VyvoUp / comparison /report.json
kadirnar's picture
Release VyvoUp v2.0 native-48k synthetic adaptation
4fba828 verified
Raw History Blame Contribute Delete
27.7 kB
{
"code_state_sha256": "e06dd72f96b63b2c8724fd27e49c364bc908d88dcc9f7fa488107bdba6b77267",
"completed_at": "2026-08-13T06:20:19.561952+00:00",
"config": {
"dataset_root": "/home/kadir/kadir_projects/github/VyvoUp/datasets/vctk-pcm-10h",
"example_count": 4,
"fft_sizes": [
256,
512,
1024
],
"flashsr_checkpoint": "/home/kadir/kadir_projects/github/VyvoUp/.model-sources/flashsr/FastAudioSR/SR48k.pth",
"flashsr_command": [
".model-envs/flashsr/bin/python",
"-m",
"audio_upscaler.comparison.flashsr_worker"
],
"flashsr_implementation_revision": "2a69326250613c0a0f6c1c8d9f0c48cb779842b8",
"flashsr_model_id": "YatharthS/FlashSR",
"flashsr_model_revision": "3e19cc92e655c2e0661e6268efdade60f42fd0b8",
"flashsr_repository": "https://github.com/ysharma3501/FlashSR",
"maximum_items": 0,
"output_root": "/home/kadir/kadir_projects/github/VyvoUp/artifacts/comparisons/vyvoup-final-vs-flashsr-v2",
"schema_version": 1,
"seed": 20260813,
"split": "test",
"vyvoup_checkpoint": "/home/kadir/kadir_projects/github/VyvoUp/artifacts/efficient-hbr-validation-soup-v4/checkpoints/step-00000300.pt",
"working_directory": "/home/kadir/kadir_projects/github/VyvoUp"
},
"config_sha256": "8b0b3765dd4509ae03fcba132d5b7e1504583860f6d65dbbca244f6c9d61cbc8",
"dataset": {
"duration_seconds": 1803.001625,
"generation_sha256": "fd8064558142ae4d45f109227cc837e9c21a61908309112122cfa79ba267f77c",
"held_out_speakers": true,
"items": 527,
"prior_use_disclosure": "This VCTK test split was previously evaluated for VyvoUp v1; the comparison is descriptive and is not used to tune v2.",
"root": "/home/kadir/kadir_projects/github/VyvoUp/datasets/vctk-pcm-10h",
"split": "test"
},
"examples": [
{
"directory": "examples/01-p260_062_mic1",
"id": "p260_062_mic1",
"sha256": {
"flashsr_48khz.wav": "cc8dab08d2423040900c2d957b960701d9c027588126a97ec90c725925b52105",
"input_common_16khz.wav": "b12971eaa38c7d6f11167edceade9e94d4349369afb3074009c061074fbaab29",
"input_equal_24khz.wav": "e5d2806350042e36eb47f08995dbe3a6038a8bed44e2b0a0b5e188978cecee00",
"input_native_24khz.wav": "0d60f8ade3b25170e5c4de8e808343b49e91dedb7a1ef634018d011cd70c6d1e",
"sinc_16to48_baseline.wav": "d543874c04c6c69ffe1225779b92e7f630a7cef76c0ea00b19729a95766c91a0",
"target_48khz.wav": "037c158582a1cb54da2478a86834b82bfab92855cf6ddb9872cc5d1eea6a65a7",
"vyvoup_equal_48khz.wav": "b597b53843ed4990a5aa93ae7154ba76d90e3f1f47bf3aa68230cc9b900753d1",
"vyvoup_native_48khz.wav": "00b4e201df06139e9366ced77a34d74ad974aa48fecc8b5649e509fcb98f9cee"
}
},
{
"directory": "examples/02-p347_025_mic1",
"id": "p347_025_mic1",
"sha256": {
"flashsr_48khz.wav": "3a8a5c7c43c01a3eb1d8f35768023b53bc66c62547c267b55c696d06eb1175b9",
"input_common_16khz.wav": "3d3b6c37628a4164731ec2ae0fb50f8ad538fa97e1bc400d7b05596a74b5384d",
