Add audited DNSMOS P.835 comparison with FlashSR

Scores all 527 held-out VCTK utterances with pinned Microsoft DNSMOS models; includes raw and RMS-aligned results, paired bootstrap intervals, full per-item evidence, protocol caveats, revisions, and hashes.

README.md

@@ -10,6 +10,7 @@ tags:
 - streaming
 - pytorch
 - synthetic-data
+- dnsmos
 ---
 
 # VyvoUp EfficientHBR v2.0
@@ -24,7 +25,8 @@ the other 97% of its trainable weights remain at the real-PCM v1 checkpoint.
 
 The release includes the model, checksum sidecar, fixed-policy outputs,
 complete validation and test reports, failed adaptation audits, synthetic-data
-provenance, RTX 5070 Ti benchmarks, and an all-utterance FlashSR comparison.
+provenance, RTX 5070 Ti benchmarks, and all-utterance FlashSR/LSD and DNSMOS
+comparisons.
 
 ## Hemen dinle: 5 giriş + 5 çıktı
 
@@ -173,6 +175,38 @@ small HiFi-GAN/HierSpeech++-style upsampler, not the diffusion model from the
 similarly named paper. Full caveats, per-item metrics, and four paired listening
 sets are in [`comparison/`](comparison/README.md).
 
+### DNSMOS P.835 diagnostic
+
+The pinned local models from
+[Microsoft DNS-Challenge](https://github.com/microsoft/DNS-Challenge) were also
+run on the same 527 held-out utterances using the regular, non-personalized
+[DNSMOS P.835](https://arxiv.org/abs/2110.01763) protocol. Higher scores are
+better. The main table below matches each output's RMS to its target with one
+scalar gain, preventing FlashSR's official per-clip peak normalization from
+dominating the result.
+
+| Protocol / system | SIG | BAK | OVRL | P.808 |
+| --- | ---: | ---: | ---: | ---: |
+| VyvoUp, native 24 kHz input | **3.5174** | **4.0397** | **3.2149** | **3.6400** |
+| VyvoUp, same 16 kHz information | **3.5167** | **4.0396** | **3.2142** | **3.6383** |
+| FlashSR, 16 kHz input | 3.4883 | 3.9483 | 3.1272 | 3.6124 |
+
+Paired VyvoUp-minus-FlashSR differences and 95% bootstrap intervals:
+
+| Protocol | SIG | BAK | OVRL | P.808 |
+| --- | ---: | ---: | ---: | ---: |
+| Native | +0.0291 [0.0216, 0.0366] | +0.0915 [0.0820, 0.1011] | +0.0877 [0.0788, 0.0966] | +0.0276 [0.0220, 0.0332] |
+| Same 16 kHz information | +0.0284 [0.0209, 0.0359] | +0.0914 [0.0819, 0.1010] | +0.0870 [0.0781, 0.0958] | +0.0259 [0.0205, 0.0313] |
+
+The raw official-wrapper OVRL scores were 3.2151 for VyvoUp native, 3.2144 for
+VyvoUp equal-information, and 2.9000 for FlashSR. DNSMOS downsamples every
+output to 16 kHz, removing the synthesized 8–24 kHz band. It is therefore a
+low-band speech/noise-quality diagnostic, not a direct high-band reconstruction
+metric or a subjective listening result. The 12–24 kHz LSD comparison remains
+primary. Full distributions, caveats, pinned Microsoft revision/model hashes,
+and all 527 per-item records are in
+[`comparison/dnsmos/`](comparison/dnsmos/README.md).
+
 ## RTX 5070 Ti benchmark
 
 One second of audio, BF16, batch one, 20 warmups and 200 measured iterations:

comparison/README.md

@@ -22,6 +22,29 @@ Native raw-output winner: **vyvoup**. Equal-information raw-output winner: **vyv
 
 The native protocol respects each model's input contract; the equal-information protocol begins both systems from the same 16 kHz observation. RMS alignment applies one scalar gain only and is reported because the official FlashSR wrapper peak-normalizes every clip.
 
+## DNSMOS P.835 diagnostic
+
+The same 527 held-out utterances were evaluated with Microsoft's pinned local
+DNSMOS P.835 models. Higher scores are better. These values apply one scalar
+target-RMS matching gain per output.
+
+| Protocol / system | SIG | BAK | OVRL | P.808 |
+| --- | ---: | ---: | ---: | ---: |
+| VyvoUp, native 24 kHz input | 3.5174 | 4.0397 | 3.2149 | 3.6400 |
+| VyvoUp, same 16 kHz information | 3.5167 | 4.0396 | 3.2142 | 3.6383 |
+| FlashSR, 16 kHz input | 3.4883 | 3.9483 | 3.1272 | 3.6124 |
+
+The native paired VyvoUp-minus-FlashSR OVRL difference is +0.0877 (95%
+bootstrap interval [+0.0788, +0.0966]). All SIG, BAK, OVRL, and P.808
+intervals favor VyvoUp in both native and equal-information protocols. Raw
+official-wrapper scores, distributions, intervals, exact revisions, hashes,
+and 527 per-item records are in [`dnsmos/`](dnsmos/README.md).
+
+DNSMOS evaluates 16 kHz audio, so it removes the synthesized 8–24 kHz band and
+cannot directly measure high-band reconstruction. This is a low-band
+speech/noise-quality diagnostic; the LSD table above remains the primary
+bandwidth-extension comparison.
+
 ## Caveats
 
