[
  {
    "phase": "sample",
    "model": "kokoro",
    "voice": "af_heart",
    "passage": "conversation",
    "text": "Hey Aku. I found a few voices worth listening to. Take your time; I care more about whether this sounds natural than whether it wins a benchmark. Does the rhythm feel right, or does it sound like someone reading a script?",
    "file": "raw/kokoro-af_heart-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 13.4,
    "generation_seconds": 2.95328495814465,
    "first_chunk_seconds": 2.9520851250272244,
    "rtf": 0.22039439986154102,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.6705461740493774,
    "rms": 0.05148355072843268,
    "peak_rss_bytes": 1450590208,
    "sha256": "35366da7adff2af0ccf9b3d0b6d3aeef156365483e147baff536e508eaeecce2",
    "matched_file": "matched/kokoro-af_heart-conversation.wav",
    "input_lufs": -25.38,
    "gain_db": -0.620000000000001,
    "matched_lufs": -26.0,
    "matched_sha256": "ef69eb70eaabd4e7573b87b48067fa1d2b7a5370260bc54d67be5850bb568e50"
  },
  {
    "phase": "sample",
    "model": "kokoro",
    "voice": "af_heart",
    "passage": "expression",
    "text": "Wait, you actually got it working? Oh, that's brilliant! I was starting to think this little machine had given up on us. All right. Take a breath. We have time to get this right.",
    "file": "raw/kokoro-af_heart-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 10.425,
    "generation_seconds": 2.274479999905452,
    "first_chunk_seconds": 2.2743354169651866,
    "rtf": 0.21817553955927596,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.6082119345664978,
    "rms": 0.052477658881336244,
    "peak_rss_bytes": 1694154752,
    "sha256": "7fafd0a6b39cd0d832128684cd9539277c69f5018a98257f2489ec9f05cf9fce",
    "matched_file": "matched/kokoro-af_heart-expression.wav",
    "input_lufs": -25.5,
    "gain_db": -0.5,
    "matched_lufs": -26.0,
    "matched_sha256": "4ae8328f88683df03ef16119681471e9a0619dd0b9dd5327f471f64fd91f2194"
  },
  {
    "phase": "sample",
    "model": "kokoro",
    "voice": "af_heart",
    "passage": "technical",
    "text": "The backup finished at 9:42 p.m. We copied 3.7 gigabytes, verified the SHA-256 checksum, and left the GPU alone. The address is 192.168.1.3. Please don't reboot the machine yet.",
    "file": "raw/kokoro-af_heart-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 16.35,
    "generation_seconds": 3.805502749979496,
    "first_chunk_seconds": 3.803753708023578,
    "rtf": 0.2327524617724462,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.5111823678016663,
    "rms": 0.053028310411030295,
    "peak_rss_bytes": 1762656256,
    "sha256": "89eec7e3af07c96a2883d0154509b77768d987bb0c262dd0110810a51c6738c5",
    "matched_file": "matched/kokoro-af_heart-technical.wav",
    "input_lufs": -25.18,
    "gain_db": -0.8200000000000003,
    "matched_lufs": -26.0,
    "matched_sha256": "689adaea14f2486ba36e59549177d0ade25638228033db6aa7b96c9e877fea48"
  },
  {
    "phase": "sample",
    "model": "kokoro",
    "voice": "af_bella",
    "passage": "conversation",
    "text": "Hey Aku. I found a few voices worth listening to. Take your time; I care more about whether this sounds natural than whether it wins a benchmark. Does the rhythm feel right, or does it sound like someone reading a script?",
    "file": "raw/kokoro-af_bella-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 13.975,
    "generation_seconds": 3.2419455419294536,
    "first_chunk_seconds": 3.24154791678302,
    "rtf": 0.23198179190908436,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.5507895946502686,
    "rms": 0.058898716763816,
    "peak_rss_bytes": 1991245824,
    "sha256": "bd693e634df0aecde2607d91cb3488b184c44128c8fe5e646a3ec4ad2a8bd1ac",
    "matched_file": "matched/kokoro-af_bella-conversation.wav",
    "input_lufs": -24.29,
