平时用 AI 做图、做视频时，我会把一些网友做的效果不错、值得留存的提示词收集起来。这次我集中整理了 30 条有用的提示词，涉及人物、产品、品牌、海报、风格化重绘等多个方向，分享给大家。希望能给你一些灵感，也欢迎收藏备用。
使用方法：复制代码块里的提示词，粘贴到你常用的 AI 生图/生视频工具里；带【】或 [ ] 的地方换成你自己的内容，需要上传参考图的会在提示词里写明。文中提到的工具有 GPT-Image-2、海螺 H3（MiniMax）、Seedance 2.0，其他条目可以在你常用的模型上试。
AI视频（5条）
1. 苹果风“白菜 Pro”产品广告片
提示词
5 seconds, 16:9 landscape format, fully adopting the visual language of Apple’s new product launch videos: pure black background, studio-grade lighting, ultra-slow rotation close-ups, minimalist oversized sans-serif titles, ethereal electronic soundtrack.
The product is—a head of ordinary Chinese cabbage.
[0–4 seconds] Against a pure black background, a plump, fresh head of Chinese cabbage slowly hovers and rotates into view, its outer leaves vibrant green, the inner core a pale yellowish-green, edges rimmed with cool white contour lighting, tiny water droplets on the leaf surfaces slowly rolling under macro lens focus. The frame feels premium, restrained, serene, as if unveiling a top-tier tech innovation. White, slender large text fades in: “Cabbage Pro” with smaller text below: “Remarkably Fresh.”
[4–8 seconds] Quick rhythmic cuts through three studio-grade close-up shots: the intricate layered textures of a cabbage cross-section in macro detail; the sculptural contours of outer leaf curves and folds under side backlight; delicate mist sprayed onto the cabbage surface, water droplets clinging to the veins in ultra-slow motion. Each shot reveals a minimalist large text title: “Plump.” “Crisp & Tender.” “Clean & Sweet.”
[8–12 seconds] The cabbage elegantly flips in mid-air, outer leaves subtly unfurling, steadied and gently rotated by a hand that trails restrained, premium light-streak effects, as the music builds to a grand crescendo. The shot maintains Apple’s signature product hero vibe, like showcasing revolutionary flagship hardware. Large text: “The Most Powerful Cabbage in History.” Smaller text below: “20% Crisper Than the Previous Generation.”
[12–15 seconds] Music fades to a close, leaving only the cabbage quietly upright and suspended at the center of the pure black frame, its contours sharp, lighting premium, ample negative space. A price tag appears beside it: “Starting at ¥3” Finally, the tagline hits: “Cabbage Pro. Rolling Out Great Flavor.” Paired with a clean, restrained end chime.
适用模型/参数：海螺 H3（MiniMax）
2. 一镜十姿：同一人物连续十个时装姿态
提示词
Upload Image 1 as the sole reference for the character’s identity and clothing.
Generate a 15-second, 9:16 vertical, fashion magazine editorial-quality continuous performance video.
The character must strictly maintain Image 1’s facial structure, features, hairstyle, skin tone, age, body type, and clothing structure. The scene is a minimalist professional photography studio, with background, floor, lighting positions, and character positioning kept stable.
The character sequentially completes ten fashion poses with distinct silhouette differences:
1. Frontal standing
2. Shoulder lean forward
3. One hand on hip
4. Body side turn
5. Looking back gaze
6. Crossing step
7. Light swing of clothing hem
8. Low crouch
9. Approach toward the camera
10. Final cover pose freeze
Each pose naturally grows out of the previous action: turning must go through center of gravity shift, stepping must complete foot landing, after swinging the clothing hem, retain realistic inertia. Key poses have a brief pause of 0.3–0.5 seconds to let viewers clearly see the body silhouette.
The camera uses a 50mm lens for stable, slow retreat along the Z-axis, always maintaining a reasonable distance from the character. Use photographic flash, clothing hem sweeping across the lens, hair flick, and quick turns to create rhythmic emphasis, but do not switch scenes.
Hard side lighting outlines body lines, soft frontal fill light preserves facial details. Realistic fabric folds, hair inertia, shoe sole contact shadows, and natural motion blur, overall with a high-fashion editorial feel.
Prohibit face swapping, age drift, clothing color change, clothing structure recombination, limb duplication, joint inversion, foot sliding, character teleportation, background jumping, meaningless circling, ten-image-style hard cuts, and restarting each action from scratch.
💡 Core Logic
One Shot, Ten Poses is built from three locks:
Identity lock, ensuring it’s always the same person; Clothing lock, proving it’s the same shoot; Action chain, making the ten poses part of one continuous performance.
The truly advanced part isn’t the number of poses, but that the power of the previous action can carry over to the next one.
⚙️ Key Control Points
First test identity stability with four to six poses, then expand to ten. Between poses, change at least two of: arm silhouette, body orientation, or center of gravity position—don’t just slightly shift the hand.
If producing static images, extract six 3:4 independent key frames from the video; the six images keep the character and clothing consistent, but cover at least frontal, side, back view looking over shoulder, low angle, close-up, and full-body six types of framing relationships.
