UGC Photo Template Identity Replacement: Structured JSON Prompt to Swap Faces While Preserving Wardrobe and Framing

AI visual creator Mira (@herprompts) shared a structured JSON prompt template on X on October 7, 2026, designed to replace a subject's facial identity while strictly preserving the pose, clothing, lighting, and framing of a reference photograph.

Image source: Mira (@herprompts) via X

In creator workflows and UGC (User Generated Content) marketing campaigns, teams frequently need to repurpose real-world photo assets. The goal is often to retain the candid smartphone texture, authentic wardrobe fit, and ambient environment of an original shot while casting a completely different virtual model. Traditional text prompts or standalone face-swap tools often struggle with this task, either regenerating random garments, altering the perspective, or introducing plastic artifacts around the facial contours.

This structured workflow solves that problem by instructing multimodal models like GPT Image 2 to treat the supplied reference as a "locked photographic template." By isolating preserved elements from altered facial anatomy across distinct JSON key-value blocks, creators can achieve reproducible identity replacement without disrupting the surrounding composition.

Complete Structured Identity Replacement JSON Prompt

Below is the full JSON prompt published by the author. When paired with an uploaded reference photo in a multimodal generation interface, it enforces strict compositional preservation while re-rendering facial geometry:

{
  "prompt_type": "reference_template_identity_replacement",

  "CORE_RULE": "Use the supplied image as a locked photographic template. Preserve the photograph as closely as possible, but replace the woman's facial identity with a completely different fictional adult woman.",

  "PRESERVE": [
    "exact pose and body position",
    "head angle and gaze direction",
    "camera angle, height, perspective and framing",
    "subject scale and crop",
    "background and environment",
    "lighting and shadows",
    "hairstyle, hair length and placement",
    "ALL clothing exactly as shown in the reference",
    "ALL accessories and jewelry exactly as shown in the reference",
    "glasses, if present",
    "clothing colors, materials, patterns, fit and silhouette",
    "jewelry type, placement and appearance",
    "overall smartphone photography aesthetic"
  ],

  "IDENTITY_CHANGE": {
    "instruction": "Create a completely different fictional adult woman. Do not preserve the reference woman's recognizable identity or facial structure.",
    "change": [
      "face shape",
      "eyes and eye spacing",
      "eyebrows",
      "nose",
      "cheekbones",
      "jawline",
      "chin",
      "lips",
      "overall facial proportions"
    ],
    "rule": "The new woman must clearly look like a different person while keeping the same photograph composition."
  },

  "CLOTHING_AND_ACCESSORIES": {
    "instruction": "Do not invent, redesign or recolor the wardrobe. Automatically copy the clothing and accessories visible in the supplied reference image.",
    "preserve": [
      "top or shirt",
      "bottoms",
      "dress or outerwear if present",
      "shoes if visible",
      "glasses if present",
      "necklaces",
      "earrings",
      "rings",
      "bracelets",
      "watches",
      "all other visible accessories"
    ],
    "important": "Whatever the woman is wearing or carrying in the reference must remain visually consistent in the generated image."
  },

  "PHOTOGRAPHY": {
    "style": "authentic candid smartphone photograph",
    "realism": "extreme photorealism",
    "skin": "natural pores and realistic imperfections",
    "processing": "natural smartphone HDR",
    "retouching": "minimal and believable"
  },

  "CLEANUP": [
    "remove text",
    "remove watermarks",
    "remove logos",
    "remove captions",
    "remove UI elements",
    "remove screen borders",
    "remove black strips or borders"
  ],

  "NEGATIVE_PROMPT": [
    "same woman",
    "same face",
    "facial duplicate",
    "recognizable identity",
    "different pose",
    "different camera angle",
    "different framing",
    "different background",
    "different clothing",
    "different accessories",
    "invented jewelry",
    "invented clothing",
    "changed clothing color",
    "missing accessories",
    "missing glasses",
    "CGI",
    "3D render",
    "plastic skin",
    "airbrushed skin",
    "anime",
    "illustration"
  ]
}

Schema Breakdown and Key Enforcement Modules

The strength of this template lies in its explicit modularization, which prevents the generative model from making creative compromises across unaffected regions.

1. Locked Photographic Template (CORE_RULE & PRESERVE)

• Template Anchoring: Declaring the input image as a locked photographic template forces the diffusion pipeline to treat camera geometry, lighting direction, and background elements as immutable ground truth.
• Physical Consistency: Locking head tilt, eye direction, subject scale, and framing ensures the synthetic replacement maintains the exact optical presence of the source image.
• Aesthetic Retention: Preserving the candid smartphone aesthetic keeps the image grounded in everyday social media realism rather than shifting into polished studio portraits.

2. Anatomical Identity Substitution (IDENTITY_CHANGE)

• Skeletal and Feature Variations: The instruction explicitly targets bone structure—jawline, cheekbones, chin, eye spacing, and nose proportions—preventing partial likenesses that look like altered versions of the original model.
• Strict Disambiguation: By mandating a distinct fictional individual, the prompt mitigates inadvertent resemblance while preserving framing continuity.

3. Wardrobe and Accessory Enforcement (CLOTHING_AND_ACCESSORIES)

• No Hallucinated Outfits: Models frequently substitute casual clothing with stylized alternatives. Explicitly forbidding redesign, recoloring, or re-styling locks in fabric types, patterns, and fit.
• Micro-Accessory Anchoring: Minor items like rings, bracelets, earrings, and watches are enumerated explicitly to ensure small details do not vanish during generation.

4. Photographic Realism and Safeguards (PHOTOGRAPHY, CLEANUP, NEGATIVE_PROMPT)

• Skin Texture Fidelity: Requiring natural pores and realistic imperfections overrides standard model tendencies toward over-smoothed, airbrushed skin.
• Artifact Removal: The CLEANUP array strips watermarks, user interface frames, captions, and black borders that might linger from screenshot references.
• Negative Exclusions: Specifying negative constraints against modified poses, altered backgrounds, CGI gloss, and 3D rendering keeps the generated output strictly photorealistic.

Original source

• X (Twitter) Post: Mira (@herprompts) on X
• Prompt Reference: Visual AI Club - ChatGPT Influencer Character Template