Companies are realizing they're burning way too much money on AI now.

{
  "hook": {
    "spoken": "Companies are realizing they're burning way too much money on AI now.",
    "ends_at": "00:04.2",
    "on_screen": "Companies token-maxxing at AI conference",
    "starts_at": "00:00.0",
    "technique": "bold_claim",
    "technique_note": "Identifies immediate financial waste in trending tech sector",
    "why_it_stops_the_scroll": "Appeals to developers and founders worried about high LLM token costs."
  },
  "tags": {
    "topics": [
      "artificial intelligence",
      "startups",
      "token optimization",
      "tech conference"
    ],
    "cta_type": "none",
    "language": "en",
    "has_music": false,
    "has_speech": true,
    "content_type": "talking_head_advice",
    "products_or_apps": [
      "Claude",
      "Codex"
    ],
    "has_burned_captions": true
  },
  "audio": {
    "music": "None",
    "voice": "Single female speaker, natural room acoustic, unamplified mic",
    "ambient": "Slight conference audience room hum during cutaways",
    "mix_notes": "Clean direct voice-over dialogue",
    "sound_effects": "None"
  },
  "shots": [
    {
      "end": "00:05.6",
      "index": 1,
      "start": "00:00.0",
      "camera": "Front-facing smartphone camera",
      "framing": "Eye-level chest up",
      "movement": "Static handheld",
      "shot_type": "medium close-up",
      "transition_in": "Cut",
      "screen_content": "None",
      "subject_action": "Speaking and gesturing slightly",
      "what_is_visible": "Speaker seated indoors talking directly to camera",
      "duration_seconds": 5.6,
      "spoken_during_shot": "Companies are realizing they're burning way too much money on AI now. I went to the AI conference this year",
      "on_screen_text_during_shot": [
        "Companies token-maxxing at AI conference",
        "Companies",
        "they're",
        "way",
        "on",
        "went",
        "conference"
      ],
      "facial_expression_or_gesture": "Serious, direct gaze"
    },
    {
      "end": "00:08.8",
      "index": 2,
      "start": "00:05.6",
      "camera": "Smartphone camera",
      "framing": "Vertical photo inset held up to camera",
      "movement": "Static",
      "shot_type": "medium close-up",
      "transition_in": "Hard cut",
      "screen_content": "Photo overlay showing conference press badge",
      "subject_action": "Holding attendee badge towards lens",
      "what_is_visible": "Photo of speaker holding up 'The AI Conference' press pass badge",
      "duration_seconds": 3.2,
      "spoken_during_shot": "and comparing to last year, most of the companies",
      "on_screen_text_during_shot": [
        "year",
        "to",
        "The AI Conference San Francisco 2024",
        "Lan Li Founder HeyEmmett.com PRESS"
      ],
      "facial_expression_or_gesture": "Neutral selfie face"
    },
    {
      "end": "00:25.7",
      "index": 3,
      "start": "00:08.8",
      "camera": "Front-facing smartphone camera",
      "framing": "Eye-level chest up",
      "movement": "Static handheld",
      "shot_type": "medium close-up",
      "transition_in": "Hard cut",
      "screen_content": "None",
      "subject_action": "Talking and hand gesturing",
      "what_is_visible": "Speaker explaining market shift and startup categories",
      "duration_seconds": 16.9,
      "spoken_during_shot": "are shifting from AI adoption to token maxxing. At least a quarter of the startups there were either building a dashboard to monitor all of the token use from across different platforms like Claude and Codex, or startups are calculating the ROI of token use.",
      "on_screen_text_during_shot": [
        "companies",
        "adoption",
        "least",
        "there",
        "dashboard",
        "all",
        "across",
        "cloud",
        "calculating",
        "use"
      ],
      "facial_expression_or_gesture": "Explaining with subtle head nods"
    },
    {
      "end": "00:28.8",
      "index": 4,
      "start": "00:25.7",
      "camera": "Rear smartphone camera",
      "framing": "Audience POV looking towards conference stage and presentation screen",
      "movement": "Static",
      "shot_type": "wide shot",
      "transition_in": "Hard cut",
      "screen_content": "Presentation slide with benchmarks graph",
      "subject_action": "Stage presenter talking beside slides",