During acceptance, focus on checking three positions: the face during turns, the legs during crossing steps, the fingers during clothing hem swings. These three moments are most prone to identity and limb errors.
适用模型/参数：Seedance 2.0
3. 塑料袋的奢侈品广告片
提示词
30 seconds, 16:9 landscape, fully adopting the visual language of top French luxury brand advertising: minimalist high-end, French restraint, equestrian metaphors, sculptural studio lighting, orange brand temperament, ultra-slow-motion material close-ups, elegant yet slightly absurd high-end humor. The advertising protagonist is not leather, not a silk scarf, but the most ordinary, cheapest semi-transparent orange plastic bag. The entire film must treat this plastic bag with utmost seriousness and solemnity, never shooting it as a comedic skit. The more high-end, the more restrained, the stronger the contrast. Music uses minimalist French experimental soundtrack: low cello, subtle leather-friction-like percussion, airy electronic tones, horseshoe rhythm samples. No voiceover, mainly relying on visuals and minimal text to convey the message.
【Visual Control】 The entire film should not use cheap advertising lights, no street-level plastic bag documentary feel, no comedic performances, no exaggerated antics.
【Cinematography Style】 French high fashion advertising; Top luxury still-life photography; Combination of 35mm and 85mm cinema lenses; Extensive 100mm macro material close-ups; Ultra-shallow depth of field; Extremely slow push-ins; Ultra-slow motion; Large areas of negative space; Extremely restrained editing.
Core colors only: Deep black; Orange; Warm brown; Ivory white; Minimal metallic silver.
Prohibit neon lights, cyberpunk, tech UI, excessive particles, cheap golden light effects.
The core temperament of the entire advertisement: “We are clearly just filming a plastic bag, but everyone must believe, this might be the most worthy plastic bag in the world to be watched seriously.”
4. 景点鸟瞰图→无人机第一视角导览
提示词
The image I gave you is an aerial view of [Gulangyu, Xiamen]. I want you to follow the order of the red line in the image, using a drone filming style, first-person perspective, with fast spatial traversal for the tour. The video should not include the red line, no text should appear at all—just film along the route.
适用模型/参数：Seedance 2.0；需两张参考图：景点鸟瞰图 + 标好游览路线（红线）的图
5. 高奢工坊风：中国传统刺绣广告片
提示词
30 seconds, 16:9 landscape format, top-tier Western haute couture and atelier advertising style, referencing Dior Haute Couture, Chanel Métiers d’Art, Valentino Haute Couture, Hermès craftsmanship footage, using the cinematic language of international luxury brands to capture high-end customization and atelier details, filming traditional Chinese embroidery. Overall aesthetic: opulent, exquisite, restrained, expensive, elegant, mysterious—like a true haute couture craftsmanship short film. Avoid clichéd ancient aesthetics, cheap national trend vibes, tourist promotional feel, or excessive imperial piling.
The protagonist is a close-up of an exquisitely beautiful traditional Chinese embroidery artwork, suggesting peony, phoenix, cloud patterns, and ruyi grass motifs as the main visual elements, embroidered on high-end silk base fabric in deep teal green, peacock blue, or ink blackish-green.
Embroidery materials include: richly layered colored silk threads; gold thread outlining; local coiled gold or metallic-effect embroidery threads; fine, dense stitching; three-dimensional petal layering; silk sheen on feathers and petals.
The entire piece maintains the same single embroidery artwork throughout—do not switch motifs too much; focus on shooting its close-ups, craftsmanship, sheen, and aura. Avoid turning it into a mundane flat display; film it like high-end fashion fabrics and atelier treasures.
Lighting focus: warm golden key light; soft side backlighting; macro highlight sweeps; silk thread highlights flickering; gold thread reflected light; dramatic chiaroscuro in a dark background; contrast between the silk base fabric and the embroidery’s three-dimensional layering.
The visuals must emphasize: the delicate sheen of silk threads under light; the order and rhythm of the stitching; the noble elegance of gold thread edges; the three-dimensionality of petal and feather motifs; the embroidery’s opulence like “wearable jewelry”; the world-class beauty of Eastern craftsmanship interpreted through Western luxury syntax.
All subtitles use bilingual Chinese-English with a design sensibility. Chinese: high-end calligraphic title font + modern thin secondary copy. English: elegant haute serif typeface. Layout: restrained, with generous negative space—like true luxury brand posters and film title systems. Subtitle colors suggested: warm gold, antique bronze gold, ivory white.
Music: ethereal strings + subtle piano + faint metallic overtones + soft low-frequency ambiance. Overall, like an atelier short film before a haute couture show—elegant, sumptuous, serene. No voiceover.
人像与自拍玩法（9条）
6. 定格动画黏土人像
提示词
Convert this portrait into a stop-motion animation style clay character. Retain the character’s original facial features, expression, and pose, but make the overall look like it’s hand-molded from colorful clay. The skin and clothing have a soft, rounded clay texture, with visible subtle carving marks and handmade textures. Use warm studio lighting, clean background, and give the image a cute, healing, handmade animation feel.
The core structure is: Original portrait + Retain similarity + Clay material + Hand-molded marks + Warm lighting + Clean background.
Why does it easily reduce the “AI flavor”?