    "gain_db": -1.7100000000000009,
    "matched_lufs": -26.0,
    "matched_sha256": "d2ca579b50d0a69af0eaeadaee0d00be3bcc0b88598b29c567357574bb66d422"
  },
  {
    "phase": "sample",
    "model": "kokoro",
    "voice": "af_bella",
    "passage": "expression",
    "text": "Wait, you actually got it working? Oh, that's brilliant! I was starting to think this little machine had given up on us. All right. Take a breath. We have time to get this right.",
    "file": "raw/kokoro-af_bella-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 10.925,
    "generation_seconds": 2.3252634168602526,
    "first_chunk_seconds": 2.325118332868442,
    "rtf": 0.21283875669201396,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.4668470621109009,
    "rms": 0.06059166797454231,
    "peak_rss_bytes": 2195701760,
    "sha256": "c912c19bf374b8711d34f55ce1abcd337d64d5d2fe1fe6a68ac3b8b1a0d98c0f",
    "matched_file": "matched/kokoro-af_bella-expression.wav",
    "input_lufs": -24.24,
    "gain_db": -1.7600000000000016,
    "matched_lufs": -26.0,
    "matched_sha256": "4815455e549a185a3b88f943ba5bdd2441db6c98d558a367750d4626ce364af0"
  },
  {
    "phase": "sample",
    "model": "kokoro",
    "voice": "af_bella",
    "passage": "technical",
    "text": "The backup finished at 9:42 p.m. We copied 3.7 gigabytes, verified the SHA-256 checksum, and left the GPU alone. The address is 192.168.1.3. Please don't reboot the machine yet.",
    "file": "raw/kokoro-af_bella-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 16.975,
    "generation_seconds": 4.111789459129795,
    "first_chunk_seconds": 4.11111633409746,
    "rtf": 0.24222618315933991,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.508679211139679,
    "rms": 0.06075397332460566,
    "peak_rss_bytes": 2195701760,
    "sha256": "d5f8ee5ca81eeac2a9908bbbcf6aad3357dfbff78f8a8c70eb47b47770008418",
    "matched_file": "matched/kokoro-af_bella-technical.wav",
    "input_lufs": -24.2,
    "gain_db": -1.8000000000000007,
    "matched_lufs": -26.0,
    "matched_sha256": "4a7e76b3cfb2ff9f552cadf437dadcddd240c379d8da5c52d81bffb3240dcbbb"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "azelma",
    "passage": "conversation",
    "text": "Hey Aku. I found a few voices worth listening to. Take your time; I care more about whether this sounds natural than whether it wins a benchmark. Does the rhythm feel right, or does it sound like someone reading a script?",
    "file": "raw/nano-azelma-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 12.44,
    "generation_seconds": 9.120418041013181,
    "first_chunk_seconds": null,
    "rtf": 0.7331525756441464,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.46450430154800415,
    "rms": 0.04954095234521797,
    "peak_rss_bytes": 2233614336,
    "sha256": "fbc45608a570f543ff867a1d20d1d4d0828c908a96df441fb1332044cc6a2abc",
    "matched_file": "matched/nano-azelma-conversation.wav",
    "input_lufs": -26.21,
    "gain_db": 0.21000000000000085,
    "matched_lufs": -26.0,
    "matched_sha256": "cd096b72544ce20a738d876c8f140c154198fcfbea6aff4a79762f6c474833f6"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "azelma",
    "passage": "expression",
    "text": "Wait, you actually got it working? Oh, that's brilliant! I was starting to think this little machine had given up on us. All right. Take a breath. We have time to get this right.",
    "file": "raw/nano-azelma-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 11.44,
    "generation_seconds": 8.131349290953949,
    "first_chunk_seconds": null,
    "rtf": 0.7107822806777928,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.4049055278301239,
    "rms": 0.04633872810536041,
    "peak_rss_bytes": 2233614336,
    "sha256": "75b110900d48f223267169101b068404043aaf8ff7793dccd1c685dbe856905b",
    "matched_file": "matched/nano-azelma-expression.wav",
