{
  "benchmark": "FARTBench",
  "test_id": "FB-C02",
  "version": "0.3",
  "title": "Multi-Agent Source Attribution",
  "canonical_spec": "FB-C02 — Multi-Agent Source Attribution\n\nTarget duration: 8 seconds\nAspect ratio: 16:9\nMode: text-to-video with native generated audio\nShot: one continuous static medium-wide live-action shot\nScene: modern office elevator, four adults visible throughout\nTarget actor: adult woman in a red jacket, rear-left from camera perspective\nEvent: exactly one brief natural fart sound at approximately 3 seconds, associated with the target actor\nTarget reaction: subtle, restrained embarrassment\nBystander reaction: the other three adults gradually recognize the sound and direct attention toward the target actor\nBackground audio: quiet elevator ventilation and faint building/mechanical ambience\nSpeech/music: none\nVisual treatment: ordinary clear air; event communicated through synchronized audio and subtle human reactions\n",
  "policy": {
    "principle": "Standardize intent, not syntax.",
    "safety_neutral_prompting": "Use clear benign positive descriptions. Do not add unrelated graphic or anatomical exclusions. If a provider refuses the benign prompt, record the moderation outcome rather than obfuscating or repeatedly rewriting around the filter.",
    "variant_rule": "Within a model family, quality/speed tiers share the exact same rendered prompt wherever the prompting architecture is the same."
  },
  "prompts": {
    "veo-3.1": "Static medium-wide photorealistic live-action shot inside a modern office elevator, showing four adults clearly throughout one continuous take. An adult woman in a red jacket stands at the rear-left from the camera's perspective. The passengers begin quietly with subtle natural movement. Around three seconds into the shot, the woman discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound. Her movement remains subtle and realistic. After the sound, she shows restrained embarrassment while the other three adults gradually recognize what happened and direct their attention toward her. Neutral commercial elevator lighting, realistic anatomy, understated performances, visually ordinary clear air.\n\nSFX: exactly one brief natural fart sound, synchronized with the woman in the red jacket and spatially associated with her position.\nAmbient noise: quiet elevator ventilation and faint building mechanical room tone.\nNo speech or music.",
    "gemini-omni-1.1-flash": "Create one single continuous, unbroken photorealistic live-action scene in a modern office elevator. Use a static medium-wide camera and keep all four adults visible throughout. An adult woman in a red jacket stands at the rear-left from the camera's perspective.\n\n[0–2.5s] The four passengers stand quietly with subtle natural movement.\n[2.5–4s] The woman in the red jacket discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound. The sound is associated with her position and her physical reaction is subtle.\n[4–8s] She shows restrained embarrassment while the other three adults gradually recognize the sound and direct their attention toward her.\n\nKeep the scene grounded, natural and understated with neutral elevator lighting and visually ordinary clear air. Sound design: quiet elevator ventilation, faint building room tone, and exactly one natural fart sound. No speech, no music, no scene cuts.",
    "seedance-2.5": "16:9 widescreen, photorealistic live action, single continuous take, static medium-wide camera. Four adults stand inside a modern office elevator and remain clearly visible throughout. An adult woman wearing a red jacket stands at the rear-left from the camera's perspective.\n\n0.0–2.5s: All four passengers stand quietly with subtle natural movement.\n\n2.5–4.0s: The woman in the red jacket discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound. The sound is associated with her position and her physical reaction remains subtle.\n\n4.0–8.0s: She shows restrained embarrassment. The other three adults gradually recognize the sound and naturally direct their attention toward her.\n\nNatural synchronized elevator ambience with quiet ventilation and faint mechanical room tone. Exactly one fart sound occurs in the scene. The event is communicated through synchronized audio and subtle human reactions; the air remains visually ordinary and clear. No speech or music.",
    "flux-3-video": "Four adults stand in a modern office elevator in a static medium-wide photorealistic live-action shot. An adult woman in a red jacket stands at the rear-left from the camera's perspective and remains clearly visible with the other three passengers. For the first few seconds everyone stands quietly with small natural movements. Around three seconds, the woman discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound. Her reaction is subtle and realistic. The other three passengers then gradually register the sound and turn their attention toward her while she shows restrained embarrassment. Keep the entire scene as one continuous take with neutral elevator lighting and visually ordinary clear air.\n\nSound: quiet elevator ventilation, faint mechanical room tone, and exactly one natural fart sound spatially associated with the woman in the red jacket. No speech or music.",
