{
  "contents": [
    {
      "role": "user",
      "parts": [
        {
          "text": "You are reviewing actual attached AUDIO for an English-to-Ukrainian music adaptation. Hear BOTH ENTIRE tracks. Targ is the English creative source; baseline is its existing Ukrainian goPure adaptation. No numerical quality scores. Describe the actual complete rhetorical-musical form, no invented bars/key/chords. Respond concise JSON <=1500 words: mediaAccess:true only if actual audio is heard; audioAccess; tracks:[{id,heardFirstWords,heardLastWords,coverageSeconds,limitations,wholeFormDevelopment,landmarks:[{timeSeconds,heardWords,speechAct,addressee,expectation,reveal,contourOrRhythmicGesture,release,accompanimentRole}]}]; recurringCells:[{track,firstTime,returnTimes,repeatedWords,changedFunction}]; comparison:[{function,targTime,baselineTime,heardDifference,changeWorthTesting}]; uncertainties. Cover beginning,middle and end of each track, minimum one landmark per narrative block. Focus on causal repeated endings, ironic aside, negation-to-reveal, advice pivot, jar handoff, result anticipation, return gaze, her-versus-own judgment, and sung coda. Natural Ukrainian singing does not require soft/legato delivery. Do not infer silence from held words, do not confuse speaking-like rhythm with non-singing. Report approximate times inside attached durations, never invent content to fill a field. Do not claim human or frame-by-frame listening. Distinguish observation from hypothesis."
        },
        {
          "text": "Attachment 1: targ: COMPLETE audio, starts at original time zero, duration 194.376s; constant gain only, mono56kbps proxy; no time stretch."
        },
        {
          "inlineData": {
            "mimeType": "audio/mpeg",
            "data": "[base64 of frozen bytes; see inputs.json]"
          }
        },
        {
          "text": "Attachment 2: baseline: COMPLETE audio, starts at original time zero, duration 197.726s; constant gain only, mono56kbps proxy; no time stretch."
        },
        {
          "inlineData": {
            "mimeType": "audio/mpeg",
            "data": "[base64 of frozen bytes; see inputs.json]"
          }
        }
      ]
    }
  ],
  "stream": true,
  "generationConfig": {
    "thinkingConfig": {
      "includeThoughts": false,
      "thinkingLevel": "high"
    }
  }
}