[
  {
    "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": 2.379999999999999,
    "matched_lufs": -23.0,
    "matched_sha256": "2d9025fe45fb686766dbffcc0661839ec27ab64a72f4d9edf3947c6d2d643586"
  },
  {
    "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": 2.5,
    "matched_lufs": -23.0,
    "matched_sha256": "e4a6a5d840d31f2b3cf9613eea6a495ecbc8977645bf76815b50d82b59a1e4f9"
  },
  {
    "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": 2.1799999999999997,
    "matched_lufs": -23.0,
    "matched_sha256": "0bc5e809836683889bec1f9663da07dc08140c3a6b8d84a7c3cf30cbca14e473"
  },
  {
    "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": 1.2899999999999991,
    "matched_lufs": -23.0,
    "matched_sha256": "6bbca45a887b1f343955cd555a09a7b27b3505acc83d0fb81e4fc4db6b0e8300"
  },
  {
    "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": 1.2399999999999984,
    "matched_lufs": -23.0,
    "matched_sha256": "20fb8fdcddca7aac178ed0c06d5ac7fa2ff3dc4f784b429b53f7f4a6344e8ab0"
  },
  {
    "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": 1.1999999999999993,
    "matched_lufs": -23.0,
    "matched_sha256": "1464de8849501de0669200b8f8d14a147341195c31234eb6b2e1388b11c1c771"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "alba",
    "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-alba-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 13.18,
    "generation_seconds": 9.606058458797634,
    "first_chunk_seconds": null,
    "rtf": 0.7288359983913227,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.34768053889274597,
    "rms": 0.051315620226438936,
    "peak_rss_bytes": 2330771456,
    "sha256": "dc010d8dd34a0bf98ee7ceb75e41e8a900a464ed822fb55291a8178efcf46a5f",
    "matched_file": "matched/nano-alba-conversation.wav",
    "input_lufs": -25.47,
    "gain_db": 2.469999999999999,
    "matched_lufs": -23.0,
    "matched_sha256": "1198d9298bc9d2505d41b4a7bd227413c9962f28ee0e8bffa89e4a012123e3e9"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "alba",
    "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-alba-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 12.18,
    "generation_seconds": 9.051912083057687,
    "first_chunk_seconds": null,
    "rtf": 0.7431783319423388,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.32097920775413513,
    "rms": 0.048870558731132704,
    "peak_rss_bytes": 2330771456,
    "sha256": "cf2aee7e387e2b0ade5d796118cabcc09849ef5186a4ad8badcb762ba27eec3a",
    "matched_file": "matched/nano-alba-expression.wav",
    "input_lufs": -25.54,
    "gain_db": 2.539999999999999,
    "matched_lufs": -23.0,
    "matched_sha256": "e863792b1e28743f6b831b4e59bc6753cda9fb5f89dbbe48ad20b5d8f6263350"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "alba",
    "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-alba-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 16.34,
    "generation_seconds": 13.281247624894604,
    "first_chunk_seconds": null,
    "rtf": 0.8128058521967322,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.3372593820095062,
    "rms": 0.05044279187418241,
    "peak_rss_bytes": 2330771456,
    "sha256": "9f2548fab87b930588918573ea260de296b21e8427636caaede17c2ee9f69aad",
    "matched_file": "matched/nano-alba-technical.wav",
    "input_lufs": -26.38,
    "gain_db": 3.379999999999999,
    "matched_lufs": -23.0,
    "matched_sha256": "44f7c04ee2f68eb9c8ade1c25fecc16ed927118aafde9e660ab768dbc72cfc76"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "alba",
    "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-alba-vocal-events.wav",
    "sample_rate": 24000,
    "duration_seconds": 8.46,
    "generation_seconds": 6.8305968751665205,
    "first_chunk_seconds": null,
    "rtf": 0.8073991578211016,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.32091695070266724,
    "rms": 0.04495194803476889,
    "peak_rss_bytes": 2330771456,
    "sha256": "10dff5c062bf1529cdbe61285103b28e4d9bd3486c75c6e6df967a22c2f93cfc",
    "matched_file": "matched/nano-alba-vocal-events.wav",
    "input_lufs": -26.7,
    "gain_db": 3.6999999999999993,
    "matched_lufs": -23.0,
    "matched_sha256": "09389796516a70b10af1b1c23c7e39c9d398ea4547b1729c612b5fdebcb091c6"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "tara-reference",
    "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-tara-reference-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 11.74,