"input_equal_24khz.wav": "2c38d199e69543b59fbc2527571a62a91c00b7a58be7ff0d058d8c9da5b2a620",
"input_native_24khz.wav": "be9de31f6b7a5a058b790203d782fba15528c748df91cd81d9719058f2107866",
"sinc_16to48_baseline.wav": "a4cb6f5be718747efe6909d727d9a8aa7c4df3e6de6eb0c4c4ad7bb825ac9be0",
"target_48khz.wav": "595ab71fb944987c6909c28274783e819fbe3b760724db13329b3a7589057201",
"vyvoup_equal_48khz.wav": "529a022fb5af7f49ef57451f356ae1de86061113f0435ee9ce139a412a21f704",
"vyvoup_native_48khz.wav": "2c461a74d81bd88d3e8eb073b07bdc8eaf51249e75001431922ed2115dad9967"
}
},
{
"directory": "examples/03-p280_127_mic1",
"id": "p280_127_mic1",
"sha256": {
"flashsr_48khz.wav": "1bc2ec063bf4c3d79def33d4987b2b3c4598854629f4c9173b835897de66ed16",
"input_common_16khz.wav": "40658b51b10c1079b2a572ac6695e86b910d39a845611e8f994fb1efb0d37ea7",
"input_equal_24khz.wav": "9eba7ddd36f29f60deb4dbd5e31ce12f9c428d853f91b2c5ddcad48a82386323",
"input_native_24khz.wav": "02c7c5240f678e29489450f31174e72dc713b1fdd71d93a83350e05f1f5e5077",
"sinc_16to48_baseline.wav": "959de747908a25e2a2831852640e621e12780814ffee9b25d1e186dbc47db298",
"target_48khz.wav": "82ed8d27c2168a89cf65e8cdbf773dfe895811f4e62d0fa600288601975f102a",
"vyvoup_equal_48khz.wav": "73ebd5a408c2d5d35a059bf9db7ba44cbcc71c6073089885c45a02e275d29af7",
"vyvoup_native_48khz.wav": "4b0cbd17b62de2230f02fda2d9c75cb054796301ea81772cdc90195ea6584131"
}
},
{
"directory": "examples/04-p276_418_mic1",
"id": "p276_418_mic1",
"sha256": {
"flashsr_48khz.wav": "b14b3b63d5299182d65026871db000f24072c6249fa52e572c4e14a68e349791",
"input_common_16khz.wav": "46c0150e0f4793ed39fef2ec30c6efde0b56351f5a9f89510272720e1c4c55eb",
"input_equal_24khz.wav": "7339397752046cb954c571ec9e958aa1001920f4cc265d3882c2192f9c060486",
"input_native_24khz.wav": "8a38378407ae35f027fc1d8f3cda781a202d472dfdf4ac778bb993e5ea26078f",
"sinc_16to48_baseline.wav": "bf0458e67f26191c824e03c26438175db429dd6c4a9feddcee8d94644afd7837",
"target_48khz.wav": "c71afaf6923cc7259601650eae408520d86856872b3d49fb31e23e780c1078c1",
"vyvoup_equal_48khz.wav": "c97915fda2d642c828d91f1cf033f532be90d7f4cfb17f07bb7b76c06945d707",
"vyvoup_native_48khz.wav": "e389f7aef51391484682191dce74a4343a4b69ecf5144e46aaa642759f189fda"
}
}
],
"format_version": 1,
"implementation_caveats": [
"FlashSR accepts 16 kHz while VyvoUp accepts 24 kHz, so native inputs carry different information.",
"The official FlashSR FASR.run wrapper peak-normalizes every output to 0.999.",
"FlashSR's linked repository describes a tiny HiFi-GAN/HierSpeech++-style upsampler, not the diffusion FlashSR paper implementation.",
"The compared official FlashSR PyTorch graph uses same-padded noncausal convolutions. Its repository also provides a distinct ONNX streaming wrapper, which is outside this quality comparison.",
"Objective spectral distance does not replace blinded listening tests.",
"Both RTF measurements include host-to-device input transfer and device-to-host output transfer; model loading and one real-input warm-up are excluded."