Because clay itself isn’t a perfect material. It doesn’t require hyper-realistic skin or perfectly precise edges. Slight irregularities, handmade marks, and material imperfections actually make the image more believable.
Oftentimes, the problem with AI images isn’t that they’re not refined enough—it’s that they’re too refined. The value of stop-motion animation style is in deliberately breaking up that “perfection” just a bit.
How can you vary it?Stop-motion family portrait: Turn a family into clay characters—perfect for parent-child, family, or commemorative photos.
Stop-motion pet: Cats and dogs don’t need realistic fur; focus on roundness, cuteness, and expressiveness.
Stop-motion food: Burgers, cakes, pizzas, ice cream—all suit the clay texture perfectly.
Stop-motion couple: Two people drinking tea, watching a movie, or cooking in a miniature house—the coziness will be strong.
Stop-motion villain: Add a dark background, low-angle lighting, and exaggerated expressions to create a clay animation villain origin vibe.
A few practical tips:
To retain the person’s features, add: Retain facial similarity, retain original expression, retain character’s pose.
For more handmade feel, add: Visible hand-molded marks, clay carving textures, slightly irregular edges.
For more healing vibe, add: Warm lighting, soft colors, rounded shapes, clean background.
For more cinematic feel, add: Miniature scenes, dramatic lighting, stop-motion animation film texture, storytelling composition.
Universal formula: Convert [subject] into a stop-motion animation style clay character, retain [key features], add [clay material], [handmade marks], [warm lighting], and [miniature scenes], overall cute, healing, with a handmade animation feel.
7. 自拍+商品图虚拟试穿
提示词
Based on this selfie photo and this product image, generate an “effect image of the person wearing that clothing.”
Requirements:
- Retain the person’s original facial features.
- Retain the person’s original body proportions and overall vibe.
- Accurately present the clothing’s fit, color, material, and drape.
- Generate multiple angles: front, side, half-body close-up.
- Clean image, studio lighting, simple background.
- Overall effect realistic and natural, like a professional fitting photo.
If there are multiple items, you can generate multiple outfit combinations separately.
If you want to make it even clearer, you can phrase it as:
Image 1: Person’s selfie
Image 2: Product image
Based on Image 1 and Image 2, generate an effect image of the person trying on that product.
Retain the facial features, hairstyle, skin tone, and body proportions from Image 1, and accurately dress the person in the clothing from Image 2.
Output the following angles:
1. Front full-body.
2. Side full-body.
3. Half-body close-up.
Requirements: Clothing details realistic, fabric and contours natural, image suitable for e-commerce display or outfit reference.
The core structure of this kind of prompt is: Selfie as the person’s base + product image as the clothing source + specified angles + lock in “what to retain” + designated output purpose.
💡 Core Logic
The key to this technique isn’t just “swapping one outfit.” It’s letting AI establish a realistic relationship between “the person” and “the clothes.”
👤 First, lock in the person
Be explicit with the AI: don’t change the face, don’t mess with the body proportions. That way, it focuses on the “try-on effect” instead of generating a whole new stranger model.
👕 Second, be clear about the clothing
If the product image is just a top, have it replace only the top. If it’s a full outfit, clearly specify which pieces are included. Otherwise, the model might get creative and add extras you don’t want.
📐 Third, angles determine practicality
Front view for checking overall proportions. Side view for fit and contours. Half-body close-up for neckline, fabric, and details.
适用模型/参数：GPT-Image-2
8. 自拍→杂志大片（AI影棚重绘）
提示词
Please base this on the selfie I uploaded and generate a high-fashion magazine-style portrait. Retain the subject’s authentic facial features, face shape, hairstyle, and personal vibe, but reimagine it with professional studio lighting, a clean high-end background, shallow depth of field, soft skin texture, and cinematic color grading. Overall, make it look like a magazine cover shot taken by a pro photographer—refined lighting and shadows, elevated composition, natural and confident subject, without overdone skin smoothing or editing.
🔁 How Can You Play With It?
💼 Professional Headshot
Perfect for resume photos, LinkedIn avatars, or business intro pages. Add keywords like: professional, clean background, business portrait, high-end gray tones.
🎨 Artistic Portrait
Try directions like vintage, black-and-white, film grain, sci-fi, or cinematic—great for a series of profile pics.
💍 Wedding / Couples Preview
Test styles, lighting, and compositions with AI first, then finalize your real shoot plan.
📱 Social Media Avatar Set
Turn one selfie into varied styles: magazine vibe, cinematic vibe, professional vibe, street-style vibe—for a cohesive personal brand visual kit.
🛍️ Product Model Shots
Use a subject’s selfie to create cleaner product showcase images, ideal for fashion, accessories, or lifestyle brand previews.
⚙️ A Few Practical Tips
Want it more like a magazine cover? Add: Fashion magazine cover, professional studio lighting, shallow depth of field, high-end commercial photography.
Want it more natural? Add: Retain authentic facial features, natural skin texture, no excessive smoothing, authentic personal vibe.
Want it more upscale? Add: High-end gray background, soft rim lighting, cinematic color grading, minimalist composition.
Avoid words like “beautify,” “smooth skin,” or “make it prettier.” These can easily turn it into generic edited vibes.
