MAYAGI · L7825 · MODEL GENERATION
提示词 / 参考图 / 参数 / 输出 · 2026-05-14
| API | https://api.apiyi.com/v1 |
|---|---|
| Model | gpt-image-2-vip · 不限并发 · 固定 $0.03/张 |
| Endpoint | client.images.edit(...) 多图 edit (4 张 ref) |
| Size | 2048x3072 (3:4 supersampling,接近 vip 上限) · 然后 PIL Lanczos downscale 到 1024×1536 |
| Quality | vip 不分 quality 档位 |
| 引用图数 | 4 张 (image[0] = wife anchor + image[1-3] = 鞋多角度) |
| Pose ref | ⛔ 不传图,全部用文字 pose_text + setting_text 描述(避免 image[0] attention 被稀释) |
| 跑数 | 4 张 task 并发 · ThreadPoolExecutor max_workers=4 |
| 耗时 | 190s / 198s / 250s / 694s (max_workers 排队) |
| 成本 | vip × 4 = $0.12 / 单 SKU 全套 model |
⭐ Image 1: SUBJECT IDENTITY ANCHOR (face / hair / bone structure ONLY). EXTRACT ONLY from image 1: her face (eyes, eyebrows, nose, lips, jawline, cheekbones), her hair color and length (long dark with subtle waves), her general bone structure and body proportions. ⛔ IGNORE everything else in image 1: her outfit/clothing, her accessories, her background, her current pose, her current facial expression. The outfit, pose, expression, and background must come fresh from the [Composition] / [Subject] / [Background] / pose_text slots below — NOT from image 1. ⭐ SKIN TRANSLATION: Image 1 may look polished or studio-finished. You must TRANSLATE her identity into a real candid iPhone snapshot moment — natural skin texture with visible micro pores, faint specular highlights, subtle uneven tone, real-life imperfection. DO NOT carry over the smooth/polished anchor finish. Image 2: SHOE GROUND TRUTH (90° side view) — pixel-perfect preserve. Image 3: SHOE STRUCTURE (top-down or 3/4) — strap layout, buckle count, footbed shape. Image 4: SHOE MATERIAL DETAIL — {HEEL_MATERIAL}. Preserve material exactly: {SHOE_MATERIAL}. Do NOT reinterpret. ═══════════════════════════════════════════════════════ [Photography Style] Photorealistic candid iPhone 15 Pro snapshot — amateur smartphone photography aesthetic, EXIF feel (26mm wide lens, f/1.8, slight handheld micro-blur), natural ambient daylight, real skin texture with visible micro pores and natural specular highlights, faint freckle hints, subtle uneven skin tone. NO professional retouching, NO beauty filter, NO airbrushing, NO AI plastic skin, NO digital over-sharpening. Honest unposed quality, like a moment captured between shots. [Composition] 3:4 portrait, 2048x3072 high resolution. {framing}. Eye-level camera at conversational distance. {pose_text}. Gaze averted, looking off-frame / down / sideways — NEVER looking at camera, NEVER eye contact, NEVER smiling, lips gently closed (no teeth). [Subject] A woman in her late twenties (face identity matches Image 1 ONLY — outfit and expression are fresh, NOT from image 1). She wears {colorway.upper()} T-strap sandals on her feet (matched to images 2-4 exactly). OUTFIT (creative brief — model freely interprets within these boundaries, ⛔ DO NOT copy outfit from Image 1): - Aesthetic: quiet luxury, Lemaire SS26 / Toteme / The Row / Khaite editorial - Color palette: muted earthy neutrals only (cream, taupe, charcoal, navy, brown, oat, dusty olive) - Cohesion: outfit pairs naturally with {colorway} T-strap sandals - Banned: bright/saturated colors, logos, prints, streetwear, Y2K, frilly/lace, tight cleavage tops, AND the specific garment shown in Image 1 (use a different fresh outfit) [Background] {setting_text}. Warm cream plaster wall (#F5F2EC base tone), gentle directional ambient daylight from upper-left (11 o'clock), max 5-8% tonal range — readable but never dramatic, NO bright hotspot, NO dark vignette. Minimalist negative space. NO clutter. ⛔ DO NOT use the background from Image 1. [Constraints] PRESERVE list (every iteration): face identity from Image 1 (face only); shoe silhouette + material + heel + hardware + colorway from Images 2-4; averted gaze; calm closed-lip expression. NEVER: looking at camera / direct gaze / smile / teeth showing / influencer pose / duck face. NO: text, watermark, logo, brand label. NO: plastic skin, airbrushing, AI yellow tint, ChatGPT-typical orange cast, beauty filter symmetry. DO NOT alter shoe details from Images 2-4. ⛔ DO NOT copy outfit, accessories, expression, or background from Image 1 — only the face identity carries over.
SHOE_MATERIAL = "matte suede upper + patent leather T-strap mix"
HEEL_MATERIAL = "tortoiseshell amber acrylic block heel — translucent
with brown veining, NOT solid black plastic, NOT wood,
NOT metal"
blackblack_side + black_top_down + black_heel_close
blackblack_side + black_top_down + black_heel_close
brownbrown_side + brown_3q + brown_dual
brownbrown_side + brown_3q + brown_dual