    "input_lufs": -26.53,
    "gain_db": 0.5300000000000011,
    "matched_lufs": -26.0,
    "matched_sha256": "1251dee5b37abc41723519715bf48c7b2149a87238e926979b43208fa4de1b6e"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "azelma",
    "passage": "technical",
    "text": "The backup finished at 9:42 p.m. We copied 3.7 gigabytes, verified the SHA-256 checksum, and left the GPU alone. The address is 192.168.1.3. Please don't reboot the machine yet.",
    "file": "raw/nano-azelma-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 16.72,
    "generation_seconds": 12.671564165968448,
    "first_chunk_seconds": null,
    "rtf": 0.757868670213424,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.3977297246456146,
    "rms": 0.04600071699768357,
    "peak_rss_bytes": 2233614336,
    "sha256": "be41a3b5d9b07af851239185dde9bc990485a779a789f44bfa1fcfc6a7cf0256",
    "matched_file": "matched/nano-azelma-technical.wav",
    "input_lufs": -26.81,
    "gain_db": 0.8099999999999987,
    "matched_lufs": -26.0,
    "matched_sha256": "3173b8e069408eaef5f68488ce5b36dd71b568a96902e6d827e6029b84f44c54"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "azelma",
    "passage": "vocal-events",
    "text": "Well, that was subtle. [chuckle] The desktop survived, and nobody had to buy a new graphics card. [sigh] I could get used to this.",
    "file": "raw/nano-azelma-vocal-events.wav",
    "sample_rate": 24000,
    "duration_seconds": 8.6,
    "generation_seconds": 6.4306144579313695,
    "first_chunk_seconds": null,
    "rtf": 0.7477458672013221,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.4217293858528137,
    "rms": 0.04359303373572576,
    "peak_rss_bytes": 2233614336,
    "sha256": "c7dec1e6e09f505e5fcb87c42043185254bcf1899cbd0425b2c6d49765f97e7f",
    "matched_file": "matched/nano-azelma-vocal-events.wav",
    "input_lufs": -26.84,
    "gain_db": 0.8399999999999999,
    "matched_lufs": -26.0,
    "matched_sha256": "66e98c8a837a31ec159d17a6e8375b4e144be9e7629c1a4c2308a732a152e156"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "eponine",
    "passage": "conversation",
    "text": "Hey Aku. I found a few voices worth listening to. Take your time; I care more about whether this sounds natural than whether it wins a benchmark. Does the rhythm feel right, or does it sound like someone reading a script?",
    "file": "raw/nano-eponine-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 14.12,
    "generation_seconds": 10.604404666926712,
    "first_chunk_seconds": null,
    "rtf": 0.7510201605472175,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.3839244246482849,
    "rms": 0.057901263531675774,
    "peak_rss_bytes": 2688090112,
    "sha256": "a598e487824d47bc5d5f9b0d33dfb498ded084dd20dc19a2f6b8d15244397d02",
    "matched_file": "matched/nano-eponine-conversation.wav",
    "input_lufs": -24.74,
    "gain_db": -1.2600000000000016,
    "matched_lufs": -26.0,
    "matched_sha256": "5b1eb1d3d7d9094c082df2e4c6bcf6ace3ff4fb200e43514130a79360cf460a9"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "eponine",
    "passage": "expression",
    "text": "Wait, you actually got it working? Oh, that's brilliant! I was starting to think this little machine had given up on us. All right. Take a breath. We have time to get this right.",
    "file": "raw/nano-eponine-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 10.44,
    "generation_seconds": 7.3946995828300714,
    "first_chunk_seconds": null,
    "rtf": 0.708304557742344,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.5286828875541687,
    "rms": 0.07256756085622434,
    "peak_rss_bytes": 2688090112,
    "sha256": "1959bfb52035497603db398941e9797e581b2e16bf12dd70fa2113a5705a6f2e",
    "matched_file": "matched/nano-eponine-expression.wav",
    "input_lufs": -22.62,
    "gain_db": -3.379999999999999,
    "matched_lufs": -26.0,