Better to use: Redraw, professional photography, magazine cover shot, studio lighting, retain authentic features.
📌 Usage Notes
AI studio redrawing is fundamentally about “recreation,” not fine-tuning the original image.
So after multiple rounds of tweaks, you might notice subtle shifts: like moles fading, eyes becoming more refined, hairline changes, or features slightly idealized.
To keep it true to you, emphasize repeatedly: Retain authentic facial features, retain face shape, retain hairstyle, retain personal vibe.
适用模型/参数：GPT-Image-2
9. 和小时候的自己合影
提示词
📝 Core Prompt Words
Please use the two photos I uploaded: one of me as a child, and one of me now. Generate a realistic photography-style image: my younger self sitting on the left, my current self sitting on the right, the two of us facing each other at a warm wooden table, with a small birthday cake or childhood toy placed in the middle. Keep the facial features of both people authentic and natural, without over-beautifying. The image should use soft natural light, slight film grain, a real family photo album texture, and warm nostalgic tones. The overall atmosphere should be quiet, tender, like a genuine photo spanning time.
🔁 You Can Change the Scene Like This
Birthday Cake Version
Your younger self and current self sitting on either side of the cake, like a birthday that’s long overdue.
Old Room Version
Change the background to a childhood bedroom, with a gaming console, paintbrushes, old computer, and comic books on the table.
Cinematic Version
Add window backlight, shallow depth of field, low saturation tones, and a real movie still texture.
Black-and-White Photography Version
Make it a black-and-white documentary portrait, more like a time archive.
Fairy Tale Version
Your current self squats down, gently looking at your younger self, with the image like a tender sci-fi fairy tale.
⚙️ Image Generation Tips
For the childhood photo, try to choose one where the face is clear. For the current photo, it’s best to use a front-facing or half-body shot. Don’t have too exaggerated differences in angles between the two photos, or it might lead to failed face fusion.
For more realism, you can add: Family photo album texture, natural light, slight film grain, authentic skin texture, no excessive beautification.
For more tenderness, you can add: Warm desk lamp, old room, childhood toys, birthday cake, quiet gaze.
For more cinematic feel, you can add: cinematic still, soft window light, shallow depth of field, nostalgic film color grading.
The most moving thing about AI sometimes isn’t creating the future, but letting us see the past again.
10. 焦外抽离双生：手机照→85mm F1.2大光圈
提示词
Based on the photo I uploaded, create a standalone 3:4 vertical “Bokeh Separation Diptych” poster.
Divide the image strictly into two equal-height sections, top and bottom, each occupying 50% of the canvas.
Both sections must preserve the same subject, face, action, pose, gaze direction, subject position, scene, camera angle, and compositional relationship. Do not redesign the person, replace the background, or change the photo’s content.
Upper section | REAL
Preserve the uploaded original photo in full, retaining its original phone, digital-camera, or candid-photo quality. Keep the person, background, and environmental details. Minor exposure and color adjustments are allowed, but do not change the depth-of-field relationship. Make the original state of the photo clearly visible.
Lower section | 85MM F/1.2
Strictly based on the same image above, simulate a professional full-frame portrait shot with a wide aperture: an 85mm full-frame prime lens at f/1.2.
Keep the subject’s eyes, face, and main contours crisp and sharp. Recalculate depth of field according to realistic spatial distances: the subject’s focal plane is sharp; the environment in front of and behind the subject progressively falls out of focus. Retain a little structure in the near background, soften the mid-distance background, and turn the distant background into natural creamy bokeh.
Streetlights, car lights, shop windows, reflections on leaves and metal, and sunlight highlights should become realistic optical bokeh of varied sizes and blur levels. The bokeh must show real lens characteristics: soft circular highlights, slight cat-eye distortion at the edges, smooth highlight gradients, layered foreground and background depth, and realistic spatial compression.
Avoid cutout edges around the subject’s hair, shoulders, and clothing. Blend hair naturally into the defocused background. Preserve subtle lens breathing, chromatic aberration, grain, exposure imperfections, and other authentic photographic flaws.
The final result should look as if the same photographer stayed in the same position and simply replaced an ordinary phone with a professional 85mm F1.2 lens.
Prohibit a uniformly Gaussian-blurred background, a flat background with no depth, cutting the person out and pasting them onto a blurred background, face swapping, changing the person’s pose, adding nonexistent buildings or scenery, illustration styles, excessive retouching, plastic skin, or an artificial studio-photo look.
适用模型/参数：GPT-Image-2
11. 极端广角畸变人像
提示词
Based on the uploaded reference photo of the person, generate an “Extreme Wide-Angle Distortion Portrait.”
Identity lock: Strictly preserve the reference subject’s face shape, facial features, hairstyle, skin tone, age characteristics, and recognizable identity. Do not replace the person or create a similar but different character.
Subject settings: Clothing: oversized shirt / oversized sweatshirt / sculptural suit / large jacket / workwear. Personality: fashionable / playful / cool / relaxed / lively. Expression: calm gaze / exaggerated surprise / confident smile / humorous expression.
