智点AI独家秘籍:告别AI模特“变脸”烦恼!Nano Banana Pro人脸一致性终极指南

参考图文件夹结构：model_ref/├── front_neutral.jpg# 正面，中性表情├── front_smile.jpg# 正面，微笑├── side_left_45.jpg# 左侧 45°├── side_right_45.jpg# 右侧 45°└── three_quarter.jpg# 3/4 侧面（最常用构图）

❌ 错误写法（无保留指令）："A young Asian woman wearing a red dress, studio lighting, 4K"✅ 正确写法（含完整保留指令）："A young Asian woman wearing a red dress, studio lighting, 4K.Preserve exact facial structure, eye shape, jawline, and skin texturefrom reference image. Realistic skin with visible pores and texture,not smooth or plastic. Natural face proportions, slim face shape.Sharp eyes with natural highlights."

正向提示词加入："slim face, defined jawline, natural facial proportions,realistic bone structure, angular cheekbones"负向概念（在提示词末尾说明要避免的效果）：Avoid: "fat face, round puffy cheeks, bloated face,overly smooth skin, plastic skin, waxy appearance,airbrushed, over-retouched, flat eyes, doll-like"

推荐固定句式："3/4 portrait shot, eye-level camera angle,medium close-up, [focal length: 85mm portrait lens equivalent]"

Refine face detail,sharpen eyes, maintain exact facial structure,improve skin texture realism. Keep everythingelse unchanged."

import google.generativeai as genaiimport base64genai.configure(api_key="YOUR_ZHIDIANAI_KEY",# 智点AI端点，$0.05/次client_options={"api_endpoint": "vip.zhidianai.com"})model = genai.GenerativeModel("gemini-3-pro-image-preview")# 固定的面部一致性提示词模板FACE_CONSISTENCY_PROMPT="""Fashion model in {outfit}, {scene}.Preserve exact facial structure, eye shape, jawline, and skin texturefrom reference image. Natural slim face proportions, defined cheekbones.Realistic skin with visible pores, not airbrushed or plastic.Sharp clear eyes with natural highlights. 3/4 portrait shot,eye-level, 85mm portrait lens equivalent.Avoid: fat face, puffy cheeks, overly smooth skin, flat eyes.4K resolution, {aspect_ratio} aspect ratio."""defgenerate_with_face_ref(reference_image_path: str,outfit: str,scene: str,output_path: str,aspect_ratio: str="2:3"):"""使用参考图生成保持人脸一致性的模特图通过 智点AI zhidianai.com 调用，$0.05/次"""# 读取参考图withopen(reference_image_path, "rb") as f:ref_data =base64.b64encode(f.read()).decode()prompt =FACE_CONSISTENCY_PROMPT.format(outfit=outfit, scene=scene, aspect_ratio=aspect_ratio)response = model.generate_content([{"inline_data": {"mime_type": "image/jpeg","data": ref_data}},prompt],generation_config=genai.GenerationConfig(response_modalities=["IMAGE"],resolution="4K",aspect_ratio=aspect_ratio))for part in response.candidates[0].content.parts:if part.inline_data andpart.inline_data.mime_type.startswith("image/"):withopen(output_path, "wb") as f:f.write(base64.b64decode(part.inline_data.data))return output_pathreturnNone# 批量生成不同场景的模特图（使用同一参考图）scenes = [{"outfit": "white linen blazer", "scene": "minimalist studio, white background", "output": "model_studio.png"},{"outfit": "casual denim jacket", "scene": "urban street, natural daylight","output": "model_street.png"},{"outfit": "elegant evening gown", "scene": "hotel lobby, warm ambient light","output": "model_evening.png"},]for s in scenes:result = generate_with_face_ref(reference_image_path="model_ref/front_neutral.jpg",outfit=s["outfit"],scene=s["scene"],output_path=s["output"])print(f"{'✓'if result else'✗'}{s['output']}")# 3张图，智点AI 总成本 $0.15

Fashion model wearing {服装描述}, posed naturally against a cleanwhite studio background. Professional e-commerce product photography.FACE PRESERVATION (critical):- Preserve exact facial structure, bone geometry, and proportionsfrom reference image- Maintain original eye shape, eyelid fold, and iris color- Keep original jawline definition and chin shape- Realistic skin texture with visible pores, natural micro-details- Slim natural face shape matching reference — avoid any wideningPHOTOGRAPHY SETTINGS:- 3/4 portrait shot, eye-level camera angle- 85mm portrait lens equivalent, shallow depth of field- Soft diffused studio lighting, Rembrandt lighting setup- Model facing slightly left, engaging with cameraAVOID: fat face, puffy cheeks, bloated jaw, plastic skin,airbrushed appearance, flat emotionless eyes, wax figure look.Resolution: 4K, aspect ratio: 2:3

[Reference image: front-facing portrait of model]Young woman in {场景描述}, wearing {服装描述}.Lifestyle photography style, candid natural pose.IDENTITY LOCK:- This is the SAME person as in the reference image- Preserve all facial features: eye shape, nose bridge width,lip shape, jawline, cheekbone height- Do not alter face geometry in any way- Natural skin tone and texture, no over-smoothingENVIRONMENT: {场景细节}, natural daylight, golden hour lightingPOSE: {姿势描述}, natural relaxed postureQUALITY: editorial fashion photography, shot on full-frame camera,natural color grading.NEGATIVE: overweight appearance, fat face, double chin,plastic surgery look, uncanny valley, distorted anatomy.Resolution: 4K, aspect ratio: 4:5

[Input: previous generation result that needs face refinement]Refine and improve the face in this image ONLY.Do not change: background, clothing, body position, lighting.FACE IMPROVEMENTS NEEDED:1. Slim the face shape slightly, reduce cheek width by 10-15%2. Sharpen the jawline definition3. Add more realistic skin texture (visible pores, subtle imperfections)4. Improve eye clarity — add natural catch lights, deepen iris color5. Soften any plastic or over-retouched appearanceKeep all other elements 100% identical to the input image.This is a targeted face-only refinement pass.Resolution: 4K, same aspect ratio as input.