 - FlashSR accepts 16 kHz while VyvoUp accepts 24 kHz, so native inputs carry different information.

comparison/dnsmos/README.md

@@ -0,0 +1,48 @@
+# VyvoUp versus FlashSR — DNSMOS P.835
+
+Official local DNSMOS models were run on 527 held-out VCTK utterances (30.05 minutes). Higher scores are better.
+
+| Output | Input information | SIG | BAK | OVRL | P.808 |
+| --- | --- | ---: | ---: | ---: | ---: |
+| Original 48 kHz target | Reference | 3.5175 | 4.0399 | 3.2150 | 3.6400 |
+| Sinc baseline | Native 24 kHz input | 3.5175 | 4.0399 | 3.2150 | 3.6400 |
+| VyvoUp | Native 24 kHz input | 3.5175 | 4.0399 | 3.2151 | 3.6400 |
+| Sinc baseline | Same 16 kHz information | 3.5168 | 4.0400 | 3.2144 | 3.6383 |
+| VyvoUp | Same 16 kHz information | 3.5168 | 4.0400 | 3.2144 | 3.6383 |
+| FlashSR | Native / same 16 kHz input | 3.4169 | 3.5680 | 2.9000 | 3.6124 |
+
+## RMS-gain-aligned diagnostic
+
+One scalar per clip matches target RMS before DNSMOS; this controls for FlashSR's official per-clip peak normalization.
+
+| Output | SIG | BAK | OVRL | P.808 |
+| --- | ---: | ---: | ---: | ---: |
+| VyvoUp — native 24 kHz | 3.5174 | 4.0397 | 3.2149 | 3.6400 |
+| VyvoUp — same 16 kHz information | 3.5167 | 4.0396 | 3.2142 | 3.6383 |
+| FlashSR — 16 kHz | 3.4883 | 3.9483 | 3.1272 | 3.6124 |
+
+## Interpretation boundary
+
+DNSMOS evaluates 16 kHz audio and therefore cannot directly measure the 8–24 kHz band synthesized by these super-resolution models. Treat this as a low-band speech/noise-quality diagnostic; the 12–24 kHz LSD report remains the primary reconstruction comparison.
+
+## Reproducibility
+
+- Microsoft DNS-Challenge revision: `591184a9fcb2cbdec02520fed81a32bbbf9d73ff`
+- Local scorer script SHA-256: `1ab566afe006daab32ac7073296a5d0ef99f8b82f91c7266f3ccf26113d7a28b`
+- `sig_bak_ovr.onnx` SHA-256: `269fbebdb513aa23cddfbb593542ecc540284a91849ac50516870e1ac78f6edd`
+- `model_v8.onnx` SHA-256: `9246480c58567bc6affd4200938e77eef49468c8bc7ed3776d109c07456f6e91`
+- Resampling: librosa `kaiser_best`, matching the pinned official script's default.
+- Aggregation: 9.01-second windows, one-second hop, regular non-personalized polynomial calibration.
+
+See [`report.json`](report.json) for complete distributions, paired intervals,
+environment versions, and evidence hashes; [`per-item.jsonl`](per-item.jsonl)
+contains all 527 item-level records.
+
+## Caveats
+
+- DNSMOS is a 16 kHz metric trained for noise-suppression quality, not for speech bandwidth extension.
+- Downsampling to 16 kHz removes the synthesized 8–24 kHz band, so DNSMOS cannot directly reward or penalize that reconstructed band.
+- The native protocol gives VyvoUp a 24 kHz observation and FlashSR a 16 kHz observation; the equal-information protocol starts both from the same 16 kHz observation.
+- FlashSR's official wrapper peak-normalizes every output to 0.999; raw scores preserve this behavior and RMS-aligned scores are reported as a separate diagnostic.
+- Short VCTK utterances are repeated to fill DNSMOS's 9.01-second input window exactly as in the official local script.
+- DNSMOS is a learned proxy for human ratings and does not replace a blinded subjective listening test.

comparison/dnsmos/config.toml

@@ -0,0 +1,14 @@
+schema_version = 1
+base_comparison_config = "vyvoup-final-vs-flashsr-v2.toml"
+dnsmos_repository = "https://github.com/microsoft/DNS-Challenge"
+dnsmos_revision = "591184a9fcb2cbdec02520fed81a32bbbf9d73ff"
+dnsmos_script = "../../.model-sources/dnsmos/dnsmos_local.py"
+dnsmos_script_sha256 = "1ab566afe006daab32ac7073296a5d0ef99f8b82f91c7266f3ccf26113d7a28b"
+primary_model = "../../.model-sources/dnsmos/sig_bak_ovr.onnx"
+primary_model_sha256 = "269fbebdb513aa23cddfbb593542ecc540284a91849ac50516870e1ac78f6edd"
+p808_model = "../../.model-sources/dnsmos/model_v8.onnx"
+p808_model_sha256 = "9246480c58567bc6affd4200938e77eef49468c8bc7ed3776d109c07456f6e91"
+output_root = "../../artifacts/comparisons/vyvoup-final-vs-flashsr-dnsmos-v1"
+maximum_items = 0
+bootstrap_resamples = 10000
+seed = 20260813