Lens and pose: Create a hyper-realistic full-body portrait using an 8–14mm ultra-wide or fisheye lens. Place the camera very low, about 20–40 cm from the subject. Extend one hand toward the lens, with the palm in the foreground visibly enlarged. Bring the head and upper body close to the lens to create exaggerated large-head, large-hand, broad-shoulder proportions. Dress the subject in loose, oversized clothing whose silhouette is further enlarged by wide-angle perspective. Place the legs farther from the lens so they appear slender, smaller, and strongly foreshortened.
All proportion changes must come from a realistic camera angle and optical perspective. Do not use post-production stretching or create melted areas or broken anatomy.
Scene settings: Background: country road / city street / seaside / travel landmark / office / neon-lit night scene. Lighting: golden sunlight / soft overcast light / street lights at night / side light at sunset.
Image quality: Realistic skin texture, realistic fabric folds, natural contact shadows, strong foreground-to-background depth, slight lens-edge distortion, and the imperfections of real photography. Hyper-realistic portrait photography, cinematic color grading, high-resolution original-photo quality, sharp but not over-retouched.
适用模型/参数：GPT-Image-2
12. 角色设定板：锁定角色一致性
提示词
Generate a high-precision “character design sheet” to lock the appearance of one character.
Based on the uploaded reference image or the description below, create a complete character design sheet. Every module must depict the same character. Do not allow any drift in face shape, hairstyle, clothing, or proportions.
Basic settings
Style: [realistic 3D / anime / semi-realistic / IP design]
Character description: [enter a detailed description or indicate that a reference image has been uploaded]
Gender / age / body type: [fill in]
Style keywords: [premium / futuristic / emotional / sci-fi, etc.]
Background: pure white or minimalist light color.
Layout: clean, professional, technical character sheet.
Requirements: no watermark, no logo, clear labels.
Required modules
1. Top information panel: name, identity, age, 3–5 personality keywords, and a description of the core theme.
2. Color palette: 6–8 main color swatches.
3. Main identity views (largest area): front, three-quarter, side, back, standard standing pose, and height-proportion guide.
4. Silhouette views: front and side.
5. Expression sheet (8): calm, curious, nervous, surprised, afraid, sad, determined, relaxed.
6. Micro-expressions (5).
7. Head structure: three-quarter, side, looking up, and looking down.
8. Pose variations: relaxed, tense, confident.
9. Emotional close-up: an expressive chest-up portrait.
10. Clothing details: hairstyle, materials, accessories, and shoes.
11. Hand poses: relaxed, tense, pointing, gripping, and touching the face.
Consistency requirements: Every module must show the same character with identical face shape, hairstyle, body proportions, and clothing. The main identity views must be the largest and clearest area and serve as the definitive reference.
Output a complete, professional, clearly organized character design sheet.
适用模型/参数：GPT-Image-2
13. 个人角色档案卡
提示词
Based on 3–5 photos of the person I uploaded, create a polished “character profile card / character design page.”
Requirements:
1. Preserve the person’s real appearance and make them recognizable.
2. Make the page look like a refined, hand-drawn character profile card, not an ordinary collage.
3. Include: one main half-body or bust portrait; one full-body character illustration; 3–5 small expression portraits showing different emotions; 1–2 small scene illustrations related to the person’s everyday life; a set of personality tags; one short, handwritten-style personal quote; and a concise description of the character’s traits.
Style direction: [hand-drawn illustration / Japanese character-design page / watercolor / flat vector / retro comic / fresh literary style]
Personality keywords: [e.g., quiet, loves exhibitions, coffee enthusiast, laid-back, weekend hiker, lightly retro]
Layout: make the whole page feel like a character guide that can be saved and shared. Include complete information without crowding; keep the hierarchy clear and the design polished for social media.
To make the person look more like you, add: “First study the person’s appearance, personality, and lifestyle from the photos, then unify them into one complete character-profile page.”
适用模型/参数：GPT-Image-2
14. 四格日式温情漫画（角色一致＋中文对话）
提示词
Generate a four-panel Japanese-style heartwarming comic in clean, hand-drawn anime line art, with a warm monochrome palette of beige and light brown, a soft cream background, simple gentle shadows, and an overall comforting, healing mood. Square 1:1 image, standard 2×2 layout, with thin brown borders between panels. At the top, write the large title “夫婦の日” in handwritten-style dark brown lettering, with one small heart on each side.
Panel contents
Panel 1: A young Asian couple standing together. The man stands behind the woman with both hands gently resting on her shoulders. They smile toward the camera.
Panel 2: The couple sit together in a cozy room, holding mugs and smiling at each other. The woman’s speech bubble must clearly read in Chinese: “谢谢你一直陪着我。” Use a white rounded speech bubble with crisp black text.
Panel 3: The couple cook together in a kitchen, standing side by side at the stove. The man stirs while the woman helps; both look focused and happy.
Panel 4: The couple embrace closely. The man gently kisses the woman on the forehead as she smiles with her eyes closed.
Mandatory character consistency: Keep the exact same face, hairstyle, clothing, and body proportions for both characters from panel 1 through panel 4. No drift. The man has short hair and a gentle appearance, wearing a simple sweater or shirt. The woman has medium-length hair and a gentle appearance, wearing comfortable home clothes.