    "matched_sha256": "b41e504ef5b72bb3c7572aa94295ba817a63b4e2457537eab6b54372bc8f6310"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "eponine",
    "passage": "technical",
    "text": "The backup finished at 9:42 p.m. We copied 3.7 gigabytes, verified the SHA-256 checksum, and left the GPU alone. The address is 192.168.1.3. Please don't reboot the machine yet.",
    "file": "raw/nano-eponine-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 15.8,
    "generation_seconds": 11.254350333008915,
    "first_chunk_seconds": null,
    "rtf": 0.712300653987906,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.5146812796592712,
    "rms": 0.05452775363597984,
    "peak_rss_bytes": 2688090112,
    "sha256": "188d705e765bcacd72173fdee431af2f26ade86d1e64a4bbe77efede43d4d1a5",
    "matched_file": "matched/nano-eponine-technical.wav",
    "input_lufs": -25.42,
    "gain_db": -0.5799999999999983,
    "matched_lufs": -26.0,
    "matched_sha256": "c6f6110acf8d1bdf006b9329e9dccbd50ca5641189bb1c361e473937e1949114"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "eponine",
    "passage": "vocal-events",
    "text": "Well, that was subtle. [chuckle] The desktop survived, and nobody had to buy a new graphics card. [sigh] I could get used to this.",
    "file": "raw/nano-eponine-vocal-events.wav",
    "sample_rate": 24000,
    "duration_seconds": 8.52,
    "generation_seconds": 6.495909999823198,
    "first_chunk_seconds": null,
    "rtf": 0.7624307511529576,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.5577936172485352,
    "rms": 0.05713508377787298,
    "peak_rss_bytes": 2688090112,
    "sha256": "035d05d52b7e6dbb4681cc9a695c7f28905f9db08a68ee6970324bbacbabf578",
    "matched_file": "matched/nano-eponine-vocal-events.wav",
    "input_lufs": -24.71,
    "gain_db": -1.2899999999999991,
    "matched_lufs": -26.0,
    "matched_sha256": "0eebfdea82b5fbf21f4a63e664a9308936867da6962158225376879bfd3bc13f"
  },
  {
    "phase": "sample",
    "model": "pocket",
    "voice": "azelma",
    "passage": "conversation",
    "text": "Hey Aku. I found a few voices worth listening to. Take your time; I care more about whether this sounds natural than whether it wins a benchmark. Does the rhythm feel right, or does it sound like someone reading a script?",
    "file": "raw/pocket-azelma-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 14.4,
    "generation_seconds": 3.359052333049476,
    "first_chunk_seconds": 0.1173213329166174,
    "rtf": 0.23326752312843585,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.7268035411834717,
    "rms": 0.09071936400878981,
    "peak_rss_bytes": 968589312,
    "sha256": "7f6de8350ab30e9611391c56fdf7c57faac4fa87c2d20bdcbfae03743991b90a",
    "matched_file": "matched/pocket-azelma-conversation.wav",
    "input_lufs": -20.4,
    "gain_db": -5.600000000000001,
    "matched_lufs": -26.0,
    "matched_sha256": "3b831eb0b0da98e05fb826829252864196165b6e73c34c78c1537e49cafc62a1"
  },
  {
    "phase": "sample",
    "model": "pocket",
    "voice": "azelma",
    "passage": "expression",
    "text": "Wait, you actually got it working? Oh, that's brilliant! I was starting to think this little machine had given up on us. All right. Take a breath. We have time to get this right.",
    "file": "raw/pocket-azelma-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 13.92,
    "generation_seconds": 3.2287692499812692,
    "first_chunk_seconds": 0.06960495794191957,
    "rtf": 0.231951813935436,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.504514217376709,
    "rms": 0.07862092260420656,
    "peak_rss_bytes": 968589312,
    "sha256": "fed8a89f25992b1c0e073f47d8af44ded9b6159867eb4ac7ca59f4de4573eb40",
    "matched_file": "matched/pocket-azelma-expression.wav",
    "input_lufs": -20.93,
    "gain_db": -5.07,
    "matched_lufs": -26.0,