comparison/dnsmos/report.json

@@ -0,0 +1,836 @@
+{
+  "caveats": [
+    "DNSMOS is a 16 kHz metric trained for noise-suppression quality, not for speech bandwidth extension.",
+    "Downsampling to 16 kHz removes the synthesized 8\u201324 kHz band, so DNSMOS cannot directly reward or penalize that reconstructed band.",
+    "The native protocol gives VyvoUp a 24 kHz observation and FlashSR a 16 kHz observation; the equal-information protocol starts both from the same 16 kHz observation.",
+    "FlashSR's official wrapper peak-normalizes every output to 0.999; raw scores preserve this behavior and RMS-aligned scores are reported as a separate diagnostic.",
+    "Short VCTK utterances are repeated to fill DNSMOS's 9.01-second input window exactly as in the official local script.",
+    "DNSMOS is a learned proxy for human ratings and does not replace a blinded subjective listening test."
+  ],
+  "code_state_sha256": "d303205f64d96b61dacf0fb2137e561cd1534987bcbeafbc26c372dfac37ab98",
+  "completed_at": "2026-08-13T10:59:29.812606+00:00",
+  "config": {
+    "base_comparison_config": "/home/kadir/kadir_projects/github/VyvoUp/configs/evaluation/vyvoup-final-vs-flashsr-v2.toml",
+    "bootstrap_resamples": 10000,
+    "dnsmos_repository": "https://github.com/microsoft/DNS-Challenge",
+    "dnsmos_revision": "591184a9fcb2cbdec02520fed81a32bbbf9d73ff",
+    "dnsmos_script": "/home/kadir/kadir_projects/github/VyvoUp/.model-sources/dnsmos/dnsmos_local.py",
+    "dnsmos_script_sha256": "1ab566afe006daab32ac7073296a5d0ef99f8b82f91c7266f3ccf26113d7a28b",
+    "maximum_items": 0,
+    "output_root": "/home/kadir/kadir_projects/github/VyvoUp/artifacts/comparisons/vyvoup-final-vs-flashsr-dnsmos-v1",
+    "p808_model": "/home/kadir/kadir_projects/github/VyvoUp/.model-sources/dnsmos/model_v8.onnx",
+    "p808_model_sha256": "9246480c58567bc6affd4200938e77eef49468c8bc7ed3776d109c07456f6e91",
+    "primary_model": "/home/kadir/kadir_projects/github/VyvoUp/.model-sources/dnsmos/sig_bak_ovr.onnx",
+    "primary_model_sha256": "269fbebdb513aa23cddfbb593542ecc540284a91849ac50516870e1ac78f6edd",
+    "schema_version": 1,
+    "seed": 20260813
+  },
+  "config_sha256": "f5743178113fcc43198060d7e089d2191c0781b7ad15eab653e7a0aaf3d8572f",
+  "dataset": {
+    "base_comparison_per_item_sha256": "091ecfd6e757b20a420791687f598831c49acbb832ccac6915ae98a909320473",
+    "base_comparison_report_sha256": "a5c52a3c19443f35dc29dfd31ec77272f48b177bc62d8a27a882a230826325bf",
+    "duration_seconds": 1803.001625,
+    "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"
+  },
+  "dnsmos": {
+    "calibration": "official non-personalized polynomial calibration",
+    "higher_is_better": true,
+    "hop_seconds": 1.0,
+    "input_sample_rate_hz": 16000,
+    "input_window_seconds": 9.01,
+    "license": "CC-BY-4.0 for DNS-Challenge assets; MIT for upstream code",
+    "mode": "regular non-personalized MOS",
+    "p808_model": {
+      "filename": "model_v8.onnx",
+      "sha256": "9246480c58567bc6affd4200938e77eef49468c8bc7ed3776d109c07456f6e91"
+    },
+    "primary_model": {
+      "filename": "sig_bak_ovr.onnx",
+      "sha256": "269fbebdb513aa23cddfbb593542ecc540284a91849ac50516870e1ac78f6edd"
+    },
+    "repository": "https://github.com/microsoft/DNS-Challenge",
+    "resampling": "librosa.resample with explicit res_type=kaiser_best, matching the pinned librosa-0.8.1 script default",
+    "revision": "591184a9fcb2cbdec02520fed81a32bbbf9d73ff",
+    "script_sha256": "1ab566afe006daab32ac7073296a5d0ef99f8b82f91c7266f3ccf26113d7a28b",
+    "session": "ONNX Runtime CPUExecutionProvider, sequential execution, one intra-op and one inter-op thread per inference, twelve concurrent independent clip workers in eight-item scheduling batches",
+    "short_clip_policy": "repeat waveform until one 9.01-second window fits"
+  },
+  "elapsed_seconds": 1836.3021753499925,
+  "environment": {
+    "cuda": "13.0",
+    "device": "NVIDIA GeForce RTX 5070 Ti",
+    "librosa": "0.11.0",
+    "numpy": "2.5.2",
+    "onnxruntime": "1.23.2",
+    "python": "3.13.14",
+    "resampy": "0.4.3",
+    "torch": "2.13.0+cu130"
+  },
+  "format_version": 1,
+  "paired_comparisons": {
+    "equal_information_raw": {
+      "bak": {
+        "flashsr_higher_items": 5,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.4719968078696209,
+        "paired_bootstrap_95_percent_ci": [
+          0.44578911576292324,
+          0.498479236637986
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 522,
+        "winner_by_interval": "vyvoup"
+      },
+      "ovrl": {
+        "flashsr_higher_items": 5,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.3143832267466196,
+        "paired_bootstrap_95_percent_ci": [
+          0.2983429257053626,
+          0.33028737987469675
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 522,
+        "winner_by_interval": "vyvoup"