Style and texture: Clean hand-drawn anime line art, warm beige, light brown, and cream monochrome palette, simple shadows, like a Japanese slice-of-life comic or healing illustration.
Text requirements: Chinese text inside the speech bubble must be correct, clear, and undistorted. Avoid photorealism, 3D rendering, vivid colors, complex backgrounds, distorted faces, garbled text, and cheap sticker aesthetics.
适用模型/参数：GPT-Image-2
海报与配图设计（11条）
15. 复古弥散渐变概念海报
提示词
Create a premium retro diffused-gradient concept poster around [theme]. Reference independent-magazine covers, art-exhibition posters, experimental-music posters, and avant-garde brand visuals. Include softly diffused color fields, hazy edges, film grain, printing noise, a scanned texture, paper grain, abstract geometric elements, and sophisticated editorial typography. The main title must be clearly legible, with an orderly layout, generous negative space, and a strong design sensibility. No human faces, e-commerce look, cheap neon, or PowerPoint-template feel.
16. 材质化解释图（文章/PPT配图）
提示词
Create a “materialized explainer graphic.”
Topic: [enter topic]
Core elements: [enter the concept / structure / process to explain]
Image style: white studio lighting; objects arranged on a white tabletop, like photography of a real physical model. Keep it clean, restrained, and premium. Give the core elements a light 3D material quality (matte plastic, translucent acrylic, metal, glass, paper cards, etc.). Use only one vivid accent color (IKB blue, orange, mint green, macaron pink, bright yellow, etc.); keep everything else white, light gray, black, or low-saturation neutrals.
Embed short Chinese labels (2–5 Chinese characters) in the image, placed in white space, on annotation panels, or beside fine pointer lines.
The result should resemble a magazine-quality educational explainer, a central visual for a high-end presentation, a product-instruction model, an educational science diagram, or a concept visualization for a work report.
Requirements: clear information, simple structure, and explicit relationships between modules. No complex backgrounds, piles of colors, cheap tech aesthetics, excessive decoration, or distorted labels.
17. 产品“破屏”强透视商业海报
提示词
Create a commercial poster with powerful spatial perspective.
Camera: Use a wide-angle lens and extreme perspective from a [low angle / eye level / overhead / front-side] view. Bring the product very close to the lens for a dramatic “breaking out of the frame” effect.
Product: [enter product name] is the undisputed visual focus in the foreground. It should appear much larger than the people and setting behind it, while retaining its complete structure and crisp edges; do not crop it excessively.
Person and action: [enter person’s features]. Show the person [handing out / throwing / sprinting / leaping / swinging / reaching for] the product in a dynamic pose with clear direction and energy. Keep the person in the middle or background.
Product details: Emphasize material qualities such as [brushed metal / transparent glass / leather texture / condensation droplets / velvet matte finish / ceramic sheen / frosted plastic].
Background: Use a simple gradient from [main color] to [secondary color]. Avoid a busy environment so the product silhouette and action stand out.
Overall style: premium commercial photography, magazine-quality advertising, minimalist composition, strong perspective and movement, with a high-value, frame-breaking effect.
Avoid an ordinary eye-level product shot, placing the product and person at the same depth of field, a complex background, cheap neon, distorted products, malformed hands, or excessive edge cropping.
适用模型/参数：GPT-Image-2
18. 笔触成体：丙烯厚笔触艺术海报
提示词
Create one complete 3:4 vertical art poster. Use a small number of broad, thick acrylic-paint strokes to form [subject] in space.
Each stroke directly carries the outline, volume, and movement; use large color blocks to summarize the form.
Palette: [palette]. On close inspection, show parallel brush grooves, accumulated paint at stroke starts, varied thickness, overlapping layers, strokes that thin and naturally split at their ends, and occasional dry-brush edges. Keep the subject’s front-to-back structure clear. Side lighting should reveal paint thickness, occlusion, and soft shadows. Make the details tactile but not greasy.
Text is limited to the large title “[title]” and the small words “笔触成体” and “AI绘画每日一词,” set in a carefully designed [modern Chinese sans-serif] with strong size contrast.
Composition and text placement: [composition]. Use a clean solid-color background, generous quiet space, few but forceful visual elements, and publication-level art direction. No smooth plastic ribbons, paper strips, metal, glass, yarn, real objects merely coated in paint, flat oil-painting appearance, decorative borders, icons, unrelated English, or other text.
Example values (variation 1):
[Subject] An abstract flower with five broad petals. The center and petals are formed from separate but overlapping, connected thick-paint sweeps; no flowerpot.
[Palette] Coral red and vermilion petals, a small stroke of bright yellow at the center, deep ink-green stem and leaves, warm ivory background, and dark charcoal text.
[Title] 花
[Composition] The flower opens diagonally from the lower right toward the upper left, occupying about two-thirds of the image width. Place a huge “花” in the upper left to balance the flower diagonally. Keep the upper right and lower left open, with the small text aligned along the lower-left baseline.
19. 一张产品图批量出8张海报方案
提示词
Based on this product image, generate 8 product-poster concepts in different styles.
