{
  "name": "event-E01-that-high",
  "model": "gemini-3-flash",
  "transport": "inline_data (base64 bytes of the actual MP4 incl. audio)",
  "endpoint": "https://api.kie.ai/gemini/v1/models/gemini-3-flash-v1betamodels:streamGenerateContent",
  "prompt": "This clip is an ORIGINAL excerpt with sound from a longer film, temporally MAGNIFIED 4x (picture and sound both slowed 4x, pitch preserved), so you receive 4 frames per original second. Clip 0.00 corresponds to source 1.80 s; source_time = 1.80 + clip_time/4. The excerpt covers source 1.80–4.60 s. Speech will sound slow; that is intended.\n\nSpecific question to answer from this clip: When exactly does the picture cut to the father relative to his words \"That high?\", what does his face do before, during and after the words, and when does the picture leave him relative to the doctor's next words \"And the scan\"?\n\nReturn ONLY valid JSON (no markdown fences):\n{\n \"access\": {\"heard_audio\": bool},\n \"timeline\": [fine-grained observable moments in order: {\"clip_time_s\", \"who\", \"face\": brow/eyes/gaze/mouth, \"hands\": what each visible hand does, \"body\": posture/lean/turn, \"prop\": object state, \"speech\": words being said at this moment and by whom, \"sound\": non-speech sound}],\n \"answer\": {\"finding\": direct answer to the question from evidence, \"onset_clip_time_s\": when the key action starts, \"completion_clip_time_s\": when it completes or null, \"relation_to_speech\": which word/pause it coincides with, \"relation_to_cut\": before/on/after any picture change in the clip, \"confidence\": 0-1},\n \"uncertainties\": [str]\n}\nDo not invent anything not visible or audible.\n",
  "prompt_sha256": "2354bee10dcf8b62826e8de7a6e77a32b66154297c65feb167cbbf49f2dada78",
  "media": {
    "path": "/workspaces/UGS/AdFactoryNarrativeFormat/trees/bf8f11bde4/analysis/original-atomic/fragments/events/E01-that-high-x4.mp4",
    "sha256": "9e6f2ebfa37197dab7a0eae18a0f6549a8da0dc0ee90cb0fa08425d480bb62f4",
    "bytes": 419013,
    "mime": "video/mp4"
  },
  "extra_media": [],
  "generation_config": {
    "thinkingConfig": {
      "includeThoughts": false,
      "thinkingLevel": "high"
    },
    "maxOutputTokens": 12000,
    "temperature": 0.2,
    "mediaResolution": "MEDIA_RESOLUTION_HIGH"
  },
  "meta": {
    "event": {
      "id": "E01-that-high",
      "start_s": 1.8,
      "end_s": 4.6,
      "question": "When exactly does the picture cut to the father relative to his words \"That high?\", what does his face do before, during and after the words, and when does the picture leave him relative to the doctor's next words \"And the scan\"?"
    },
    "clip": {
      "path": "/workspaces/UGS/AdFactoryNarrativeFormat/trees/bf8f11bde4/analysis/original-atomic/fragments/events/E01-that-high-x4.mp4",
      "sha256": "9e6f2ebfa37197dab7a0eae18a0f6549a8da0dc0ee90cb0fa08425d480bb62f4",
      "bytes": 419013,
      "source_start_s": 1.8,
      "source_end_s": 4.6,
      "duration_s": 11.166667,
      "recipe": [
        "ffmpeg",
        "-hide_banner",
        "-loglevel",
        "error",
        "-y",
        "-ss",
        "1.800",
        "-to",
        "4.600",
        "-i",
        "/workspaces/UGS/AdFactoryNarrativeFormat/data/references_liked/resilia-aged-garlic-family-cholesterol-story.mp4",
        "-vf",
        "setpts=4*PTS,scale=-2:854",
        "-af",
        "atempo=0.5,atempo=0.5",
        "-r",
        "30",
        "-c:v",
        "libx264",
        "-preset",
        "medium",
        "-crf",
        "26",
        "-pix_fmt",
        "yuv420p",
        "-c:a",
        "aac",
        "-b:a",
        "96k",
        "-movflags",
        "+faststart",
        "/workspaces/UGS/AdFactoryNarrativeFormat/trees/bf8f11bde4/analysis/original-atomic/fragments/events/E01-that-high-x4.mp4"
      ],
      "time_scale": 4,
      "note": "Clip runs 4x slower than the source; clip_time/4 + 1.800 = source time."
    },
    "source_start_s": 1.8,
    "time_scale": 4.0
  }
}