+      },
+      "p808": {
+        "flashsr_higher_items": 177,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.025897316130409503,
+        "paired_bootstrap_95_percent_ci": [
+          0.020507140093240415,
+          0.03133601335069939
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 350,
+        "winner_by_interval": "vyvoup"
+      },
+      "sig": {
+        "flashsr_higher_items": 99,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.0999298305491633,
+        "paired_bootstrap_95_percent_ci": [
+          0.08977523578598925,
+          0.10991626132603068
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 428,
+        "winner_by_interval": "vyvoup"
+      }
+    },
+    "equal_information_rms_gain_aligned": {
+      "bak": {
+        "flashsr_higher_items": 91,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.09136705014963444,
+        "paired_bootstrap_95_percent_ci": [
+          0.08186295178036535,
+          0.10096829965207188
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 436,
+        "winner_by_interval": "vyvoup"
+      },
+      "ovrl": {
+        "flashsr_higher_items": 92,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.08698550333882181,
+        "paired_bootstrap_95_percent_ci": [
+          0.07806604800925487,
+          0.09584389766978445
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 435,
+        "winner_by_interval": "vyvoup"
+      },
+      "p808": {
+        "flashsr_higher_items": 177,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.025897328888830302,
+        "paired_bootstrap_95_percent_ci": [
+          0.0205071739837752,
+          0.03133602817376649
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 350,
+        "winner_by_interval": "vyvoup"
+      },
+      "sig": {
+        "flashsr_higher_items": 191,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.028399427998410935,
+        "paired_bootstrap_95_percent_ci": [
+          0.02092995326571376,
+          0.03587896430254628
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 336,
+        "winner_by_interval": "vyvoup"
+      }
+    },
+    "native_raw": {
+      "bak": {
+        "flashsr_higher_items": 5,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.47187599614294323,
+        "paired_bootstrap_95_percent_ci": [
+          0.445692636668131,
+          0.4983589863174942
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 522,
+        "winner_by_interval": "vyvoup"
+      },
+      "ovrl": {
+        "flashsr_higher_items": 5,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.31502301557809614,
+        "paired_bootstrap_95_percent_ci": [
+          0.29897670893977535,
+          0.33101253201806774
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 522,
+        "winner_by_interval": "vyvoup"
+      },
+      "p808": {
+        "flashsr_higher_items": 170,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.027577239833486054,
+        "paired_bootstrap_95_percent_ci": [
+          0.022007122639099767,
+          0.03318909253356545
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 357,
+        "winner_by_interval": "vyvoup"
+      },
+      "sig": {
+        "flashsr_higher_items": 96,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.100615846415849,
+        "paired_bootstrap_95_percent_ci": [
+          0.09048739370252327,
+          0.11060698234226848
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 431,
+        "winner_by_interval": "vyvoup"
+      }
+    },
+    "native_rms_gain_aligned": {
+      "bak": {
+        "flashsr_higher_items": 90,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.09145623790343727,
+        "paired_bootstrap_95_percent_ci": [
+          0.08195550664779873,
+          0.1010666739170723
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 437,
+        "winner_by_interval": "vyvoup"
+      },
+      "ovrl": {
+        "flashsr_higher_items": 93,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.08770961714963284,
+        "paired_bootstrap_95_percent_ci": [
+          0.078804227683616,
+          0.09659971420270105
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 434,
+        "winner_by_interval": "vyvoup"
+      },
+      "p808": {
+        "flashsr_higher_items": 170,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.027577239066009614,
+        "paired_bootstrap_95_percent_ci": [
+          0.02200711884158591,
+          0.03318908620521048
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 357,
+        "winner_by_interval": "vyvoup"
+      },
+      "sig": {
+        "flashsr_higher_items": 187,
+        "higher_is_better": true,
+        "mean_difference_vyvoup_minus_flashsr": 0.02911654833324659,
+        "paired_bootstrap_95_percent_ci": [
+          0.02164265214341567,
+          0.036597018873276904
+        ],
