{
  "schema_version": 1,
  "demo_scope": "real-model-inpainting-component-only",
  "target_model_evaluated": false,
  "samples": [
    {
      "sample_id": "goldberg_variation20_excerpt",
      "display_name": "Goldberg Variation 20 \u2014 30 s excerpt",
      "source_audio": "https://commons.wikimedia.org/wiki/File:Kimiko_Ishizaka_-_21_-_Variatio_20_a_2_Clav.ogg",
      "source_attribution": "J. S. Bach, Goldberg Variations, Variatio 20 a 2 Clav.; performed by Kimiko Ishizaka",
      "source_license": "CC0 1.0 Universal",
      "duration_seconds": 30.0,
      "original_audio": "data/audio/real-demo/goldberg_variation20_excerpt_original.wav",
      "masked_audio": "data/audio/real-demo/goldberg_variation20_excerpt_masked.wav",
      "original_spectrogram": "data/spectrograms/real-demo/goldberg_variation20_excerpt_original.png",
      "masked_spectrogram": "data/spectrograms/real-demo/goldberg_variation20_excerpt_masked.png",
      "original_region_seconds": [
        [
          5.0,
          5.4
        ],
        [
          12.0,
          12.3
        ],
        [
          18.0,
          18.5
        ]
      ],
      "effective_regions": [
        {
          "original": [
            5.0,
            5.4
          ],
          "effective": [
            5.015510204081632,
            5.38702947845805
          ],
          "start_frame": 432,
          "end_frame": 464,
          "center_shift_seconds": 0.0012698412698410877,
          "alignment_reason": "GACELA maestro-short requires a fixed 32-frame gap; the interval was aligned around the preserved legacy center."
        },
        {
          "original": [
            12.0,
            12.3
          ],
          "effective": [
            11.969886621315192,
            12.34140589569161
          ],
          "start_frame": 1031,
          "end_frame": 1063,
          "center_shift_seconds": 0.005646258503400503,
          "alignment_reason": "GACELA maestro-short requires a fixed 32-frame gap; the interval was aligned around the preserved legacy center."
        },
        {
          "original": [
            18.0,
            18.5
          ],
          "effective": [
            18.06512471655329,
            18.436643990929706
          ],
          "start_frame": 1556,
          "end_frame": 1588,
          "center_shift_seconds": 0.0008843537414975344,
          "alignment_reason": "GACELA maestro-short requires a fixed 32-frame gap; the interval was aligned around the preserved legacy center."
        }
      ],
      "crossfade_seconds": 0.02,
      "model_name": "GACELA maestro-short (375 ms gap)",
      "model_repository": "https://github.com/andimarafioti/GACELA",
      "checkpoint_name": "01_310000.pt",
      "checkpoint_hash": "bd32992080febfbc37f7e77b625ca11a4580c804a12cfa6bfdba4b2b14688970",
      "checkpoint_archive_md5": "1ac767255cf9edd53b8215a4301132dc",
      "code_commit": "34649fb01bdecbcb266db046a8b9c48c141f16e1",
      "sample_rate": 22050,
      "input_hash": "a87c76339dc1dc16f39f59cc2ebd55cdcd9b346420625338c76c976f31f1e726",
      "masked_hash": "3bc5777d54ff769dfce15323195265cc081a34be016724e41384200efa704080",
      "generation_date": "2026-09-04",
      "target_model_evaluated": false,
      "selected_seed": 1401,
      "selection_rule": "Among candidates without NaN, silence, or clipping, select the lowest mean boundary jump; use absolute RMS difference as a tiebreaker.",
      "candidates": [
        {
          "random_seed": 1401,
          "region_seeds": [
            1401,
            1402,
            1403
          ],
          "regenerated_audio": "data/audio/real-demo/goldberg_variation20_excerpt_regenerated_seed-1401.wav",
          "output_hash": "d818d7f9c67ee9f5d808f03b3ccd39f4da0f7def2992731c4327ce843fcc753a",
          "regenerated_spectrogram": "data/spectrograms/real-demo/goldberg_variation20_excerpt_seed-1401.png",
          "difference_spectrogram": "data/spectrograms/real-demo/goldberg_variation20_excerpt_seed-1401_difference.png",
          "diagnostics": {
            "peak": 0.58050537109375,
            "rms_difference_db": -0.10030563114548038,
            "mean_boundary_jump": 0.011606852213541666,
            "outside_max_abs_difference": 0.0
          }
        },
        {
          "random_seed": 2207,
          "region_seeds": [
            2207,
            2208,
            2209
          ],
          "regenerated_audio": "data/audio/real-demo/goldberg_variation20_excerpt_regenerated_seed-2207.wav",
          "output_hash": "b8c408f3f50f9f6e657fa3f94da337536d558f7bf165d590fa321bfa0325dd43",
          "regenerated_spectrogram": "data/spectrograms/real-demo/goldberg_variation20_excerpt_seed-2207.png",
          "difference_spectrogram": "data/spectrograms/real-demo/goldberg_variation20_excerpt_seed-2207_difference.png",
          "diagnostics": {
            "peak": 0.58050537109375,
            "rms_difference_db": -0.10046722183523174,
            "mean_boundary_jump": 0.011606852213541666,
            "outside_max_abs_difference": 0.0
          }
        },
        {
          "random_seed": 3511,
          "region_seeds": [
            3511,
            3512,
            3513
          ],
          "regenerated_audio": "data/audio/real-demo/goldberg_variation20_excerpt_regenerated_seed-3511.wav",
          "output_hash": "bdfdddfda9c4788e725512903a9f3b12a2eafff91843fe59d84a4454e4707f43",
          "regenerated_spectrogram": "data/spectrograms/real-demo/goldberg_variation20_excerpt_seed-3511.png",
          "difference_spectrogram": "data/spectrograms/real-demo/goldberg_variation20_excerpt_seed-3511_difference.png",
          "diagnostics": {
            "peak": 0.58050537109375,
            "rms_difference_db": -0.10060939357453441,
            "mean_boundary_jump": 0.011606852213541666,
            "outside_max_abs_difference": 0.0
          }
        }
      ],
      "inference_verification": {
        "checkpoint_loaded": true,
        "loaded_parameter_count": 9482017,
        "loaded_parameter_abs_sum": 369462.7235278487,
        "torch_version": "2.0.1"
      },
      "notes": "The fixed interval centers come from the legacy MAIA page metadata. They were aligned to the GACELA maestro-short 32-frame gap. No target classifier or retrieval model was run for this interactive sample."
    }
  ]
}