],
"models": {
"flashsr": {
"active_forward_parameter_count": 88128,
"architecture": {
"activation": "anti-aliased SnakeBeta",
"active_residual_blocks": [
0,
2
],
"causality": "noncausal same-padded convolutions",
"constructed_but_inactive_residual_blocks": [
1
],
"initial_channels": 32,
"input_sample_rate_hz": 16000,
"output_sample_rate_hz": 48000,
"residual_block_kernel_sizes": [
3,
11
],
"upsampling": "3x linear interpolation"
},
"checkpoint": {
"bytes": 1715101,
"path": "/home/kadir/kadir_projects/github/VyvoUp/.model-sources/flashsr/FastAudioSR/SR48k.pth",
"sha256": "62c70874ac4efeb4dc9c8aa9dc0a611a951e1c36292abeb4c406d7fb91e0eefc"
},
"checkpoint_tensor_elements": 132360,
"cuda": "12.8",
"cuda_peak_allocated_bytes": 415837184,
"device": "cuda:0",
"device_name": "NVIDIA GeForce RTX 5070 Ti",
"format_version": 1,
"implementation": "official FastAudioSR.FASR.run",
"implementation_revision": "2a69326250613c0a0f6c1c8d9f0c48cb779842b8",
"license": "Apache-2.0 (repository README and Hugging Face metadata)",
"model_id": "YatharthS/FlashSR",
"model_load_seconds": 0.1684477689996129,
"model_revision": "3e19cc92e655c2e0661e6268efdade60f42fd0b8",
"official_per_clip_peak_normalization": true,
"parameter_count": 131904,
"precision": "float16",
"process_peak_rss_bytes": 1496055808,
"repository": "https://github.com/ysharma3501/FlashSR",
"restricted_checkpoint_loading": true,
"runtime_boundary": "CPU tensor -> official GPU FASR.run including peak normalization -> CPU tensor",
"torch": "2.11.0+cu128",
"warmup_policy": "first real input once, excluded, then repeated and measured"
},
"vyvoup": {
"checkpoint": "/home/kadir/kadir_projects/github/VyvoUp/artifacts/efficient-hbr-validation-soup-v4/checkpoints/step-00000300.pt",
"checkpoint_sha256": "133a1d019d05c9a146f3ea618b868021268a58842f09841f85f900e5b05774c7",
"cuda_peak_allocated_bytes": 1013220352,
"model_load_seconds": 0.09039203499560244,
"precision": "bfloat16",
"profile": {
"algorithmic_delay_ms": 6.416666666666667,
"algorithmic_delay_output_samples": 308,
"architecture": "EfficientHBR",
"estimated_macs_per_second": 801288000,
"parameter_bytes_fp32": 128448,
"receptive_field_input_samples": 2059,
"receptive_field_ms": 85.79166666666667,
"streaming_state_bytes_fp32": 398932,
"total_parameters": 32112,
"trainable_parameters": 32112
},
"runtime_boundary": "CPU tensor -> GPU inference -> aligned FP32 CPU waveform"
}
},
"objective_result": {
"equal_information_rms_gain_aligned_shared_12_24khz_lsd_margin_db": 3.204423308883298,
"equal_information_rms_gain_aligned_shared_12_24khz_lsd_winner": "vyvoup",
"equal_information_shared_12_24khz_lsd_margin_db": 1.9116385290224152,
"equal_information_shared_12_24khz_lsd_winner": "vyvoup",
"lower_lsd_is_better": true,
"native_rms_gain_aligned_shared_12_24khz_lsd_margin_db": 4.669459507806819,
"native_rms_gain_aligned_shared_12_24khz_lsd_winner": "vyvoup",
"native_shared_12_24khz_lsd_margin_db": 3.3886021022637856,
"native_shared_12_24khz_lsd_winner": "vyvoup",
"raw_metrics_follow_each_official_wrapper": true,
"rms_gain_aligned_metrics_are_diagnostic": true
},
"protocols": {
"equal_information": {
"common_observation": "48 kHz target -> deterministic 16 kHz Kaiser",
"flashsr_adapter": "none",
"missing_band_hz": [