Product name: [enter product name]
Use case: [enter product purpose]
Number of outputs: 8
Aspect ratio: 3:4
Platform: [Xiaohongshu / Taobao / WeChat Official Account / Douyin image post]
Requirements:
1. Accurately preserve the product’s appearance, packaging, colors, and key identifying features.
2. Give every image a different setting, composition, atmosphere, and visual style.
3. Each must feel like a finished commercial poster, not merely the same image with a changed background.
4. Make each suitable for product discovery and sharing, with a clear visual focus and atmosphere.
Possible styles: minimalist white, camping, Japanese-inspired, moody premium, fresh summer, literary retro, breakfast setting, or lifestyle.
Vary the product angle, surrounding scene, lighting atmosphere, and compositional rhythm. The overall work should be polished and ready for an e-commerce proposal or social-media presentation.
适用模型/参数：GPT-Image-2
20. 极简公共空间杂志海报
提示词
Aspect ratio: [5:2 / 16:9 / 3:4]
Main title: [enter title]
Subtitle: [optional]
Scene: [enter a specific setting]
Core metaphor: [enter an action or object]
Create a sophisticated editorial poster.
Overall style: calm, restrained, orderly, spacious, and geometrically composed.
Possible settings: pool, corridor, stairs, colonnade, bench, doorway, tiled space, or public architecture. Keep the space clean and generously open.
Use very few people. Their expressions should be calm; no exaggeration or acting. People serve only as compositional elements and metaphors.
Palette: mint green, ice blue, cream, and light wood, with a small tomato-red or coral accent as a visual anchor.
Typography: retain only the main title and subtitle. Make the main title large and integrate it into the composition.
Prohibit information panels, numbering, data icons, barcodes, decorative English, stacks of magazine microcopy, PowerPoint styling, excessive tech aesthetics, or too many elements.
Final result: minimalist art photography, a premium magazine-cover feel, and integrated image and typography.
适用模型/参数：GPT-Image-2
21. 极简杂志海报
提示词
Design a 3:4 vertical “minimalist magazine poster” for [a person / product / viewpoint / exhibition theme].
Use an independent design magazine-cover layout. Keep it restrained, quiet, avant-garde, and editorial. Use a warm white, gray-white, or low-saturation solid-color background with ample clean negative space.
Include only one core subject: [person / product / building / object / abstract shape]. It may be shown in full, boldly cropped, partly outside the frame, or asymmetrically placed, but the visual focus must remain clear.
Set the title in oversized bold sans-serif type as the main visual anchor. Keep it to 2–6 characters or words. Some text may be partly covered by the subject, extend beyond the image edge, or use an asymmetric layout, but it must remain legible.
Use small neutral sans-serif type for minimal supporting information, such as an issue number, date, section name, or short note.
Limit the hierarchy to: 1) oversized main title; 2) one core subject; 3) a small amount of supporting information.
Use a strict Swiss grid and asymmetric editorial layout. Limit the palette to black, white, gray, and one low-saturation accent such as dark red, olive green, gray-blue, or earthy brown. Add subtle paper fibers, ink grain, and slight print misregistration, but do not over-distress the image.
Make it feel like a limited-edition cover by an independent design studio, not an ordinary promotional poster.
Avoid crowded elements, complex backgrounds, multiple people, decorative type, glowing gradients, metallic effects, cheap tech aesthetics, dense copy, sticker-like decoration, or meaningless graphics.
22. 干净旧报纸风封面（文章配图）
提示词
Main title: [enter title]
Short words at the bottom: [enter 1–3 short phrases, or leave blank]
Aspect ratio: [5:2 / 16:9 / 3:2 / 4:3 / 1:1 / 4:5 / 3:4 / 9:16]
Create a cover poster in a “clean vintage newspaper + retro ink type + minimalist halftone metaphor” style.
Overall visual: Use a uniform, clean, solid light warm gray-beige background close to #DED9CF. No obvious paper grain, stains, folds, gradients, lighting, or large areas of distressing.
Use only three colors throughout: light warm gray-beige for the background; deep black for the title and illustration; dark gray for the words at the bottom. Do not add any other colors.
Typography: Use one consistent retro mechanical inked-type style: slightly narrow, square, and steady, with subtle ink spread and uneven vintage printing, like an old typewriter, letterpress newspaper, or archival heading. Set the main title in bold, clearly and completely, in no more than two lines. Keep its size moderate; do not enlarge it automatically just because it is short, and do not let it fill the image. Set the bottom words in a thin version of the same typeface, up to three groups separated by middle dots. If no bottom words are provided, leave that area blank; do not add text.
Layout: For landscape formats, use text on the left and image on the right, with the title and bottom words on the left and a small halftone illustration on the right, separated by generous space. For square and portrait formats, use text above and image below: title at the top, halftone illustration in the middle-lower area, and bottom words in the lowest safe area. Keep the illustration small and subordinate.
Visual metaphor: Understand the central meaning of the theme and distill one simple, direct visual metaphor. Include only one title, one core metaphor, one main visual subject, and one group of bottom words. The metaphor may be an ordinary object, a simple action, a relationship between two objects, a gap, break, balance, occlusion, or change of direction. Do not illustrate every concept in the theme.
No photorealism, 3D, high-saturation colors, gradients, glow, complex backgrounds, or large dirty distressed areas.