    "matched_sha256": "88d657b600fd87d97a32721f94f47951f4af5b1aab191bd5d103336e49507f0b"
  },
  {
    "phase": "sample",
    "model": "pocket",
    "voice": "azelma",
    "passage": "technical",
    "text": "The backup finished at 9:42 p.m. We copied 3.7 gigabytes, verified the SHA-256 checksum, and left the GPU alone. The address is 192.168.1.3. Please don't reboot the machine yet.",
    "file": "raw/pocket-azelma-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 19.2,
    "generation_seconds": 4.3870860419701785,
    "first_chunk_seconds": 0.05527379200793803,
    "rtf": 0.2284940646859468,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.5844667553901672,
    "rms": 0.08327605954232069,
    "peak_rss_bytes": 980762624,
    "sha256": "7cf19e327199710258ddc7e8e7cf0e396fbba8afc76ca157b1722cce6dec4a5b",
    "matched_file": "matched/pocket-azelma-technical.wav",
    "input_lufs": -20.91,
    "gain_db": -5.09,
    "matched_lufs": -26.0,
    "matched_sha256": "cbccf4ba476c9baae02c9a6eeea50aba09d50c08e90f1e62ad3c61990c22a17a"
  },
  {
    "phase": "sample",
    "model": "pocket",
    "voice": "eponine",
    "passage": "conversation",
    "text": "Hey Aku. I found a few voices worth listening to. Take your time; I care more about whether this sounds natural than whether it wins a benchmark. Does the rhythm feel right, or does it sound like someone reading a script?",
    "file": "raw/pocket-eponine-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 12.24,
    "generation_seconds": 2.8496928338427097,
    "first_chunk_seconds": 0.10243804194033146,
    "rtf": 0.23281804198061354,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.5936885476112366,
    "rms": 0.08718797500261921,
    "peak_rss_bytes": 990052352,
    "sha256": "bbf2afd3440f569cb75ebe05d65476000aacdff8fde79d8716d6477379733dff",
    "matched_file": "matched/pocket-eponine-conversation.wav",
    "input_lufs": -21.23,
    "gain_db": -4.77,
    "matched_lufs": -26.0,
    "matched_sha256": "175ed70af3d68853292bdff05eff35ce44c889268faae14fcd5fc566e99664f5"
  },
  {
    "phase": "sample",
    "model": "pocket",
    "voice": "eponine",
    "passage": "expression",
    "text": "Wait, you actually got it working? Oh, that's brilliant! I was starting to think this little machine had given up on us. All right. Take a breath. We have time to get this right.",
    "file": "raw/pocket-eponine-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 10.16,
    "generation_seconds": 2.3368184159044176,
    "first_chunk_seconds": 0.0652735410258174,
    "rtf": 0.23000181258901747,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.6303151249885559,
    "rms": 0.0766076759242342,
    "peak_rss_bytes": 990117888,
    "sha256": "09705f4af28185268b450fcf16850ab61314b6d48ccd3e996b5e432ffcb54442",
    "matched_file": "matched/pocket-eponine-expression.wav",
    "input_lufs": -22.36,
    "gain_db": -3.6400000000000006,
    "matched_lufs": -26.0,
    "matched_sha256": "393bf915aa9385cf224e091e540b5845467a6650b6a153ffd2d355fb93b73433"
  },
  {
    "phase": "sample",
    "model": "pocket",
    "voice": "eponine",
    "passage": "technical",
    "text": "The backup finished at 9:42 p.m. We copied 3.7 gigabytes, verified the SHA-256 checksum, and left the GPU alone. The address is 192.168.1.3. Please don't reboot the machine yet.",
    "file": "raw/pocket-eponine-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 17.28,
    "generation_seconds": 3.944596708053723,
    "first_chunk_seconds": 0.05502320802770555,
    "rtf": 0.22827527245681264,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.9284368753433228,
    "rms": 0.07740794934118653,
    "peak_rss_bytes": 990347264,
    "sha256": "b8679edacbeda321b82240ff9a39a21b8b9420458f3eafb2c461ef3768ba996a",
    "matched_file": "matched/pocket-eponine-technical.wav",
    "input_lufs": -22.09,
    "gain_db": -3.91,