      "what_is_visible": "Speaker on stage beside large slide titled 'Look at task-specific benchmarks'",
      "duration_seconds": 3.1,
      "spoken_during_shot": "So tokens used per feature built.",
      "on_screen_text_during_shot": [
        "tokens",
        "built",
        "Look at task-specific benchmarks"
      ],
      "facial_expression_or_gesture": "N/A"
    },
    {
      "end": "00:40.5",
      "index": 5,
      "start": "00:28.8",
      "camera": "Front-facing smartphone camera",
      "framing": "Eye-level chest up",
      "movement": "Static handheld",
      "shot_type": "medium close-up",
      "transition_in": "Hard cut",
      "screen_content": "None",
      "subject_action": "Talking and gesturing",
      "what_is_visible": "Speaker summarizing dynamic model routing",
      "duration_seconds": 11.7,
      "spoken_during_shot": "There are other startups doing model selection for you so they can switch to different models depending on your task so it can switch to a less expensive model when your task is not complicated.",
      "on_screen_text_during_shot": [
        "startup",
        "for",
        "different",
        "task",
        "expensive",
        "complicated"
      ],
      "facial_expression_or_gesture": "Engaged and concluding"
    }
  ],
  "video": {
    "pacing": "Moderate, conversational tempo",
    "setting": "Home studio/office with guitar and floor lamp",
    "subject": "Asian creator with long dark hair and glasses speaking to camera",
    "lighting": "Warm indoor ambient lighting from side lamp",
    "color_grade": "Natural warm indoor tones",
    "orientation": "vertical",
    "aspect_ratio": "9:16",
    "overall_format": "talking-head with b-roll cutaways",
    "tone_and_energy": "Analytical, grounded, insider tech report",
    "duration_seconds": 40.5,
    "wardrobe_and_look": "Charcoal graphic t-shirt, round wireframe glasses, natural appearance"
  },
  "script": [
    {
      "end": "00:04.2",
      "line": "Companies are realizing they're burning way too much money on AI now.",
      "start": "00:00.0",
      "delivery": "Assertive, conversational",
      "uncertain": false
    },
    {
      "end": "00:12.7",
      "line": "I went to the AI conference this year and comparing to last year, most of the companies are shifting from AI adoption to token-maxxing.",
      "start": "00:04.3",
      "delivery": "Reflective, observant",
      "uncertain": false
    },
    {
      "end": "00:22.4",
      "line": "At least a quarter of the startups there were either building a dashboard to monitor all of the token use from across different platforms like Claude and Codex,",
      "start": "00:13.0",
      "delivery": "Descriptive, informative",
      "uncertain": false
    },
    {
      "end": "00:28.6",
      "line": "or startups are calculating the ROI of token use. So tokens used per feature built.",
      "start": "00:22.5",
      "delivery": "Analytical breakdown",
      "uncertain": false
    },
    {
      "end": "00:35.3",
      "line": "There are other startups doing model selection for you, so they can switch to different models depending on your task,",
      "start": "00:28.8",
      "delivery": "Clear explanatory",
      "uncertain": false
    },
    {
      "end": "00:40.5",
      "line": "so it can switch to a less expensive model when your task is not complicated.",
      "start": "00:35.4",
      "delivery": "Decisive conclusion",
      "uncertain": false
    }
  ],
  "editing": {
    "cut_count": 4,
    "transitions": "Direct jump cuts",
    "caption_style": "Minimal single-word white centered subtitles",
    "end_card_or_cta": "None",
    "average_shot_seconds": 8.1,
    "effects_and_overlays": "Static title pill overlay at start"
  },
  "teardown": {
    "hook": "Companies are realizing they're burning way too much money on AI now.",
    "format": "talking-head",
    "mechanism": "relatable_pain",
    "structure": [
      "States bold industry shift: companies burning too much money on AI tokens",
      "Presents conference attendee credential proof and reveals the shift to 'token-maxxing'",
      "Lists 3 startup categories solving token efficiency (dashboards, ROI metrics, model routers)"
    ],
    "hook_source": "transcript",
    "why_it_worked": "Taps into enterprise AI cost fatigue with relatable industry insider observations backed by conference footage."