8000,
24000
],
"shared_metric_band_hz": [
12000,
24000
],
"vyvoup_adapter": "deterministic 16->24 kHz Kaiser interpolation only"
},
"native": {
"flashsr_observation": "same target -> deterministic 16 kHz Kaiser",
"native_missing_bands_hz": {
"flashsr": [
8000,
24000
],
"vyvoup": [
12000,
24000
]
},
"shared_metric_band_hz": [
12000,
24000
],
"vyvoup_observation": "48 kHz target -> deterministic 24 kHz Kaiser"
},
"resampling": "scipy.signal.resample_poly, Kaiser beta 12, zero padded",
"target_alignment": "tail cropped to a multiple of six 48 kHz samples"
},
"runtime_summary": {
"flashsr_latency_seconds": {
"maximum": 0.05793205999361817,
"mean": 0.013586752478545135,
"minimum": 0.006570892001036555,
"p50": 0.011913184003788047,
"p95": 0.020413357400684617
},
"flashsr_real_time_factor": {
"maximum": 0.007099886490375154,
"mean": 0.004041951123977624,
"minimum": 0.003021606683274107,
"p50": 0.0038471696651965533,
"p95": 0.005060941355754778
},
"vyvoup_equal_latency_seconds": {
"maximum": 0.08827425799972843,
"mean": 0.01021124375685855,
"minimum": 0.003992500991444103,
"p50": 0.005120126996189356,
"p95": 0.04665261829795781
},
"vyvoup_equal_real_time_factor": {
"maximum": 0.01871369373922161,
"mean": 0.0029592607350939263,
"minimum": 0.0012343360382725678,
"p50": 0.0016109893797496598,
"p95": 0.013002484880782464
},
"vyvoup_native_latency_seconds": {
"maximum": 0.1762528649996966,
"mean": 0.013987051349868194,
"minimum": 0.0043041299941251054,
"p50": 0.008340574000612833,
"p95": 0.04872271240019471
},
"vyvoup_native_real_time_factor": {
"maximum": 0.019370634275561665,
"mean": 0.004044014354151535,
"minimum": 0.0013026777881033353,
"p50": 0.0026810720173388878,
"p95": 0.013998405001075943
}
},
"scope": "objective reconstruction and runtime; no subjective listening claim",
"summary": {
"equal_information": {
"flashsr": {
"clipping_fraction": {
"maximum": 0.0,
"mean": 0.0,
"minimum": 0.0,
"p50": 0.0,
"p95": 0.0
},
"missing_band_energy_absolute_error_db": {
"maximum": 8.17873732652954,
"mean": 2.853031641180934,
"minimum": 0.007908186454169505,
"p50": 2.498233671753503,
"p95": 6.389616040835624
},
"missing_band_energy_fraction": {
"maximum": 0.0701722920554289,
"mean": 0.008530730285938967,
"minimum": 0.0001111897656481945,
"p50": 0.0052925725112752155,
"p95": 0.028354182492214407
},
"missing_band_energy_ratio_db": {
"maximum": 8.17873732652954,
"mean": 1.493739380125258,
"minimum": -7.713555332703459,
"p50": 1.6424948550134846,
"p95": 6.237564007732604
},
"native_missing_band_lsd_db": {
"maximum": 20.149064302959452,
"mean": 13.34394736113413,
"minimum": 10.620370541426224,
"p50": 12.879473178761174,
"p95": 16.662897062920624
},
"observed_band_relative_error_db": {
"maximum": 22.05600110534771,
"mean": 5.450567142810624,
"minimum": -8.280983058853852,
"p50": 5.498183411555109,
"p95": 16.604154451341213
},
"peak": {
"maximum": 0.9990234375,
"mean": 0.9990234375,
"minimum": 0.9990234375,
"p50": 0.9990234375,
"p95": 0.9990234375
},
"protected_observed_band_relative_error_db": {
"maximum": 22.04910818561469,
"mean": 5.443691663236877,
"minimum": -8.294674314868436,
"p50": 5.496186654278052,
"p95": 16.601936311381717
},
"rms_gain_aligned_native_missing_band_lsd_db": {
"maximum": 14.41596168071016,