    "generation_seconds": 8.71362041705288,
    "first_chunk_seconds": null,
    "rtf": 0.742216389868218,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.3171693980693817,
    "rms": 0.052685035290583195,
    "peak_rss_bytes": 2501918720,
    "sha256": "44d8d0ead98426a431bbaea49097618e4fd3393c298e6ae0d3078fc221d1c6fa",
    "matched_file": "matched/nano-tara-reference-conversation.wav",
    "input_lufs": -25.71,
    "gain_db": 2.710000000000001,
    "matched_lufs": -23.0,
    "matched_sha256": "b0d4e7c1c19fc00e0018bdbdf890544d3cdae52cfe745d64347b48402e84b7a2"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "tara-reference",
    "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-tara-reference-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 10.02,
    "generation_seconds": 7.067639000015333,
    "first_chunk_seconds": null,
    "rtf": 0.7053531936143047,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.46547234058380127,
    "rms": 0.05695168611732161,
    "peak_rss_bytes": 2501918720,
    "sha256": "c7c6b068cfd9b184d5710013c8ceb6104daeffccff6e0cad8958d96bf9daf368",
    "matched_file": "matched/nano-tara-reference-expression.wav",
    "input_lufs": -25.01,
    "gain_db": 2.0100000000000016,
    "matched_lufs": -23.0,
    "matched_sha256": "c66f2afbb73ec9225eda5a7ca017fe3a05c5fbef8dc0765241b83603fa387f83"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "tara-reference",
    "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-tara-reference-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 14.94,
    "generation_seconds": 10.798345667077228,
    "first_chunk_seconds": null,
    "rtf": 0.7227808344763875,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.5653565526008606,
    "rms": 0.043534834894862616,
    "peak_rss_bytes": 2501918720,
    "sha256": "cb84c21d5edca4e847cc1a315771f6e387345b1960b7232068dd9dedea981820",
    "matched_file": "matched/nano-tara-reference-technical.wav",
    "input_lufs": -27.25,
    "gain_db": 3.95,
    "matched_lufs": -23.3,
    "matched_sha256": "3d73286212e2e53147d71ed433d7d98fd445afea9d29a2975e7c68c3bef722ea"
  },
  {
    "phase": "sample",
    "model": "nano",
    "voice": "tara-reference",
    "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-tara-reference-vocal-events.wav",
    "sample_rate": 24000,
    "duration_seconds": 8.74,
    "generation_seconds": 6.106178165879101,
    "first_chunk_seconds": null,
    "rtf": 0.6986473874003547,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.4486549496650696,
    "rms": 0.048153157111494815,
    "peak_rss_bytes": 2501918720,
    "sha256": "00e209114ead7ae9ee04e8b6a9257a0e951a5be19e394b923c24ace965e7ac40",
    "matched_file": "matched/nano-tara-reference-vocal-events.wav",
    "input_lufs": -26.35,
    "gain_db": 3.3500000000000014,
    "matched_lufs": -23.0,
    "matched_sha256": "a16ca7a0f8deb831b0e46679b9a7b46938424ec4c57c6794979db57f285a93b2"
  },
  {
    "phase": "sample",
    "model": "pocket",
    "voice": "alba",
    "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-alba-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 10.8,
    "generation_seconds": 2.5721984170377254,
    "first_chunk_seconds": 0.10800512507557869,
    "rtf": 0.2381665200960857,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.6293071508407593,
    "rms": 0.1322757804712993,
    "peak_rss_bytes": 1000423424,
    "sha256": "45f4e7e067d8c6780ca088d9905a3f6d8087b26b00d602996d1284583fcc796d",
    "matched_file": "matched/pocket-alba-conversation.wav",
    "input_lufs": -18.17,
    "gain_db": -4.829999999999998,
    "matched_lufs": -23.0,
    "matched_sha256": "e43fef4be6d7fc14ee226058588fca75fee38bc29d617c90f6c5a1eeb5e2d36c"
  },
  {
    "phase": "sample",
    "model": "pocket",
    "voice": "alba",
    "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-alba-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 8.64,
    "generation_seconds": 2.0386684590484947,
    "first_chunk_seconds": 0.0676284171640873,
    "rtf": 0.23595699757505725,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.5788008570671082,
    "rms": 0.11126821325438589,
    "peak_rss_bytes": 1000423424,
    "sha256": "e8c733029efdb6e3aa9cadba400aa5e6f35bf61bea2e3c22aec44641af360060",
    "matched_file": "matched/pocket-alba-expression.wav",
    "input_lufs": -19.91,
    "gain_db": -3.09,
    "matched_lufs": -23.0,
    "matched_sha256": "bb8865f38d06dcfbaf1b4e48c9531fd45ac68117ebc0acb022da784af8bc51c6"
