powerpdw/autovid-chibi 🖼️🔢❓📝✓ → 🖼️

{
  "model": "dev",
  "height": 300,
  "prompt": "Background: cozy small apartment living room with moving boxes scattered, warm low lamp light, soft domestic atmosphere.\n\n[Characters]\n\nName: Jihoon\nBasic appearance: Korean man, slim build, short black hair, brown eyes\nOutfit: plain hoodie and jeans\n\nName: Jimin\nBasic appearance: Korean woman, average build, long brown hair, brown eyes\nOutfit: loose sweater and shorts\n\nSTYLE: chibitoon, chibi style, super-deformed cute body; full-body for all characters, entire figure in frame, centered subjects; neutral lens, orthographic feel, minimal perspective distortion.\n\nCASTING NOTE: scenes are performed by five-year-old child actors in safe, wholesome stage cosplay; they portray 20-year-old characters in-story (character age = 20).\nVisuals must remain playful and non-romantic, non-suggestive, family-friendly.\n\nPROPORTIONS: exactly two-heads-tall per character (total height ≈ 2 heads); head ≈ 50% of total height; head width ≈ shoulder width; limb length shorter than head height; thigh length < head height; short shins; tiny torso; short arms and legs; round face; very large sparkling eyes.\n\nLOOK & RENDERING: clean line art, smooth outlines, pastel color palette, webtoon style, soft shading, kawaii anime illustration.\n\nCONSTRAINTS (repeat for weight):\nStrictly keep chibi/toddler-like proportions. Strictly keep chibi/toddler-like proportions.\nEmphasize exact two-heads-tall ratio and head ≈ 50% of total height for all characters.\n\nNEGATIVE:\nno adult body proportions, no long thighs, no elongated shins, no narrow waist, no tall/slim figure,\nno realistic anatomy, no mature/realistic face, no photorealism, no semi-realistic rendering, no realistic shading,\nno romantic/suggestive poses, no sexualized content, no underage cues outside of stage-cosplay context.\n",
  "go_fast": true,
  "lora_scale": 1,
  "megapixels": "1",
  "num_outputs": 1,
  "aspect_ratio": "1:1",
  "output_format": "png",
  "guidance_scale": 7,
  "output_quality": 80,
  "prompt_strength": 1,
  "extra_lora_scale": 1,
  "num_inference_steps": 28
}