{
 "0": {
  "class_type": "LoadImage",
  "inputs": {
   "image": "kate.png"
  },
  "is_changed": [
   "cfb8226fc997a27ee3886170555d86a5fc8ab8b08d2d1ab2a055816649c04f7f"
  ]
 },
 "2": {
  "class_type": "CLIPLoader",
  "inputs": {
   "clip_name": "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
   "type": "minimax",
   "device": "default"
  }
 },
 "3": {
  "class_type": "VAELoader",
  "inputs": {
   "vae_name": "minimax_h3_video_vae_fp16.safetensors"
  }
 },
 "4": {
  "class_type": "VAELoader",
  "inputs": {
   "vae_name": "minimax_h3_audio_vae_fp32.safetensors"
  }
 },
 "5": {
  "class_type": "UNETLoader",
  "inputs": {
   "unet_name": "minimax_h3_ref2va_pruned_int8_convrot.safetensors",
   "weight_dtype": "default"
  }
 },
 "6": {
  "class_type": "MiniMaxH3ReferenceToVideo",
  "inputs": {
   "clip": [
    "2",
    0
   ],
   "vae": [
    "3",
    0
   ],
   "audio_vae": [
    "4",
    0
   ],
   "prompt": "subject_definitions:\n<Subject 1> is the young woman in <Picture 1>: shoulder-length dark brown wavy hair with a faint violet sheen, fair skin and blue-grey eyes, wearing a chunky oatmeal cable-knit wool sweater. Her exact facial structure, features and proportions stay identical to <Picture 1> in every frame.\n\nsummary:\n[reference generation] The target video shows <Subject 1> walking along a sunlit city sidewalk on the phone, getting bumped by a passer-by and falling backward in extreme slow motion while one continuous camera whip pans from her to her floating flip phone, coffee cup and red leather notebook and back to her, before time snaps back and she lands flat on the pavement with her belongings scattered.\n\nretention_analysis:\n<Subject 1> (appears in [Shot 1] through [Shot 3]): fully_preserved - her face, identity and clothing are held identical to <Picture 1> in every frame with zero drift.\n\ndetailed_description:\nThe target video is live-action and cinematic, a warm golden-hour city street with shallow depth of field, soft backlight through trees and a gently desaturated palette.\n[Shot 1] A tight close-up frames the torso of <Subject 1> as she walks along a red-brick sidewalk, a takeaway coffee cup with a white lid in one hand and a red leather notebook tucked under her arm; the camera tilts up with medium amplitude at medium speed to her face as she lifts a small dark clamshell flip phone to her ear, relaxed and half-smiling, while in the foreground a man in a dark jacket, seen only from behind with his back to the camera, walks away from the lens down the sidewalk straight toward her.\n[Shot 2] At 00:02.400, the shot cuts to a wide shot down the sidewalk and holds as one continuous take with no further cuts: the man in the dark jacket, his back to the camera, keeps walking away from the lens toward her as she walks toward the camera, and he walks straight into her head-on, his shoulder slamming hard into her chest; the force of the impact knocks <Subject 1> backward off her feet, arms flung wide, while he stumbles on past, the phone, the coffee cup and the notebook flying up out of her hands. Time slows to a crawl and she tips backward, hanging in the air. The camera pushes in toward her, then starts framed on <Subject 1> and whip pans left away from her to land on the flip phone turning over in mid-air in the foreground, sharp and detailed: the same small dark clamshell phone, hinged across its short top edge, the screen half above the hinge and the keypad half below it, <Subject 1> soft and out of focus behind it. The camera then whip pans right from the phone to land on the coffee cup drifting weightlessly, its lid lifting away as brown droplets of coffee hang suspended beside it. The camera then whip pans from the cup to land on the red leather notebook tumbling slowly in mid-air and pushes in with large amplitude at slow speed until its gold metal clasp fills the entire frame. The camera then whip pans back from the notebook to land on <Subject 1> in the wide shot: time snaps back to full speed and the phone, cup and notebook drop as she falls the last distance backward onto the pavement.\n[Shot 3] At 00:11.000, the shot cuts to a low ground-level static shot along the pavement: <Subject 1> lies flat on her back on the brick sidewalk, arms out, blinking up at the sky in stunned embarrassment, the same clamshell flip phone, the spilled cup and the notebook scattered around her while pedestrians walk on in the distance.\n\noverall_soundscape: Busy street ambience with distant traffic and footsteps, a muffled phone voice, a soft shoulder bump, then the sound stretches into a slow, hollow, underwater-like drone during the floating moment, snapping back to full-speed street sound with the clatter of the phone, cup and notebook hitting the pavement and a soft thud as she lands.\n\nnon_diegetic_music: A light, playful pizzicato string motif under the walk, dissolving into a suspended airy synth pad during the slow-motion float, then cut off on the landing.\n",
   "width": 1184,
   "height": 672,
   "length": 294,
   "ref_image_size": "max",
   "ref_images.ref_image_0": [
    "0",
    0
   ]
  }
 },
 "7": {
  "class_type": "RandomNoise",
  "inputs": {
   "noise_seed": 552013742
  }
 },
 "8": {
  "class_type": "KSamplerSelect",
  "inputs": {
   "sampler_name": "res_multistep"
  }
 },
 "9": {
  "class_type": "BasicScheduler",
  "inputs": {
   "model": [
    "5",
    0
   ],
   "scheduler": "simple",
   "steps": 20,
   "denoise": 1.0
  }
 },
 "10": {
  "class_type": "BasicGuider",
  "inputs": {
   "model": [
    "5",
    0
   ],
   "conditioning": [
    "6",
    0
   ]
  }
 },
 "11": {
  "class_type": "SamplerCustomAdvanced",
  "inputs": {
   "noise": [
    "7",
    0
   ],
   "guider": [
    "10",
    0
   ],
   "sampler": [
    "8",
    0
   ],
   "sigmas": [
    "9",
    0
   ],
   "latent_image": [
    "6",
    1
   ]
  }
 },
 "12": {
  "class_type": "VAEDecode",
  "inputs": {
   "samples": [
    "11",
    0
   ],
   "vae": [
    "3",
    0
   ]
  }
 },
 "13": {
  "class_type": "VAEDecodeAudio",
  "inputs": {
   "samples": [
    "11",
    0
   ],
   "vae": [
    "4",
    0
   ]
  }
 },
 "14": {
  "class_type": "CreateVideo",
  "inputs": {
   "images": [
    "12",
    0
   ],
   "audio": [
    "13",
    0
   ],
   "fps": 24.0
  }
 },
 "15": {
  "class_type": "SaveVideo",
  "inputs": {
   "video": [
    "14",
    0
   ],
   "filename_prefix": "video/higgsfield-vs-h3/floating-fall",
   "format": "auto",
   "codec": "auto"
  }
 }
}