    "wan3.0-video": "Shot 1 — single continuous photorealistic live-action shot. A static medium-wide camera shows four adults inside a modern office elevator under neutral commercial lighting. An adult woman wearing a red jacket stands at the rear-left from the camera's perspective. The four passengers begin by standing quietly with subtle natural movement. Around three seconds, only the woman in the red jacket discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound. Her physical reaction is subtle. After the sound, she shows restrained embarrassment while the other three adults gradually recognize the event and direct their attention toward her. The air remains visually ordinary and clear.\n\nSound: quiet elevator ventilation and faint mechanical room tone. Exactly one brief natural fart sound is synchronized with and spatially associated with the woman in the red jacket. No speech or music.",
    "vidu-q3": "Photorealistic 8-second live-action scene in a modern office elevator, filmed as one continuous static medium-wide shot with all four adults visible. An adult woman in a red jacket stands at the rear-left from the camera's perspective. During the opening few seconds the passengers stand quietly with subtle natural movement. Around three seconds, the woman in the red jacket discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound. The sound is associated with her position. She reacts only slightly, then shows restrained embarrassment. The other three adults gradually recognize the sound and direct their attention toward her. Keep the reactions natural and understated, with ordinary clear elevator air.\n\nAudio: quiet elevator ventilation, faint room tone, and exactly one natural fart sound. No speech or music.",
    "ltx-2.5": "A static medium-wide photorealistic live-action shot holds continuously inside a modern office elevator, keeping four adults visible throughout. An adult woman wearing a red jacket stands at the rear-left from the camera's perspective. For roughly the first two and a half seconds, everyone stands quietly with subtle natural movement. Around three seconds, the woman in the red jacket discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound associated with her position and accompanied by only a subtle physical reaction. She then shows restrained embarrassment as the other three adults gradually recognize the sound and turn their attention toward her. The elevator remains visually ordinary with neutral commercial lighting and clear air. The continuous soundscape is quiet elevator ventilation and faint mechanical room tone, interrupted once by the single natural fart sound. No speech or music.",
    "minimax-h3-hosted": "Create a photorealistic live-action single continuous shot inside a modern office elevator. A static medium-wide camera keeps four adults clearly visible for the full scene. An adult woman in a red jacket stands at the rear-left from the camera's perspective. At first, everyone stands quietly with subtle natural movement. Around three seconds, the woman in the red jacket discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound associated with her position and synchronized with a very subtle physical reaction. She then shows restrained embarrassment while the other three adults gradually recognize the sound and naturally direct their attention toward her. Keep all performances understated and physically realistic. The air stays visually ordinary and clear. The soundscape is quiet elevator ventilation and faint building ambience, with exactly one natural fart sound and no speech or music.",
    "minimax-h3-direct-structured": {
      "integrated_multimodal_description": "[Shot 1] Photorealistic live-action single continuous shot. A static medium-wide camera frames four adults inside a modern office elevator under neutral commercial lighting, with all four clearly visible. An adult woman wearing a red jacket stands at the rear-left from the camera's perspective. For the opening portion of the eight-second shot, all four passengers stand quietly with subtle natural movement. Around three seconds, the woman discreetly passes gas, producing exactly one brief natural fart sound associated with her position and synchronized with a subtle physical reaction. After the sound, she shows restrained embarrassment while the other three adults gradually recognize it and direct their attention toward her. Performances remain understated and physically realistic. The air remains visually ordinary and clear.",
      "overall_soundscape": "Quiet elevator ventilation and faint building mechanical ambience continue throughout. Exactly one brief, natural, moderately audible fart sound occurs around three seconds and is associated with the woman in the red jacket. No speech or other prominent vocal sounds.",
      "non_diegetic_music": "N/A"
    },
    "kling-3": "Single continuous photorealistic live-action shot in a modern office elevator. A static medium-wide camera keeps four adults clearly visible. An adult woman in a red jacket stands at the rear-left from the camera's perspective.\n\nShot 1 (0–3s): The passengers stand quietly with subtle natural movement.\nShot 1 (around 3s): The woman in the red jacket discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound. The sound is associated with her position and her physical response remains subtle.\nShot 1 (3–8s): She shows restrained embarrassment while the other three adults gradually recognize the sound and naturally direct their attention toward her.\n\nNeutral commercial lighting, realistic movement, understated performances, visually ordinary clear air. Native audio: quiet elevator ventilation and faint room tone, with exactly one natural fart sound. No speech or music.",
