{
  "aggregate": {
    "maximum_observed_recovered_wer_basis_points": 3889,
    "micro_wer_basis_points": 1293,
    "recordings": 8,
    "reference_words": 317,
    "sparse_initial_to_recovered_improvement_basis_points": 5294,
    "word_errors": 41
  },
  "corpus": {
    "license": "CC-BY-4.0",
    "name": "VOiCES",
    "registry_source": "https://registry.opendata.aws/lab41-sri-voices/",
    "selection_disclosure": "This is a post-screening characterization and regression canary selected from the 113 recordings in the pinned VOiCES_90_1 shard. It is not a prospective or statistically representative holdout.",
    "subset_revision": "6284c9f83d1190c0a5c5054c32bcb666cb3d44a9"
  },
  "initial_gate": {
    "condition": "real room, concurrent music, close microphone; severe full-file ASR dropout",
    "defer_recovery": true,
    "hypothesis_words": 4,
    "minimum_time_coverage_basis_points": 2500,
    "minimum_wer_basis_points": 8000,
    "passed": true,
    "reason": "audio_transcript_sparse",
    "reference_words": 34,
    "severity": "critical",
    "source_id": "music-dropout-close",
    "speech_enhancement_requested": true,
    "transcript_time_coverage_basis_points": 1654,
    "wer_basis_points": 9118,
    "word_errors": 31
  },
  "model": {
    "name": "ggml-small.bin",
    "sha256": "1be3a9b2063867b937e64e2ec7483364a79917e157fa98c5d94b5c1fffea987b",
    "source_revision": "c521a4b02f422512d734391fdf08bb08c0862f68"
  },
  "passed": true,
  "protocol": {
    "claim_boundary": "Eight post-screening English read-speech recordings characterize exact real-room/distractor conditions. They qualify one sparse-transcript lazy gate and one VAD recovery path for the pinned runtime/model, not representative field audio, conversations, languages, diarization, generic denoising, calibrated confidence, production SLOs or certification.",
    "overlap_ms": 1500,
    "voice_activity": {
      "frame_ms": 20,
      "minimum_rms_basis_points": 50,
      "minimum_voice_ms": 1200,
      "pre_roll_ms": 400,
      "trailing_silence_ms": 650
    },
    "window_ms": 30000
  },
  "recovery_results": [
    {
      "condition": "real room, no played distractor, close microphone",
      "detected_voice_ms": 16515,
      "hypothesis_words": 44,
      "maximum_wer_basis_points": 4500,
      "passed": true,
      "reference_words": 44,
      "source_id": "same-speech-none-close",
      "sparse_transcript": false,
      "transcript_time_coverage_basis_points": 9978,
      "wer_basis_points": 0,
      "word_errors": 0
    },
    {
      "condition": "real room, concurrent babble, close microphone",
      "detected_voice_ms": 16515,
      "hypothesis_words": 44,
      "maximum_wer_basis_points": 4500,
      "passed": true,
      "reference_words": 44,
      "source_id": "same-speech-babble-close",
      "sparse_transcript": false,
      "transcript_time_coverage_basis_points": 9978,
      "wer_basis_points": 0,
      "word_errors": 0
    },
    {
      "condition": "real room, concurrent babble, microphone behind source",
      "detected_voice_ms": 16515,
      "hypothesis_words": 44,
      "maximum_wer_basis_points": 4500,
      "passed": true,
      "reference_words": 44,
      "source_id": "same-speech-babble-behind",
      "sparse_transcript": false,
      "transcript_time_coverage_basis_points": 9688,
      "wer_basis_points": 227,
      "word_errors": 1
    },
    {
      "condition": "real room, concurrent television, far microphone",
      "detected_voice_ms": 16515,
      "hypothesis_words": 44,
      "maximum_wer_basis_points": 4500,
      "passed": true,
      "reference_words": 44,
      "source_id": "same-speech-television-far",
      "sparse_transcript": false,
      "transcript_time_coverage_basis_points": 9978,
      "wer_basis_points": 0,
      "word_errors": 0
    },
    {
      "condition": "real room, concurrent babble, far microphone",
      "detected_voice_ms": 7460,
      "hypothesis_words": 15,
      "maximum_wer_basis_points": 4500,
      "passed": true,
      "reference_words": 18,
      "source_id": "babble-far-hard",
      "sparse_transcript": false,
      "transcript_time_coverage_basis_points": 9222,
      "wer_basis_points": 3889,
      "word_errors": 7
    },
    {
      "condition": "real room, concurrent music, far microphone",
      "detected_voice_ms": 15830,
      "hypothesis_words": 46,
      "maximum_wer_basis_points": 4500,
      "passed": true,
      "reference_words": 45,
      "source_id": "music-far-hard",
      "sparse_transcript": false,
      "transcript_time_coverage_basis_points": 9500,
      "wer_basis_points": 1778,
      "word_errors": 8
    },
    {
      "condition": "real room, concurrent television, far microphone",
      "detected_voice_ms": 16010,
      "hypothesis_words": 44,
      "maximum_wer_basis_points": 4500,
      "passed": true,
      "reference_words": 44,
      "source_id": "television-far-hard",
      "sparse_transcript": false,
      "transcript_time_coverage_basis_points": 9943,
      "wer_basis_points": 2727,
      "word_errors": 12
    },
    {
      "condition": "real room, concurrent music, close microphone; severe full-file ASR dropout",
      "detected_voice_ms": 12090,
      "hypothesis_words": 26,
      "maximum_wer_basis_points": 4500,
      "passed": true,
      "reference_words": 34,
      "source_id": "music-dropout-close",
      "sparse_transcript": false,
      "transcript_time_coverage_basis_points": 9925,
      "wer_basis_points": 3824,
      "word_errors": 13
    }
  ],
  "runtime": {
    "name": "whisper.cpp",
    "server_sha256": "639cf23c070e0c111a57a443f11cef906941f7d1280918e9263f60be73ddbb3e",
    "version": "v1.8.3"
  },
  "schema_version": "mentaview.real-room-audio-result/v1"
}