+        "tied_items": 0,
+        "vyvoup_higher_items": 340,
+        "winner_by_interval": "vyvoup"
+      }
+    }
+  },
+  "scope": "non-intrusive 16 kHz speech/noise-quality diagnostic; not a direct high-band super-resolution metric and not a subjective listening claim",
+  "summary": {
+    "rms_gain_aligned_scores": {
+      "flashsr": {
+        "bak": {
+          "maximum": 4.188754639792337,
+          "mean": 3.9482760563293846,
+          "minimum": 3.30274020521787,
+          "p50": 3.9665659015682877,
+          "p95": 4.128725060692873,
+          "standard_deviation": 0.1328117297619954
+        },
+        "bak_raw": {
+          "maximum": 4.526457866032918,
+          "mean": 4.04032224200116,
+          "minimum": 3.0695571899414062,
+          "p50": 4.060595154762268,
+          "p95": 4.386069854100545,
+          "standard_deviation": 0.23458803804416886
+        },
+        "ovrl": {
+          "maximum": 3.5074407234343923,
+          "mean": 3.127234510084684,
+          "minimum": 2.4465458997954235,
+          "p50": 3.149234752972989,
+          "p95": 3.3736744696362364,
+          "standard_deviation": 0.17151810145638235
+        },
+        "ovrl_raw": {
+          "maximum": 4.1459559599558515,
+          "mean": 3.516902771630649,
+          "minimum": 2.5448920726776123,
+          "p50": 3.5438714981079102,
+          "p95": 3.91121643781662,
+          "standard_deviation": 0.26359148494740015
+        },
+        "p808": {
+          "maximum": 4.231863498687744,
+          "mean": 3.6124037638970417,
+          "minimum": 2.7246737480163574,
+          "p50": 3.6294984817504883,
+          "p95": 3.9464235484600065,
+          "standard_deviation": 0.22553679719535674
+        },
+        "sig": {
+          "maximum": 3.781728687221951,
+          "mean": 3.488304743753213,
+          "minimum": 2.8235975968115103,
+          "p50": 3.5108230844749584,
+          "p95": 3.674518146546538,
+          "standard_deviation": 0.14711852293215186
+        },
+        "sig_raw": {
+          "maximum": 4.464244556427002,
+          "mean": 3.9080800177858825,
+          "minimum": 2.879090428352356,
+          "p50": 3.938454747200012,
+          "p95": 4.2452921311060585,
+          "standard_deviation": 0.2526741749905674
+        }
+      },
+      "vyvoup_equal_information": {
+        "bak": {
+          "maximum": 4.212169120375947,
+          "mean": 4.03964310647902,
+          "minimum": 3.5112475548085005,
+          "p50": 4.054154496740116,
+          "p95": 4.168139679566963,
+          "standard_deviation": 0.09849020475423781
+        },
+        "bak_raw": {
+          "maximum": 4.580577850341797,
+          "mean": 4.2094805890788685,
+          "minimum": 3.34213924407959,
+          "p50": 4.229577159881591,
+          "p95": 4.474968409538269,
+          "standard_deviation": 0.18706945925357854
+        },
+        "ovrl": {
+          "maximum": 3.5266712997814214,
+          "mean": 3.214220013423506,
+          "minimum": 2.5292147990276326,
+          "p50": 3.2260478230507674,
+          "p95": 3.4362321194974426,
+          "standard_deviation": 0.16073990750665212
+        },
+        "ovrl_raw": {
+          "maximum": 4.180505434672038,
+          "mean": 3.65433264548802,
+          "minimum": 2.654279669125875,
+          "p50": 3.6661410331726074,
+          "p95": 4.019303526878357,
+          "standard_deviation": 0.2545093365776293
+        },
+        "p808": {
+          "maximum": 4.211267948150635,
+          "mean": 3.6383010927858717,
+          "minimum": 2.7457475066184998,
+          "p50": 3.6634799242019653,
+          "p95": 3.9560792803764344,
+          "standard_deviation": 0.223860410794245
+        },
+        "sig": {
+          "maximum": 3.7746965037985896,
+          "mean": 3.5167041717516234,
+          "minimum": 2.829131524091924,
+          "p50": 3.5327477593391743,
+          "p95": 3.685788262396077,
+          "standard_deviation": 0.13307320736162837
+        },
+        "sig_raw": {
+          "maximum": 4.449194431304932,
+          "mean": 3.957113272325988,
+          "minimum": 2.887879172960917,
+          "p50": 3.9782106578350067,
+          "p95": 4.267317748069763,
+          "standard_deviation": 0.23178679270797023
+        }
+      },
+      "vyvoup_native": {
+        "bak": {
+          "maximum": 4.213284743235378,
+          "mean": 4.039732294232822,
+          "minimum": 3.5106467519343214,
+          "p50": 4.05401021033057,
+          "p95": 4.16837698740378,
+          "standard_deviation": 0.0983029425897935
+        },
+        "bak_raw": {
+          "maximum": 4.583457946777344,
+          "mean": 4.209651875681183,
+          "minimum": 3.341315269470215,
+          "p50": 4.228875637054443,
+          "p95": 4.475713288784027,
+          "standard_deviation": 0.1868633445069282
