{
  "contents": [
    {
      "role": "user",
      "parts": [
        {
          "text": "Two excerpts of the same scene from two sung story-songs (Attachment 1 English original, Attachment 2 Ukrainian candidate): three weeks later at the mirror, lines lighter; the daughter says you look different, brighter; by week six she cannot believe her eyes; skin more even than in a decade; she goes out with her neck uncovered; compare how the time jumps are marked (held word, breath, thinning), the daughter's colour, the surprise attack, the plain understated payoff, and syllable density. Give timed landmarks (attachment, local seconds, quoted words). Then: mechanismsMatched[{mechanism,attachment1,attachment2,matched(boolean)}], defects[{localSeconds,words,problem,severity}], wordsMisheardOrUnclear[], whatCandidateDoesWell[], remainingGaps[]. Return ONLY JSON (no prose outside JSON), <=1900 words, English. Required keys: mediaAccess(boolean), audioAccess(boolean), attachmentsHeard[{attachment,firstWordsHeard,lastWordsHeard,durationEstimateSeconds}], then the analysis keys listed below. No script is supplied: quote only words you actually hear and mark uncertain words with (?). Use attachment number plus LOCAL seconds of that attachment. Do not rate quality with numbers, do not assume either language or the original is better, do not reward louder mastering, more notes or softer timbre. Identify concrete mechanisms, not taste."
        },
        {
          "text": "Attachment 1: Attachment 1; parent targ.mp4; EXACT parent range [122.6, 146.1]; local audio starts0. Use attachment number and local seconds, never concatenate timestamps. Fixed gain, no retiming."
        },
        {
          "inlineData": {
            "mimeType": "audio/mpeg",
            "data": "[base64 of frozen bytes; see inputs.json]"
          }
        },
        {
          "text": "Attachment 2: Attachment 2; parent gopure-targ-uk-master.wav; EXACT parent range [112.46, 135.68]; local audio starts0. Use attachment number and local seconds, never concatenate timestamps. Fixed gain, no retiming."
        },
        {
          "inlineData": {
            "mimeType": "audio/mpeg",
            "data": "[base64 of frozen bytes; see inputs.json]"
          }
        }
      ]
    }
  ],
  "stream": true,
  "generationConfig": {
    "thinkingConfig": {
      "includeThoughts": false,
      "thinkingLevel": "high"
    }
  }
}