  },
  "on_screen_text": [
    {
      "end": "00:03.9",
      "kind": "hook",
      "text": "Companies token-maxxing at AI conference",
      "start": "00:00.0",
      "style": "Black sans-serif text inside white rounded badge",
      "source": "creator",
      "position": "Top third center"
    },
    {
      "end": "00:08.8",
      "kind": "embedded",
      "text": "The AI Conference San Francisco 2024 / Lan Li Founder",
      "start": "00:05.6",
      "style": "Printed conference badge lanyard",
      "source": "creator",
      "position": "Center"
    },
    {
      "end": "00:28.8",
      "kind": "embedded",
      "text": "Look at task-specific benchmarks",
      "start": "00:25.7",
      "style": "Conference presentation slide title",
      "source": "creator",
      "position": "Center stage screen"
    }
  ],
  "structure_beats": [
    {
      "end": "00:04.2",
      "beat": "Problem Hook",
      "start": "00:00.0",
      "purpose": "Call out rising AI compute expenses"
    },
    {
      "end": "00:12.7",
      "beat": "Context & Proof",
      "start": "00:04.3",
      "purpose": "Establish conference credibility and introduce 'token-maxxing'"
    },
    {
      "end": "00:28.6",
      "beat": "Category Breakdown",
      "start": "00:13.0",
      "purpose": "Detail dashboard tracking and ROI metrics startups"
    },
    {
      "end": "00:40.5",
      "beat": "Routing Solution",
      "start": "00:28.8",
      "purpose": "Explain dynamic model selection for cost reduction"
    }
  ],
  "recreation_blueprint": {
    "shot_template": [
      {
        "slot": "Hook Statement",
        "visual": "Talking head stating an industry cost issue with top text sticker",
        "seconds": 5.6,
        "voiceover": "Bold assertion on industry spending or waste",
        "text_overlay": "[Trend name] at [Industry event]"
      },
      {
        "slot": "Proof B-Roll",
        "visual": "Cut to photo holding event badge",
        "seconds": 3.2,
        "voiceover": "Personal presence and year-over-year comparison",
        "text_overlay": "None"
      },
      {
        "slot": "Market Shift & Types",
        "visual": "Talking head breaking down startup categories",
        "seconds": 16.9,
        "voiceover": "Listing specific product types tackling the issue",
        "text_overlay": "Dynamic subtitles"
      },
      {
        "slot": "Event Stage B-Roll",
        "visual": "View of speaker presenting relevant slides on stage",
        "seconds": 3.1,
        "voiceover": "Concrete metric or formula example",
        "text_overlay": "Dynamic subtitles"
      },
      {
        "slot": "Final Solution Breakdown",
        "visual": "Talking head explaining practical implementation",
        "seconds": 11.7,
        "voiceover": "How intelligent routing or cost reduction works",
        "text_overlay": "Dynamic subtitles"
      }
    ],
    "what_can_change": [
      "Niche topic (e.g. cloud hosting costs, ad spend)",
      "B-roll environment and event type"
    ],
    "template_summary": "Founder breaks down a tech industry trend seen at a conference with badge proof and stage B-roll.",
    "what_must_be_kept": [
      "Conference credential b-roll for authority",
      "Direct problem-led opening hook",
      "Categorical breakdown of emerging solutions"
    ],
    "generation_prompts": [
      {
        "prompt": "Vertical framing, young founder talking to smartphone camera indoors, casual attire, soft natural lamp lighting, clean modern studio aesthetic.",
        "shot_index": 1,
        "negative_prompt": "Over-processed, high contrast studio flash, chaotic background."
      },
      {
        "prompt": "Smartphone selfie holding up physical conference pass badge showing attendee name and press tag, industrial ceiling background.",
        "shot_index": 2,
        "negative_prompt": "Blurry text, artificial 3d render, distorted hands."
      },
      {
        "prompt": "Vertical eye-level shot, tech creator speaking calmly into camera, subtle explanatory hand gestures, warm room background.",
        "shot_index": 3,
        "negative_prompt": "Jerky camera, oversaturated colors, dramatic expressions."
      },
      {
        "prompt": "Audience POV looking up at tech conference presentation stage with giant screen displaying data graphs, seated attendees in foreground.",
        "shot_index": 4,
        "negative_prompt": "Empty room, blurry screen, cartoon style."
      },
      {
        "prompt": "Vertical close-up of tech founder explaining software logic to camera, neutral relaxed demeanor, indoor cozy background.",
        "shot_index": 5,
        "negative_prompt": "Distorted face, harsh backlighting, noisy video."
      }
    ]
  },
  "props_apps_and_products": [
    {
      "name": "The AI Conference 2024 Press Badge",
      "how_used": "Held up as visual proof of attendance",
      "first_seen": "00:05.6"
    },
    {
      "name": "Claude / OpenAI Codex",
      "how_used": "Referenced as LLM tooling platforms",
      "first_seen": "00:21.0"
    }
  ]
}