+        },
+        "ovrl": {
+          "maximum": 3.526503092424965,
+          "mean": 3.214944127234317,
+          "minimum": 2.5229964721375975,
+          "p50": 3.2272690420635364,
+          "p95": 3.436185525940128,
+          "standard_deviation": 0.1604035352567778
+        },
+        "ovrl_raw": {
+          "maximum": 4.18020486831665,
+          "mean": 3.655475431634361,
+          "minimum": 2.6460090478261313,
+          "p50": 3.6685317039489744,
+          "p95": 4.01924919128418,
+          "standard_deviation": 0.25401589094427385
+        },
+        "p808": {
+          "maximum": 4.2414422035217285,
+          "mean": 3.6399810029630517,
+          "minimum": 2.7435120940208435,
+          "p50": 3.6595144271850586,
+          "p95": 3.9546747183799744,
+          "standard_deviation": 0.22710222950887507
+        },
+        "sig": {
+          "maximum": 3.7746271478056013,
+          "mean": 3.517421292086459,
+          "minimum": 2.8221475805175005,
+          "p50": 3.5326231873069593,
+          "p95": 3.6877168855714664,
+          "standard_deviation": 0.13267403476365797
+        },
+        "sig_raw": {
+          "maximum": 4.449050108591716,
+          "mean": 3.9583578450393038,
+          "minimum": 2.878344416618347,
+          "p50": 3.9784825325012205,
+          "p95": 4.271277848879496,
+          "standard_deviation": 0.23111072411000186
+        }
+      }
+    },
+    "rms_match_gain_db": {
+      "flashsr": {
+        "maximum": -2.4618952767019433,
+        "mean": -9.510630755567183,
+        "minimum": -22.598701988494607,
+        "p50": -9.054288443172911,
+        "p95": -3.455075433118714,
+        "standard_deviation": 4.182229411328972
+      },
+      "vyvoup_equal_information": {
+        "maximum": 0.8800690999488524,
+        "mean": 0.03966718953694786,
+        "minimum": 0.0003382845005462897,
+        "p50": 0.01681529771745078,
+        "p95": 0.16277054433809182,
+        "standard_deviation": 0.0768049797529481
+      },
+      "vyvoup_native": {
+        "maximum": 0.4245181818308427,
+        "mean": 0.01514023797223691,
+        "minimum": -7.664365060497681e-05,
+        "p50": 0.0037452483955403492,
+        "p95": 0.06872585711980608,
+        "standard_deviation": 0.041353066793800075
+      }
+    },
+    "scores": {
+      "flashsr": {
+        "bak": {
+          "maximum": 4.120457691677491,
+          "mean": 3.568033077914179,
+          "minimum": 2.5751555158233916,
+          "p50": 3.6152995155522083,
+          "p95": 3.988392568128686,
+          "standard_deviation": 0.3086137748749187
+        },
+        "bak_raw": {
+          "maximum": 4.36879841486613,
+          "mean": 3.456631774756712,
+          "minimum": 2.2669315338134766,
+          "p50": 3.48953914642334,
+          "p95": 4.100485634803772,
+          "standard_deviation": 0.42940049476816494
+        },
+        "ovrl": {
+          "maximum": 3.4035131609032683,
+          "mean": 2.9000279012858834,
+          "minimum": 1.8967707950883836,
+          "p50": 2.906117704523656,
+          "p95": 3.2456626729347993,
+          "standard_deviation": 0.2234709712270248
+        },
+        "ovrl_raw": {
+          "maximum": 3.9624659220377603,
+          "mean": 3.1775639971442002,
+          "minimum": 1.8716660737991333,
+          "p50": 3.1758424043655396,
+          "p95": 3.697956222295761,
+          "standard_deviation": 0.32307538718990486
+        },
+        "p808": {
+          "maximum": 4.231863021850586,
+          "mean": 3.6124037721583195,
+          "minimum": 2.7246737480163574,
+          "p50": 3.6294984817504883,
+          "p95": 3.9464236378669737,
+          "standard_deviation": 0.22553679430400536
+        },
+        "sig": {
+          "maximum": 3.7640985535616864,
+          "mean": 3.416864839165122,
+          "minimum": 2.3882590689326886,
+          "p50": 3.4383650511394954,
+          "p95": 3.640567535027024,
+          "standard_deviation": 0.16747851100355868
+        },
+        "sig_raw": {
+          "maximum": 4.4268412590026855,
+          "mean": 3.785747662187969,
+          "minimum": 2.3231801986694336,
+          "p50": 3.8111703395843506,
+          "p95": 4.179217500686645,
+          "standard_deviation": 0.2770886464061986
+        }
+      },
+      "sinc_common_16to48": {
+        "bak": {
+          "maximum": 4.2121348452941145,
+          "mean": 4.040030566026372,
+          "minimum": 3.51131499489417,
+          "p50": 4.054688392586572,
+          "p95": 4.168261900587745,
+          "standard_deviation": 0.09825557892081427
+        },
+        "bak_raw": {
+          "maximum": 4.580493211746216,
+          "mean": 4.210219835994301,
+          "minimum": 3.3422317504882812,
+          "p50": 4.230890464782715,
+          "p95": 4.47525235414505,