"mean": 12.521003563056972,
"minimum": 11.607526505922914,
"p50": 12.477616103639397,
"p95": 13.376109251071345
},
"rms_gain_aligned_observed_band_relative_error_db": {
"maximum": -3.196712865687566,
"mean": -11.650457552549138,
"minimum": -18.160824495164224,
"p50": -11.996831809430077,
"p95": -5.369240352311118
},
"rms_gain_aligned_protected_observed_band_relative_error_db": {
"maximum": -3.2290233902698837,
"mean": -11.706977801981484,
"minimum": -18.201533011462093,
"p50": -12.09957314413576,
"p95": -5.390803548493131
},
"rms_gain_aligned_shared_12_24khz_lsd_db": {
"maximum": 15.224097401929845,
"mean": 12.999414409502373,
"minimum": 11.79981718023143,
"p50": 12.963171732540687,
"p95": 13.99779798168718
},
"rms_match_gain_db": {
"maximum": -2.4618952767019433,
"mean": -9.510630755567183,
"minimum": -22.598701988494607,
"p50": -9.054288443172911,
"p95": -3.455075433118714
},
"shared_0_7_5khz_relative_error_db": {
"maximum": 22.05600110534771,
"mean": 5.450567142810624,
"minimum": -8.280983058853852,
"p50": 5.498183411555109,
"p95": 16.604154451341213
},
"shared_12_24khz_lsd_db": {
"maximum": 16.843445932073262,
"mean": 11.720615693761514,
"minimum": 9.611810644902528,
"p50": 11.465033382734797,
"p95": 13.82289465964888
},
"target_missing_band_energy_fraction": {
"maximum": 0.17605699220498794,
"mean": 0.007850915959391254,
"minimum": 5.870587536028849e-05,
"p50": 0.003131335441556233,
"p95": 0.0349447857457594
}
},
"vyvoup": {
"clipping_fraction": {
"maximum": 0.0,
"mean": 0.0,
"minimum": 0.0,
"p50": 0.0,
"p95": 0.0
},
"missing_band_energy_absolute_error_db": {
"maximum": 24.912006142984072,
"mean": 14.54454569064087,
"minimum": 3.6173581524895617,
"p50": 14.39762232873658,
"p95": 20.899851372055917
},
"missing_band_energy_fraction": {
"maximum": 0.0017766968237126338,
"mean": 0.00015370681627781773,
"minimum": 1.1376883017810926e-05,
"p50": 9.928330614546078e-05,
"p95": 0.00042198273066988115
},
"missing_band_energy_ratio_db": {
"maximum": -3.6173581524895617,
"mean": -14.54454569064087,
"minimum": -24.912006142984072,
"p50": -14.39762232873658,
"p95": -7.3209519202605815
},
"native_missing_band_lsd_db": {
"maximum": 26.9940238252708,
"mean": 22.382642661878272,
"minimum": 20.4081183340989,
"p50": 22.238048059317677,
"p95": 24.257618014476204
},
"observed_band_relative_error_db": {
"maximum": -82.01733864731136,
"mean": -106.69258459104823,
"minimum": -137.2964400294285,
"p50": -106.3472858245164,
"p95": -96.4160780483937
},
"peak": {
"maximum": 0.8477678298950195,
"mean": 0.4064270267563482,
"minimum": 0.09183477610349655,
"p50": 0.3804367184638977,
"p95": 0.7526394546031951
},
"protected_observed_band_relative_error_db": {
"maximum": -126.34086198787335,
"mean": -135.06498671079225,
"minimum": -141.5414120963394,
"p50": -135.63581328150698,
"p95": -129.22181443295574
},
"rms_gain_aligned_native_missing_band_lsd_db": {
"maximum": 26.700343392595954,
"mean": 22.36352276396841,
"minimum": 20.40084972170445,
"p50": 22.2267513308353,
"p95": 24.230807299342654
},
"rms_gain_aligned_observed_band_relative_error_db": {
"maximum": -19.44220184968457,
"mean": -54.68862726948278,
"minimum": -88.07153846997153,
"p50": -54.253388243599915,
"p95": -34.46388610637935
},
