "model_id": "fastspeech2",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_fastspeech2_02_LJ015-0187.wav",
          "utmos": 3.4584195613861084,
          "sha256": "9825da686fbbddda7fc6e255d3f42878a6d4237398f726ebcf889795a3ea3ef6",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 166912,
          "duration_seconds": 7.569705215419501
        },
        {
          "model_id": "fastspeech2",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_fastspeech2_03_LJ002-0217.wav",
          "utmos": 4.364602565765381,
          "sha256": "8c7210eb1e536ac05797f8410bc438944c485d8b9855b232cfd9f1e590fc55d5",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 164608,
          "duration_seconds": 7.465215419501134
        },
        {
          "model_id": "fastspeech2",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_fastspeech2_04_LJ037-0157.wav",
          "utmos": 4.212923049926758,
          "sha256": "def348b0b4e7b36760426fee39426a2d25ebafd16eb0cd04ffab144749bf0456",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 26880,
          "duration_seconds": 1.2190476190476192
        },
        {
          "model_id": "fastspeech2",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_fastspeech2_05_LJ050-0062.wav",
          "utmos": 2.98838210105896,
          "sha256": "f088f272ede3f4d3de2b5d50d5085fe4ca6cf2c19aadff33677b07e79b388f59",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 117760,
          "duration_seconds": 5.340589569160998
        }
      ]
    },
    {
      "group": "reference",
      "model_id": "ground-truth",
      "display_name": "Ground Truth",
      "plot_label": "GT original audio",
      "parameters": null,
      "mean_utmos_128": 4.366496236994863,
      "files": [
        {
          "model_id": "ground-truth",
          "sample_order": 0,
          "sample_id": "LJ015-0096",
          "path": "audio/audio_ground-truth_01_LJ015-0096.wav",
          "utmos": 4.446346282958984,
          "sha256": "573774b03d0c940f2cbcf16b0ac0d4d0366e88d9ddae691efcf9e88406934d73",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 201373,
          "duration_seconds": 9.132562358276644
        },
        {
          "model_id": "ground-truth",
          "sample_order": 1,
          "sample_id": "LJ015-0187",
          "path": "audio/audio_ground-truth_02_LJ015-0187.wav",
          "utmos": 3.9252371788024902,
          "sha256": "badfa1c616c7d529edd9037ab5ef31d5ab62cc147bbc17cc8aca797abea84154",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 166813,
          "duration_seconds": 7.565215419501134
        },
        {
          "model_id": "ground-truth",
          "sample_order": 2,
          "sample_id": "LJ002-0217",
          "path": "audio/audio_ground-truth_03_LJ002-0217.wav",
          "utmos": 4.450497150421143,
          "sha256": "f624d50656d9db70dfa54ddfa6c86c807bf51b1ec4147cd56f5ee12a017ebda6",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 165533,
          "duration_seconds": 7.507165532879819
        },
        {
          "model_id": "ground-truth",
          "sample_order": 3,
          "sample_id": "LJ037-0157",
          "path": "audio/audio_ground-truth_04_LJ037-0157.wav",
          "utmos": 4.246911525726318,
          "sha256": "3670de37720d0f04a61e84c5cd38db464ec1c5aace3e078d9f3c48b0a26d30f2",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 28573,
          "duration_seconds": 1.295827664399093
        },
        {
          "model_id": "ground-truth",
          "sample_order": 4,
          "sample_id": "LJ050-0062",
          "path": "audio/audio_ground-truth_05_LJ050-0062.wav",
          "utmos": 4.357795715332031,
          "sha256": "4bcc962a0f2e05550e30610426543a58125960f89dc96db776bb7a6730b7514a",
          "sample_rate": 22050,
          "channels": 1,
          "sample_width_bytes": 2,
          "frames": 138653,
          "duration_seconds": 6.2881179138322
        }
      ]
    }
  ]
}