    "matched_lufs": -26.0,
    "matched_sha256": "4fe1653c79035ecd5f53f56e429f11124befba727aba5dacba263c443805c5d6"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "azelma",
    "passage": "conversation",
    "text": "Hey Aku. I found a few voices worth listening to. Take your time; I care more about whether this sounds natural than whether it wins a benchmark. Does the rhythm feel right, or does it sound like someone reading a script?",
    "file": "raw/turbo-azelma-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 13.8,
    "generation_seconds": 23.046501000178978,
    "first_chunk_seconds": null,
    "rtf": 1.6700363043607953,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.35200944542884827,
    "rms": 0.04570819428229728,
    "peak_rss_bytes": 2731933696,
    "sha256": "02602b358e310f2b924c8d83ae4294160821cff94c446b37e25115f09f05735b",
    "matched_file": "matched/turbo-azelma-conversation.wav",
    "input_lufs": -26.65,
    "gain_db": 0.6499999999999986,
    "matched_lufs": -26.0,
    "matched_sha256": "eccf710c678af1b3c38a3f2a5005b9846e7cb96e5a223d8e59f66213141b8534"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "azelma",
    "passage": "expression",
    "text": "Wait, you actually got it working? Oh, that's brilliant! I was starting to think this little machine had given up on us. All right. Take a breath. We have time to get this right.",
    "file": "raw/turbo-azelma-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 11.48,
    "generation_seconds": 19.221622249810025,
    "first_chunk_seconds": null,
    "rtf": 1.674357338833626,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.5308325290679932,
    "rms": 0.04631215568049965,
    "peak_rss_bytes": 2731933696,
    "sha256": "c6bdd2ee71481f55cca311f2a698c30115d2e2a2c10643b3ea95d9144dfc59de",
    "matched_file": "matched/turbo-azelma-expression.wav",
    "input_lufs": -26.1,
    "gain_db": 0.10000000000000142,
    "matched_lufs": -26.0,
    "matched_sha256": "566b5cf5f5281e536d5d3709ba539b96b9a7e4a8c70a70fe264dae894945bfdd"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "azelma",
    "passage": "technical",
    "text": "The backup finished at 9:42 p.m. We copied 3.7 gigabytes, verified the SHA-256 checksum, and left the GPU alone. The address is 192.168.1.3. Please don't reboot the machine yet.",
    "file": "raw/turbo-azelma-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 20.44,
    "generation_seconds": 40.198295457987115,
    "first_chunk_seconds": null,
    "rtf": 1.9666485057723637,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.6008608341217041,
    "rms": 0.048267130492170525,
    "peak_rss_bytes": 2731933696,
    "sha256": "6a1348bdfc590ae312fa17b9fbf17b2d36e7f4598c7c4b17452d92f16fbf9a33",
    "matched_file": "matched/turbo-azelma-technical.wav",
    "input_lufs": -26.33,
    "gain_db": 0.3299999999999983,
    "matched_lufs": -26.0,
    "matched_sha256": "4a3a1e13425fc520fe7597ba4fb66db7efb51009918173c71909d145bfd58a18"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "azelma",
    "passage": "vocal-events",
    "text": "Well, that was subtle. [chuckle] The desktop survived, and nobody had to buy a new graphics card. [sigh] I could get used to this.",
    "file": "raw/turbo-azelma-vocal-events.wav",
    "sample_rate": 24000,
    "duration_seconds": 9.44,
    "generation_seconds": 15.97521841689013,
    "first_chunk_seconds": null,
    "rtf": 1.6922900865349715,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.4519299566745758,
    "rms": 0.044236999257296274,
    "peak_rss_bytes": 2869673984,
    "sha256": "e2f0ed59456912505c74ca28beaacc47a30e8a218205bd753989895c61ba4773",
    "matched_file": "matched/turbo-azelma-vocal-events.wav",
    "input_lufs": -26.51,
    "gain_db": 0.5100000000000016,
    "matched_lufs": -26.0,