"rms_gain_aligned_protected_observed_band_relative_error_db": {
"maximum": -19.44220172639951,
"mean": -54.69137305581687,
"minimum": -88.18051448454214,
"p50": -54.253391843960046,
"p95": -34.46388609401511
},
"rms_gain_aligned_shared_12_24khz_lsd_db": {
"maximum": 14.493717322568699,
"mean": 9.794991100619075,
"minimum": 8.031604989773605,
"p50": 9.636571301809168,
"p95": 11.899158418460182
},
"rms_match_gain_db": {
"maximum": 0.8800690999488524,
"mean": 0.03966718953694786,
"minimum": 0.0003382845005462897,
"p50": 0.01681529771745078,
"p95": 0.16277054433809182
},
"shared_0_7_5khz_relative_error_db": {
"maximum": -80.94632747104487,
"mean": -105.63364594876788,
"minimum": -123.64638715822359,
"p50": -105.66551007960612,
"p95": -95.59200848211124
},
"shared_12_24khz_lsd_db": {
"maximum": 14.760180204937347,
"mean": 9.8089771647391,
"minimum": 8.031661940508862,
"p50": 9.644090932066737,
"p95": 11.94088826440865
},
"target_missing_band_energy_fraction": {
"maximum": 0.17605699220498794,
"mean": 0.007850915959391254,
"minimum": 5.870587536028849e-05,
"p50": 0.003131335441556233,
"p95": 0.0349447857457594
}
}
},
"native": {
"flashsr": {
"clipping_fraction": {
"maximum": 0.0,
"mean": 0.0,
"minimum": 0.0,
"p50": 0.0,
"p95": 0.0
},
"missing_band_energy_absolute_error_db": {
"maximum": 8.17873732652954,
"mean": 2.853031641180934,
"minimum": 0.007908186454169505,
"p50": 2.498233671753503,
"p95": 6.389616040835624
},
"missing_band_energy_fraction": {
"maximum": 0.0701722920554289,
"mean": 0.008530730285938967,
"minimum": 0.0001111897656481945,
"p50": 0.0052925725112752155,
"p95": 0.028354182492214407
},
"missing_band_energy_ratio_db": {
"maximum": 8.17873732652954,
"mean": 1.493739380125258,
"minimum": -7.713555332703459,
"p50": 1.6424948550134846,
"p95": 6.237564007732604
},
"native_missing_band_lsd_db": {
"maximum": 20.149064302959452,
"mean": 13.34394736113413,
"minimum": 10.620370541426224,
"p50": 12.879473178761174,
"p95": 16.662897062920624
},
"observed_band_relative_error_db": {
"maximum": 22.05600110534771,
"mean": 5.450567142810624,
"minimum": -8.280983058853852,
"p50": 5.498183411555109,
"p95": 16.604154451341213
},
"peak": {
"maximum": 0.9990234375,
"mean": 0.9990234375,
"minimum": 0.9990234375,
"p50": 0.9990234375,
"p95": 0.9990234375
},
"protected_observed_band_relative_error_db": {
"maximum": 22.04910818561469,
"mean": 5.443691663236877,
"minimum": -8.294674314868436,
"p50": 5.496186654278052,
"p95": 16.601936311381717
},
"rms_gain_aligned_native_missing_band_lsd_db": {
"maximum": 14.41596168071016,
"mean": 12.521003563056972,
"minimum": 11.607526505922914,
"p50": 12.477616103639397,
"p95": 13.376109251071345
},
"rms_gain_aligned_observed_band_relative_error_db": {
"maximum": -3.196712865687566,
"mean": -11.650457552549138,
"minimum": -18.160824495164224,
"p50": -11.996831809430077,
"p95": -5.369240352311118
},
"rms_gain_aligned_protected_observed_band_relative_error_db": {
"maximum": -3.2290233902698837,
"mean": -11.706977801981484,
"minimum": -18.201533011462093,
"p50": -12.09957314413576,
"p95": -5.390803548493131
},
"rms_gain_aligned_shared_12_24khz_lsd_db": {
"maximum": 15.224097401929845,
"mean": 12.999414409502373,
"minimum": 11.79981718023143,
"p50": 12.963171732540687,
"p95": 13.99779798168718
},