+          "standard_deviation": 0.18671493069759648
+        },
+        "ovrl": {
+          "maximum": 3.5266412079673977,
+          "mean": 3.2144109440840993,
+          "minimum": 2.529302604123133,
+          "p50": 3.225695035209915,
+          "p95": 3.4362872126605586,
+          "standard_deviation": 0.16068082210530926
+        },
+        "ovrl_raw": {
+          "maximum": 4.180450598398845,
+          "mean": 3.6546352722483877,
+          "minimum": 2.6543960173924765,
+          "p50": 3.665571451187134,
+          "p95": 4.019393444061279,
+          "standard_deviation": 0.25442066148326437
+        },
+        "p808": {
+          "maximum": 4.211269855499268,
+          "mean": 3.638304058227089,
+          "minimum": 2.745749592781067,
+          "p50": 3.663482189178467,
+          "p95": 3.9560786843299867,
+          "standard_deviation": 0.22385918362618099
+        },
+        "sig": {
+          "maximum": 3.7746778565286316,
+          "mean": 3.5167937178248825,
+          "minimum": 2.8293440155869796,
+          "p50": 3.5327387196615323,
+          "p95": 3.6856331821639845,
+          "standard_deviation": 0.13303931801349206
+        },
+        "sig_raw": {
+          "maximum": 4.4491550127665205,
+          "mean": 3.957269978735454,
+          "minimum": 2.888166308403015,
+          "p50": 3.978191614151001,
+          "p95": 4.267066756884257,
+          "standard_deviation": 0.23172250734801095
+        }
+      },
+      "sinc_native_24to48": {
+        "bak": {
+          "maximum": 4.213277639486524,
+          "mean": 4.039908528187333,
+          "minimum": 3.5106756140303355,
+          "p50": 4.053998231363649,
+          "p95": 4.168388805126898,
+          "standard_deviation": 0.09818633488032158
+        },
+        "bak_raw": {
+          "maximum": 4.583439985911052,
+          "mean": 4.209987548215825,
+          "minimum": 3.3413548469543457,
+          "p50": 4.229214429855347,
+          "p95": 4.475740993022919,
+          "standard_deviation": 0.18668605647686165
+        },
+        "ovrl": {
+          "maximum": 3.5264866213986643,
+          "mean": 3.215049566698938,
+          "minimum": 2.522978902829976,
+          "p50": 3.227261656538288,
+          "p95": 3.4361866553330715,
+          "standard_deviation": 0.16037114420770981
+        },
+        "ovrl_raw": {
+          "maximum": 4.180174827575684,
+          "mean": 3.655642857735338,
+          "minimum": 2.6459860801696777,
+          "p50": 3.668519973754883,
+          "p95": 4.019251222610474,
+          "standard_deviation": 0.2539692757017117
+        },
+        "p808": {
+          "maximum": 4.24144172668457,
+          "mean": 3.6399836456316796,
+          "minimum": 2.7435015439987183,
+          "p50": 3.659514904022217,
+          "p95": 3.954674916267395,
+          "standard_deviation": 0.2271028342450336
+        },
+        "sig": {
+          "maximum": 3.7746170139445407,
+          "mean": 3.5174790934075055,
+          "minimum": 2.822171550940093,
+          "p50": 3.5325440871078855,
+          "p95": 3.6877111347499882,
+          "standard_deviation": 0.13265664004750807
+        },
+        "sig_raw": {
+          "maximum": 4.449028650919597,
+          "mean": 3.958459906745778,
+          "minimum": 2.878376841545105,
+          "p50": 3.978350067138672,
+          "p95": 4.271266158421834,
+          "standard_deviation": 0.23108073870573495
+        }
+      },
+      "target_48khz": {
+        "bak": {
+          "maximum": 4.213162787591714,
+          "mean": 4.039910629058792,
+          "minimum": 3.510712473456792,
+          "p50": 4.0539888588032635,
+          "p95": 4.168383797544631,
+          "standard_deviation": 0.09817405470973299
+        },
+        "bak_raw": {
+          "maximum": 4.583150863647461,
+          "mean": 4.209989903288675,
+          "minimum": 3.3414053916931152,
+          "p50": 4.229223012924194,
+          "p95": 4.475729775428772,
+          "standard_deviation": 0.18666776485155961
+        },
+        "ovrl": {
+          "maximum": 3.526476967796004,
+          "mean": 3.2150421609060884,
+          "minimum": 2.5228681521085603,
+          "p50": 3.2272578958031466,
+          "p95": 3.4362389316784117,
+          "standard_deviation": 0.16037916905104935
+        },
+        "ovrl_raw": {
+          "maximum": 4.18015718460083,
+          "mean": 3.655631600146243,
+          "minimum": 2.6458380619684854,
+          "p50": 3.668513870239258,
+          "p95": 4.019342293739319,
+          "standard_deviation": 0.253981653583714
+        },
+        "p808": {
+          "maximum": 4.241440773010254,
+          "mean": 3.6399986026958584,
+          "minimum": 2.7434436082839966,
+          "p50": 3.6595282554626465,
+          "p95": 3.9546756196022033,
+          "standard_deviation": 0.22710408639821644