    "grok-imagine-video-1.5": "A single continuous photorealistic live-action shot inside a modern office elevator, static medium-wide framing, with four adults visible throughout. An adult woman wearing a red jacket stands at the rear-left from the camera's perspective. The passengers stand quietly for the opening few seconds with subtle natural movement. Around three seconds, the woman in the red jacket discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound. The sound clearly belongs to her position and her physical response is subtle. She then looks mildly embarrassed while the other three passengers gradually register the sound and turn their attention toward her. Keep the pacing restrained and realistic, with neutral elevator lighting and visually ordinary clear air. Sound design: quiet elevator ventilation, faint mechanical ambience, and exactly one natural fart sound. No speech or music.",
    "happyhorse-1.1": "Photorealistic modern office elevator, four adults, static medium-wide camera, one continuous 8-second live-action take. An adult woman in a red jacket stands at the rear-left from the camera's perspective and all four people remain visible.\n\n0–2.5s: Everyone stands quietly with subtle natural movement.\n2.5–4s: The woman in the red jacket discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound. Her reaction is physically subtle and the sound is associated with her position.\n4–8s: She shows restrained embarrassment while the other three adults gradually recognize the sound and naturally direct their attention toward her.\n\nNeutral commercial lighting, realistic anatomy and understated performances, visually ordinary clear air. Audio: quiet elevator ventilation, faint mechanical room tone, exactly one natural fart sound, no speech or music.",
    "pixverse-v6": "A photorealistic modern office elevator in one continuous static medium-wide live-action shot. Four adults remain visible throughout. An adult woman wearing a red jacket stands at the rear-left from the camera's perspective. The passengers initially stand quietly with subtle natural movement. Around three seconds, the woman in the red jacket discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound associated with her position. Her reaction stays subtle. She then shows restrained embarrassment while the other three adults gradually recognize the sound and direct their attention toward her. Keep the scene physically realistic, naturally paced, and visually ordinary with clear air and neutral commercial lighting. Generate synchronized audio with quiet elevator ventilation, faint building ambience, and exactly one natural fart sound. No speech or music.",
    "pixverse-c1": "Single-shot cinematic realism, not a multi-shot storyboard. One continuous static medium-wide live-action view inside a modern office elevator keeps four adults visible throughout. An adult woman in a red jacket stands at the rear-left from the camera's perspective. The passengers begin quietly with subtle natural movement. Around three seconds, the woman discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound associated with her position and a subtle physical response. She then shows restrained embarrassment while the other three adults gradually register the sound and direct their attention toward her. Keep the performances understated, the motion physically realistic, and the air visually ordinary and clear. Synchronized audio: quiet elevator ventilation, faint mechanical room tone, and exactly one natural fart sound. No speech or music.",
    "sora-2": "A photorealistic 8-second live-action video inside a modern office elevator, filmed as one continuous static medium-wide shot. Four adults remain clearly visible throughout. An adult woman wearing a red jacket stands at the rear-left from the camera's perspective. The passengers begin by standing quietly with subtle natural movement. Around three seconds, the woman discreetly passes gas, producing exactly one brief, natural, moderately audible fart sound associated with her position. She reacts subtly, then shows restrained embarrassment. The other three adults gradually recognize the sound and turn their attention toward her. Neutral commercial lighting, realistic body movement, understated performances, visually ordinary clear air. Quiet elevator ventilation and faint mechanical ambience continue underneath exactly one natural fart sound. No speech or music."
  },
  "configurations": [
    {
      "family": "Google Veo 3.1",
      "configuration": "Veo 3.1",
      "model_id": "veo-3.1-generate-001",
      "prompt_key": "veo-3.1",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; native audio"
    },
    {
      "family": "Google Veo 3.1",
      "configuration": "Veo 3.1 Fast",
      "model_id": "veo-3.1-fast-generate-001",
      "prompt_key": "veo-3.1",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; native audio"
    },
    {
      "family": "Google Veo 3.1",
      "configuration": "Veo 3.1 Lite",
      "model_id": "veo-3.1-lite-generate-001",
      "prompt_key": "veo-3.1",
      "status": "If accessible",
      "settings": "8s; 16:9; 720p; native audio"
    },
    {
      "family": "Google Gemini Omni",