"rms_match_gain_db": {
"maximum": -2.4618952767019433,
"mean": -9.510630755567183,
"minimum": -22.598701988494607,
"p50": -9.054288443172911,
"p95": -3.455075433118714
},
"shared_12_24khz_lsd_db": {
"maximum": 16.843445932073262,
"mean": 11.720615693761514,
"minimum": 9.611810644902528,
"p50": 11.465033382734797,
"p95": 13.82289465964888
},
"target_missing_band_energy_fraction": {
"maximum": 0.17605699220498794,
"mean": 0.007850915959391254,
"minimum": 5.870587536028849e-05,
"p50": 0.003131335441556233,
"p95": 0.0349447857457594
}
},
"vyvoup": {
"clipping_fraction": {
"maximum": 0.0,
"mean": 0.0,
"minimum": 0.0,
"p50": 0.0,
"p95": 0.0
},
"missing_band_energy_absolute_error_db": {
"maximum": 15.811488737586565,
"mean": 4.512099880044187,
"minimum": 0.023686504036251395,
"p50": 4.175260227653971,
"p95": 10.174781659662761
},
"missing_band_energy_fraction": {
"maximum": 0.004536029431366374,
"mean": 0.0006083976658990276,
"minimum": 1.4394573770868025e-05,
"p50": 0.00038682374549211103,
"p95": 0.0018681097413708679
},
"missing_band_energy_ratio_db": {
"maximum": 1.411082407993702,
"mean": -4.476814926262968,
"minimum": -15.811488737586565,
"p50": -4.175260227653971,
"p95": -0.4946479339906577
},
"native_missing_band_lsd_db": {
"maximum": 9.706765653787752,
"mean": 8.332013591497729,
"minimum": 7.78395426067389,
"p50": 8.266177014553781,
"p95": 9.083215746966607
},
"observed_band_relative_error_db": {
"maximum": -78.5384454226004,
"mean": -99.24018123646523,
"minimum": -126.02010557193218,
"p50": -98.9617732819709,
"p95": -87.72739399396043
},
"peak": {
"maximum": 0.8201649785041809,
"mean": 0.4074106758188251,
"minimum": 0.0919647365808487,
"p50": 0.38096046447753906,
"p95": 0.7519272089004515
},
"protected_observed_band_relative_error_db": {
"maximum": -126.24017135241324,
"mean": -134.87582301532177,
"minimum": -140.9234056565594,
"p50": -135.40896246087,
"p95": -129.1619022349269
},
"rms_gain_aligned_native_missing_band_lsd_db": {
"maximum": 9.65105789665323,
"mean": 8.329954901695555,
"minimum": 7.783934363354061,
"p50": 8.266051431487705,
"p95": 9.068110617952746
},
"rms_gain_aligned_observed_band_relative_error_db": {
"maximum": -26.005236348543114,
"mean": -66.66718458535235,
"minimum": -95.69798483609165,
"p50": -67.3037452669461,
"p95": -42.00190269826101
},
"rms_gain_aligned_protected_observed_band_relative_error_db": {
"maximum": -26.005236751005896,
"mean": -66.87253755369855,
"minimum": -102.20273929362484,
"p50": -67.30391558359956,
"p95": -42.00191877203445
},
"rms_gain_aligned_shared_12_24khz_lsd_db": {
"maximum": 9.65105789665323,
"mean": 8.329954901695555,
"minimum": 7.783934363354061,
"p50": 8.266051431487705,
"p95": 9.068110617952746
},
"rms_match_gain_db": {
"maximum": 0.4245181818308427,
"mean": 0.01514023797223691,
"minimum": -7.664365060497681e-05,
"p50": 0.0037452483955403492,
"p95": 0.06872585711980608
},
"shared_12_24khz_lsd_db": {
"maximum": 9.706765653787752,
"mean": 8.332013591497729,
"minimum": 7.78395426067389,
"p50": 8.266177014553781,
"p95": 9.083215746966607
},
"target_missing_band_energy_fraction": {
"maximum": 0.0851301795087686,
"mean": 0.0031311609662697114,
"minimum": 3.622106029596592e-05,
"p50": 0.001006699661518464,
"p95": 0.013890986060368662
}
}
}
}
}