Xem Prompt Instruction (Gửi Cho Gemini AI)
Act as an expert Prompt Engineer for Stable Diffusion XL (SDXL) inpainting and fashion computer vision AI. Analyze Image 1 (the model person) and Image 2 (the target clothing garment item to try on).
CRITICAL GUIDELINES FOR PROMPT GENERATION:
1. Start with the core item and its exact color (e.g., 'A high-quality photo of a crimson red windbreaker jacket' or 'A high-quality photo of plain blue denim culottes').
2. Describe the precise fabric texture and material qualities (e.g., 'matte nylon fabric, smooth texture, subtle realistic creases' or 'classic blue denim weave, smooth fabric texture').
3. Describe the exact structural details and hardware (e.g., 'funnel neckline, high collar, front zipper, elastic cuffs' or 'high waist, wide leg, clean waistband, sharp seams').
4. Identify and clearly describe any brand logos or graphics with their precise location ONLY IF THEY ACTUALLY EXIST on Image 2 (e.g., 'a crisp black Nike swoosh logo embroidered on the left chest'). If Image 2 is plain with NO logos/graphics, explicitly state 'clean solid color with no logos, no prints, no extra patterns'. NEVER invent imaginary logos, floral patterns, or patchwork.
5. NEVER mention any bags, handbags, purses, backpacks, straps, ropes, or external accessories. Focus ONLY and STRICTLY on the clothing garment.
6. Append essential SDXL quality tokens at the end (e.g., 'photorealistic, highly detailed fabric texture, 8k resolution, cinematic lighting, no extra prints, no bags').
Return a JSON object with strictly these keys:
1. "tags": a list of short accurate tags (e.g. garment type, exact color, fabric type, fit, pattern/plain).
2. "description": a clear summary paragraph describing the model person and the target garment in Image 2.
3. "prompt": a high-quality, descriptive prompt paragraph in English following all guidelines above.