    "matched_sha256": "38957f35e17c01e3a3b80a749a2a93f2a4316cd31e865ef82c5c2c3247ea2584"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "eponine",
    "passage": "conversation",
    "text": "Hey Aku. I found a few voices worth listening to. Take your time; I care more about whether this sounds natural than whether it wins a benchmark. Does the rhythm feel right, or does it sound like someone reading a script?",
    "file": "raw/turbo-eponine-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 14.0,
    "generation_seconds": 26.20427562482655,
    "first_chunk_seconds": null,
    "rtf": 1.8717339732018965,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.5081843137741089,
    "rms": 0.060915993345354066,
    "peak_rss_bytes": 3116318720,
    "sha256": "2c58d0fede3ad8545edce6fb0e188da486a6c05f8eef41e55a67dea6f0f11366",
    "matched_file": "matched/turbo-eponine-conversation.wav",
    "input_lufs": -24.43,
    "gain_db": -1.5700000000000003,
    "matched_lufs": -26.0,
    "matched_sha256": "5b0c9621af5dba382024e0f7bab2be379869e3828ffb56418416eddc58467a5e"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "eponine",
    "passage": "expression",
    "text": "Wait, you actually got it working? Oh, that's brilliant! I was starting to think this little machine had given up on us. All right. Take a breath. We have time to get this right.",
    "file": "raw/turbo-eponine-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 11.28,
    "generation_seconds": 19.41038970905356,
    "first_chunk_seconds": null,
    "rtf": 1.720779229526025,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.7424607276916504,
    "rms": 0.05655162319084954,
    "peak_rss_bytes": 3116318720,
    "sha256": "6e7f3de3816c95f54b58ae11770e0d9afff9b816363ad508c85f109f8dac96b8",
    "matched_file": "matched/turbo-eponine-expression.wav",
    "input_lufs": -24.51,
    "gain_db": -1.4899999999999984,
    "matched_lufs": -26.0,
    "matched_sha256": "bfaba78d21bcf9c5e6ee0ef518c11247a5917492141ffa80bfd98e80b36bcc62"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "eponine",
    "passage": "technical",
    "text": "The backup finished at 9:42 p.m. We copied 3.7 gigabytes, verified the SHA-256 checksum, and left the GPU alone. The address is 192.168.1.3. Please don't reboot the machine yet.",
    "file": "raw/turbo-eponine-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 18.16,
    "generation_seconds": 32.437263041967526,
    "first_chunk_seconds": null,
    "rtf": 1.7861928987867581,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.9771501421928406,
    "rms": 0.05730704041971529,
    "peak_rss_bytes": 3116318720,
    "sha256": "20b01d25422d3817a06e368bbbb13682a7ba2aebe30512d275dba0beb987c836",
    "matched_file": "matched/turbo-eponine-technical.wav",
    "input_lufs": -24.64,
    "gain_db": -1.3599999999999994,
    "matched_lufs": -26.0,
    "matched_sha256": "970fffa182fa928940c652dfe1adac296c5c39e6f2181e98666bcbb10a5ada5f"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "eponine",
    "passage": "vocal-events",
    "text": "Well, that was subtle. [chuckle] The desktop survived, and nobody had to buy a new graphics card. [sigh] I could get used to this.",
    "file": "raw/turbo-eponine-vocal-events.wav",
    "sample_rate": 24000,
    "duration_seconds": 8.52,
    "generation_seconds": 14.76461087493226,
    "first_chunk_seconds": null,
    "rtf": 1.7329355487009694,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.6130111217498779,
    "rms": 0.051019377152887735,
    "peak_rss_bytes": 3116318720,
    "sha256": "76561eb4a9bfe96c6c5798afe77bdfba28b544aff8f90854eab9932b9ed74a68",
    "matched_file": "matched/turbo-eponine-vocal-events.wav",
    "input_lufs": -25.28,
    "gain_db": -0.7199999999999989,
    "matched_lufs": -26.0,
    "matched_sha256": "0d7081cdd9a28eba6e2b4a7a4f9f7a590da7183299db4a3e914f642c9dd6ea3e"
  }
]