+        },
+        "sig": {
+          "maximum": 3.774611318496073,
+          "mean": 3.5174688037342072,
+          "minimum": 2.822071109756651,
+          "p50": 3.532557755996092,
+          "p95": 3.68774193258524,
+          "standard_deviation": 0.13266985959171665
+        },
+        "sig_raw": {
+          "maximum": 4.449016571044922,
+          "mean": 3.9584428856762117,
+          "minimum": 2.878238081932068,
+          "p50": 3.9783740520477293,
+          "p95": 4.271327463785807,
+          "standard_deviation": 0.23110214285051464
+        }
+      },
+      "vyvoup_equal_information": {
+        "bak": {
+          "maximum": 4.212138424291818,
+          "mean": 4.0400298857838,
+          "minimum": 3.511305609037856,
+          "p50": 4.054667656597016,
+          "p95": 4.1682623114767505,
+          "standard_deviation": 0.09825555368403384
+        },
+        "bak_raw": {
+          "maximum": 4.580502033233643,
+          "mean": 4.210218489526614,
+          "minimum": 3.3422188758850098,
+          "p50": 4.23084716796875,
+          "p95": 4.475253307819367,
+          "standard_deviation": 0.18671501365308074
+        },
+        "ovrl": {
+          "maximum": 3.5266503459852423,
+          "mean": 3.2144111280325034,
+          "minimum": 2.529304470710219,
+          "p50": 3.2256925246326342,
+          "p95": 3.436285069433616,
+          "standard_deviation": 0.16067966960302368
+        },
+        "ovrl_raw": {
+          "maximum": 4.180467287699382,
+          "mean": 3.654635480424625,
+          "minimum": 2.654398481051127,
+          "p50": 3.665567398071289,
+          "p95": 4.019389677047729,
+          "standard_deviation": 0.25441901961895513
+        },
+        "p808": {
+          "maximum": 4.211268424987793,
+          "mean": 3.6383010882887294,
+          "minimum": 2.745747447013855,
+          "p50": 3.6634799242019653,
+          "p95": 3.956079399585724,
+          "standard_deviation": 0.22386041461429806
+        },
+        "sig": {
+          "maximum": 3.7746833995366487,
+          "mean": 3.5167946697142853,
+          "minimum": 2.8293456688880716,
+          "p50": 3.53272505275086,
+          "p95": 3.6856424364902654,
+          "standard_deviation": 0.13303757831284632
+        },
+        "sig_raw": {
+          "maximum": 4.449166774749756,
+          "mean": 3.957271502695939,
+          "minimum": 2.8881685733795166,
+          "p50": 3.9781669974327087,
+          "p95": 4.267085313796997,
+          "standard_deviation": 0.23171979829800263
+        }
+      },
+      "vyvoup_native": {
+        "bak": {
+          "maximum": 4.2132873058098435,
+          "mean": 4.039909074057122,
+          "minimum": 3.5106625739777972,
+          "p50": 4.053998879793741,
+          "p95": 4.168382860046258,
+          "standard_deviation": 0.0981866498859593
+        },
+        "bak_raw": {
+          "maximum": 4.58346430460612,
+          "mean": 4.209988686888589,
+          "minimum": 3.341336965560913,
+          "p50": 4.229211091995239,
+          "p95": 4.475727236270904,
+          "standard_deviation": 0.18668636253162701
+        },
+        "ovrl": {
+          "maximum": 3.526496019495875,
+          "mean": 3.2150509168639796,
+          "minimum": 2.5229999893330537,
+          "p50": 3.227257797851414,
+          "p95": 3.436181661340717,
+          "standard_deviation": 0.16036956877709588
+        },
+        "ovrl_raw": {
+          "maximum": 4.180191993713379,
+          "mean": 3.655644910333602,
+          "minimum": 2.6460142930348716,
+          "p50": 3.668513774871826,
+          "p95": 4.019242429733277,
+          "standard_deviation": 0.2539669041772187
+        },
+        "p808": {
+          "maximum": 4.241442680358887,
+          "mean": 3.6399810119918055,
+          "minimum": 2.7435120940208435,
+          "p50": 3.659514904022217,
+          "p95": 3.954674618244171,
+          "standard_deviation": 0.22710223683412958
+        },
+        "sig": {
+          "maximum": 3.774622705818226,
+          "mean": 3.5174806855809715,
+          "minimum": 2.8221897346855407,
+          "p50": 3.5325485555269225,
+          "p95": 3.6877149215356155,
+          "standard_deviation": 0.1326543793148034
+        },
+        "sig_raw": {
+          "maximum": 4.4490407307942705,
+          "mean": 3.9584625253124144,
+          "minimum": 2.87840207417806,
+          "p50": 3.978358507156372,
+          "p95": 4.271273930867513,
+          "standard_deviation": 0.23107710210380236
+        }
+      }
+    }
+  },
+  "systems": {
+    "flashsr_checkpoint_sha256": "62c70874ac4efeb4dc9c8aa9dc0a611a951e1c36292abeb4c406d7fb91e0eefc",
+    "flashsr_official_per_clip_peak_normalization": true,
+    "vyvoup_checkpoint_sha256": "133a1d019d05c9a146f3ea618b868021268a58842f09841f85f900e5b05774c7"
+  }
+}