      "configuration": "Gemini Omni 1.1 Flash",
      "model_id": "gemini-omni-1.1-flash",
      "prompt_key": "gemini-omni-1.1-flash",
      "status": "API-first",
      "settings": "single continuous shot; target ~8s; native audio"
    },
    {
      "family": "ByteDance Seedance",
      "configuration": "Seedance 2.5",
      "model_id": "provider-specific",
      "prompt_key": "seedance-2.5",
      "status": "API-first",
      "settings": "8s; 16:9; native audio"
    },
    {
      "family": "Black Forest Labs FLUX 3",
      "configuration": "FLUX 3 Full",
      "model_id": "flux-3-video",
      "prompt_key": "flux-3-video",
      "status": "API-first",
      "settings": "8s; 16:9; HD; full render; native audio"
    },
    {
      "family": "Black Forest Labs FLUX 3",
      "configuration": "FLUX 3 Draft",
      "model_id": "flux-3-video",
      "prompt_key": "flux-3-video",
      "status": "API-first",
      "settings": "8s; 16:9; HD; draft mode; native audio"
    },
    {
      "family": "Alibaba Wan 3.0",
      "configuration": "Wan 3.0",
      "model_id": "wan3.0-video",
      "prompt_key": "wan3.0-video",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; audio on; prompt extension off if exposed"
    },
    {
      "family": "Alibaba Wan 3.0",
      "configuration": "Wan 3.0 Prime",
      "model_id": "wan3.0-video-prime",
      "prompt_key": "wan3.0-video",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; audio on; prompt extension off if exposed"
    },
    {
      "family": "Vidu Q3",
      "configuration": "Vidu Q3 Pro",
      "model_id": "viduq3-pro",
      "prompt_key": "vidu-q3",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; audio=true"
    },
    {
      "family": "Vidu Q3",
      "configuration": "Vidu Q3 Turbo",
      "model_id": "viduq3-turbo",
      "prompt_key": "vidu-q3",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; audio=true"
    },
    {
      "family": "LTX-2.5",
      "configuration": "LTX-2.5 Pro",
      "model_id": "ltx-2-5-pro",
      "prompt_key": "ltx-2.5",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; generate_audio=true"
    },
    {
      "family": "LTX-2.5",
      "configuration": "LTX-2.5 Fast",
      "model_id": "ltx-2-5-fast",
      "prompt_key": "ltx-2.5",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; generate_audio=true"
    },
    {
      "family": "MiniMax H3",
      "configuration": "H3 hosted / 768P",
      "model_id": "provider-specific",
      "prompt_key": "minimax-h3-hosted",
      "status": "API-first",
      "settings": "8s; 16:9-ish; 768P; native audio; record Context-IR use"
    },
    {
      "family": "MiniMax H3",
      "configuration": "H3 hosted / 2K",
      "model_id": "provider-specific",
      "prompt_key": "minimax-h3-hosted",
      "status": "Optional quality tier",
      "settings": "8s; 16:9-ish; 2K; native audio; record Context-IR use"
    },
    {
      "family": "Kuaishou Kling 3",
      "configuration": "Kling VIDEO 3.0 Omni",
      "model_id": "provider-specific",
      "prompt_key": "kling-3",
      "status": "API-first",
      "settings": "8s; 16:9; ~720p; native audio"
    },
    {
      "family": "Kuaishou Kling 3",
      "configuration": "Kling VIDEO 3.0",
      "model_id": "provider-specific",
      "prompt_key": "kling-3",
      "status": "If exposed",
      "settings": "8s; 16:9; ~720p; native audio if supported by route"
    },
    {
      "family": "Kuaishou Kling 3",
      "configuration": "Kling VIDEO 3.0 Turbo",
      "model_id": "provider-specific",
      "prompt_key": "kling-3",
      "status": "API-first if available",
      "settings": "8s; 16:9; ~720p; native audio"
    },
    {
      "family": "xAI Grok Imagine",
      "configuration": "Grok Imagine Video 1.5",
      "model_id": "grok-imagine-video-1.5",
      "prompt_key": "grok-imagine-video-1.5",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; generated audio on"
    },
    {
      "family": "xAI Grok Imagine",
      "configuration": "Grok Imagine Video 1.5 Fast",
      "model_id": "product-only / no public API model id verified",
      "prompt_key": "grok-imagine-video-1.5",
      "status": "Defer unless selectable",
      "settings": "same semantic prompt; record product route and actual settings"
    },
    {
      "family": "HappyHorse",
      "configuration": "HappyHorse 1.1",
      "model_id": "happyhorse-1.1",
      "prompt_key": "happyhorse-1.1",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; audio on"
    },
    {
      "family": "PixVerse",
      "configuration": "PixVerse V6",
      "model_id": "v6",
      "prompt_key": "pixverse-v6",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; generate audio on"
    },
    {
      "family": "PixVerse",
      "configuration": "PixVerse C1",
      "model_id": "c1",
      "prompt_key": "pixverse-c1",
      "status": "API-first",
      "settings": "8s; 16:9; 720p; generate audio on"
    },
    {
      "family": "OpenAI Sora (Legacy)",
      "configuration": "Sora 2",
      "model_id": "sora-2",
      "prompt_key": "sora-2",
      "status": "Legacy capture if access remains",
      "settings": "8s; 16:9 / 1280x720; preserve raw output"
    },
    {
      "family": "OpenAI Sora (Legacy)",
      "configuration": "Sora 2 Pro",
      "model_id": "sora-2-pro",
      "prompt_key": "sora-2",
      "status": "Legacy capture if access remains",
      "settings": "8s; 16:9 / 1280x720; preserve raw output"
    }
  ]
}