  },
  {
    "phase": "sample",
    "model": "pocket",
    "voice": "alba",
    "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-alba-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 13.84,
    "generation_seconds": 3.2513584159314632,
    "first_chunk_seconds": 0.055011624936014414,
    "rtf": 0.23492474103551036,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.7276608347892761,
    "rms": 0.12074285869100881,
    "peak_rss_bytes": 1005748224,
    "sha256": "933a655a8a7a7917dd232db6af6cf208447ebb83abb1da834d7c3a62d4af70bb",
    "matched_file": "matched/pocket-alba-technical.wav",
    "input_lufs": -19.17,
    "gain_db": -3.8299999999999983,
    "matched_lufs": -23.0,
    "matched_sha256": "89327a4d424b0cc7e199201883e49efaf3e6cf38252b2d1ad3d8e4d100cb25bf"
  },
  {
    "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": 3.4060247079469264,
    "first_chunk_seconds": 0.10649112495593727,
    "rtf": 0.23652949360742542,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.7268035411834717,
    "rms": 0.09071936400878981,
    "peak_rss_bytes": 1019101184,
    "sha256": "9ac88e8ad640cb213830e7d8fc6e1e127dca2e150a0d4bf6527b156e0abcdc93",
    "matched_file": "matched/pocket-azelma-conversation.wav",
    "input_lufs": -20.4,
    "gain_db": -2.6000000000000014,
    "matched_lufs": -23.0,
    "matched_sha256": "4f4d68a6d317d4c77d86fb429aec5a650b3bfb891c6a396510e1e143513293b8"
  },
  {
    "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": 3.2776262920815498,
    "first_chunk_seconds": 0.06646608305163682,
    "rtf": 0.23546165891390444,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.504514217376709,
    "rms": 0.07862092260420656,
    "peak_rss_bytes": 1020477440,
    "sha256": "790cf21cd85a6ea8db14869fd470646e0966ef688f790e3ff8bd46fc77ae8147",
    "matched_file": "matched/pocket-azelma-expression.wav",
    "input_lufs": -20.93,
    "gain_db": -2.0700000000000003,
    "matched_lufs": -23.0,
    "matched_sha256": "1d03c401d6903df3879e362bcd0a2f57e999fe8df9c1ea869081dc18dca904e5"
  },
  {
    "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": 4.452680541900918,
    "first_chunk_seconds": 0.055658499943092465,
    "rtf": 0.2319104448906728,
    "first_chunk_note": "Model API yield, not network/playback latency",
    "peak": 0.5844667553901672,
    "rms": 0.08327605954232069,
    "peak_rss_bytes": 1021050880,
    "sha256": "5b5897d940b3ccfc185c477a24155e86a6ebe1aaf3ea4bab3b9e8387d29438ea",
    "matched_file": "matched/pocket-azelma-technical.wav",
    "input_lufs": -20.91,
    "gain_db": -2.09,
    "matched_lufs": -23.0,
    "matched_sha256": "12fbda2d77891051501adb1daff7824b00fd50256775380f318aea7ad5815707"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "alba",
    "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-alba-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 11.7,
    "generation_seconds": 19.296644374961033,
    "first_chunk_seconds": null,
    "rtf": 1.6492858440137637,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.3526850640773773,
    "rms": 0.05570618395815061,
    "peak_rss_bytes": 2411315200,
    "sha256": "3c0c3b218ae14d866cfbeede4f0faa4d2241d1e25ecc3ef8794aec08884d0867",
    "matched_file": "matched/turbo-alba-conversation.wav",
    "input_lufs": -25.3,
    "gain_db": 2.3000000000000007,
    "matched_lufs": -23.0,
    "matched_sha256": "ceff3b5f4fd62f4c0dedb57fad1be9494446fb1c8a48cd66eb447561f0f73140"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "alba",
    "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-alba-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 10.14,
    "generation_seconds": 17.79226537491195,
    "first_chunk_seconds": null,
    "rtf": 1.7546612795771153,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.3096488118171692,
    "rms": 0.047118402716066106,
    "peak_rss_bytes": 2665644032,
    "sha256": "f10032d98247065a87ea83879d3a98f78cbf240d87ed4807f6839321db91c6b0",
    "matched_file": "matched/turbo-alba-expression.wav",
    "input_lufs": -26.75,
    "gain_db": 3.75,
    "matched_lufs": -23.0,
    "matched_sha256": "8504448c0e25fe250a35bdeeb380096b8c43a4c5f852d41589749375a571cd0c"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "alba",
    "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-alba-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 19.94,
    "generation_seconds": 32.7587022499647,
    "first_chunk_seconds": null,
    "rtf": 1.642863703609062,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.3713817298412323,
    "rms": 0.05509306118365247,
    "peak_rss_bytes": 2665644032,
    "sha256": "723402fd150ce66ec43f225f51db33d1e0cf5af83d077ffc557499bd0f05a36e",
    "matched_file": "matched/turbo-alba-technical.wav",
    "input_lufs": -25.5,
    "gain_db": 2.5,
    "matched_lufs": -23.0,
    "matched_sha256": "1cc934191d72bac255ae7744d143d976808c261180e2d18d416a8368f9db73bf"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "alba",
    "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-alba-vocal-events.wav",
    "sample_rate": 24000,
    "duration_seconds": 9.3,
    "generation_seconds": 14.803965459112078,
    "first_chunk_seconds": null,
    "rtf": 1.5918242429152771,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.3368680775165558,
    "rms": 0.05007736593485111,
    "peak_rss_bytes": 2856402944,
    "sha256": "a91268a70ae5adb446cbafbc5938bbc2565d2b07442e9857e4a8fb5e19ae4d70",
    "matched_file": "matched/turbo-alba-vocal-events.wav",
    "input_lufs": -25.8,
    "gain_db": 2.8000000000000007,
    "matched_lufs": -23.0,
    "matched_sha256": "44d8ee84a9ae92ccc7035c74cf7f049569ce60549531fbdfd637d204b96cf746"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "tara-reference",
    "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-tara-reference-conversation.wav",
    "sample_rate": 24000,
    "duration_seconds": 12.46,
    "generation_seconds": 19.013627249980345,
    "first_chunk_seconds": null,
    "rtf": 1.5259732945409585,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.37406855821609497,
    "rms": 0.05181985319455458,
    "peak_rss_bytes": 3380985856,
    "sha256": "a55870fd140386f4cb9ce93349966f85e008ac0da8596680ed64651f10984910",
    "matched_file": "matched/turbo-tara-reference-conversation.wav",
    "input_lufs": -25.92,
    "gain_db": 2.9200000000000017,
    "matched_lufs": -23.0,
    "matched_sha256": "96dbfce185af14c381c6413b9c9c089148fd4d00be11226f8351a0944ae055a6"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "tara-reference",
    "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-tara-reference-expression.wav",
    "sample_rate": 24000,
    "duration_seconds": 9.22,
    "generation_seconds": 13.942735916934907,
    "first_chunk_seconds": null,
    "rtf": 1.5122273228779726,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.3585938811302185,
    "rms": 0.04882915390236058,
    "peak_rss_bytes": 3380985856,
    "sha256": "002f4476ed0b85fcf53c2aacff3221e701a9c98a5adb7806f0f7af4aed5d58e2",
    "matched_file": "matched/turbo-tara-reference-expression.wav",
    "input_lufs": -26.41,
    "gain_db": 3.41,
    "matched_lufs": -23.0,
    "matched_sha256": "e780935b89e938b12bb49a56146923e8d0107b389a52aa1860c1dadf2f4faa83"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "tara-reference",
    "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-tara-reference-technical.wav",
    "sample_rate": 24000,
    "duration_seconds": 16.5,
    "generation_seconds": 27.67278891708702,
    "first_chunk_seconds": null,
    "rtf": 1.677138722247698,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.4292708933353424,
    "rms": 0.04413977664604779,
    "peak_rss_bytes": 3380985856,
    "sha256": "1cc3e820dc5e9ff3b8d1a798e82f6665db5ccd59596c0fa0f168cbb942ea2ab9",
    "matched_file": "matched/turbo-tara-reference-technical.wav",
    "input_lufs": -27.13,
    "gain_db": 4.129999999999999,
    "matched_lufs": -23.0,
    "matched_sha256": "8e2baade8d6ccb1331e96c44164c000f4f6b04f2ec7f972298faaded7de4e2f6"
  },
  {
    "phase": "sample",
    "model": "turbo",
    "voice": "tara-reference",
    "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-tara-reference-vocal-events.wav",
    "sample_rate": 24000,
    "duration_seconds": 7.38,
    "generation_seconds": 12.70307149994187,
    "first_chunk_seconds": null,
    "rtf": 1.7212834010761342,
    "first_chunk_note": "Not measured; batch API",
    "peak": 0.46249374747276306,
    "rms": 0.040079764128770734,
    "peak_rss_bytes": 3380985856,
    "sha256": "721295763f629d18be55a2312c3383d8586e40b13ea194757ae6978484da04f8",
    "matched_file": "matched/turbo-tara-reference-vocal-events.wav",
    "input_lufs": -27.95,
    "gain_db": 4.949999999999999,
    "matched_lufs": -23.0,
    "matched_sha256": "17225172b967c37426a5fc25792bdefac80bbcb43b083439c518b54eeec45ea0"
  }
]