23. 上下双生海报：照片＋几何平涂插画
提示词
Turn the photo I uploaded into one standalone, sophisticated design poster. Do not combine multiple images; output each photo separately.
Use a 3:4 vertical composition, strictly divided into upper and lower sections, each occupying 50% of the height.
Upper section: Preserve the original photo, including the subject’s identity, structure, pose, realistic texture, natural lighting, and original color atmosphere. Apply only subtle premium color grading for an art-magazine, independent-publication, or exhibition-image feel. The environment may be naturally extended to fit the frame, but do not stretch, distort, or alter the subject.
Lower section: Based on the same photo, extract its most recognizable subject, silhouette, structure, pose, and narrative relationship, then reconstruct it as a modernist geometric flat illustration.
Do not merely outline the subject with geometric lines. Reduce it directly to rectangles, trapezoids, triangles, circles, and a few freeform blocks. Preserve the key identifying features with as few shapes as possible, so the original subject is immediately recognizable.
Use clearly flat spatial expression: reduce realistic perspective, volume, and lighting; build space mainly through overlap, scale, layered color blocks, and vertical relationships. Combine complex details of people, buildings, vehicles, plants, objects, or environments into a few large shape systems. Summarize repeated elements as unified color blocks rather than drawing each one. If the source is not an urban scene, do not force it into a city; use the same geometric approach for people, animals, still life, landscapes, or other subjects.
Emphasize large-shape proportions, visual balance, layering, positive and negative shapes, color rhythm, and ordered negative space. The subject may be off-center, cropped, overlapped, or naturally spread out, but the narrative focus must remain clear.
Use large flat color areas, slightly irregular hand-drawn edges, and a matte gouache or acrylic texture. The paint should feel soft and delicate, with a subtle handmade, layered quality. Avoid digital gradients, realistic material simulation, complex shadows, photo filters, or slick vector art.
Final result: the upper half feels like a premium photographic poster; the lower half feels like a modernist geometric collage illustration. Both depict the same subject and form a clear top-and-bottom relationship.
24. 超大中文字＋物体穿插海报
提示词
Create a Chinese poster with a strong graphic-design sensibility.
Place oversized Chinese sans-serif characters “[enter keyword]” in the center, occupying most of the visual area. Use bold, clear, accurate strokes with a strong structural presence.
Let “[enter object]” pass naturally through the gaps between the strokes, creating clear foreground-background occlusion and spatial interweaving.
Use “[enter solid background color]” for the background, with halftone grain, subtle print texture, and a street-poster feel.
Style: high contrast, solid color blocks, integrated image and typography, accurate Chinese lettering, with a blend of avant-garde magazine, exhibition-poster, and street-art aesthetics.
Do not make an ordinary illustrated poster or treat the text as decoration. Do not simply place the object in front of the text; make the object genuinely enter the text’s space.
Example substitutions: keyword 坐 / object wooden chair / vermilion background; keyword 飞 / paper airplane / deep-blue background; keyword 光 / prism or light beam / black background; keyword 书 / open book pages / off-white background; keyword 山 / mountain silhouette / gray-green background.
适用模型/参数：GPT-Image-2
25. 生活化UGC种草图：手机随手拍质感的产品照
提示词
Based on the uploaded product reference image, generate 6 separate lifestyle-style UGC product photos.
Product name: [enter product]
Target audience: [enter primary users]
Platform: [Xiaohongshu / Douyin image post / e-commerce detail page / social media]
Aspect ratio: 4:5
Output: 6 independent images, not a collage or multi-panel layout.
Scene setup
Scene 1: [enter everyday setting and action]
Scene 2: [enter everyday setting and action]
Scene 3: [enter everyday setting and action]
Scene 4: [enter everyday setting and action]
Scene 5: [enter everyday setting and action]
Scene 6: [enter everyday setting and action]
Give every image a different environment, camera position, lighting, product placement, and human action, while keeping the product packaging, colors, logo, text placement, and physical design consistent.
Photography style: Authentic native smartphone photography, not a commercial studio advertisement. Allow slight highlight clipping, slightly gray shadows, mild desaturation, natural white-balance variation, traces of phone auto-exposure, slight motion blur, realistic depth of field, minor compression, and an imperfect handheld composition.
People: Show people in natural everyday moments, not looking at the camera or only glancing at it briefly. Preserve real skin texture, pores, fine lines, under-eye circles, flyaway hair, and natural expressions. No excessive smoothing, influencer-style faces, or stiff studio-model posing.
Lighting: Use naturally occurring mixed light, such as window daylight with ceiling lights, a desk lamp with screen light, midday hard light with architectural shade, warm evening light with cool indoor light, or tree shadows with outdoor bounce. Allow slight color-temperature differences between light sources; do not make the lighting excessively uniform.
Product integration: Place the product naturally in a real-use situation: held in a hand, set near the edge of a table, opened for use, next to everyday items, just removed from its packaging, or used for a while. Do not always center it or face it directly toward the camera. Add realistic traces appropriate to the product, such as condensation on a cup, fingerprints on a bottle, slight packaging creases, natural fabric folds, screen reflections, subtle signs of use around a bottle opening, or minor everyday clutter on a tabletop. Every trace must make sense; do not make the product dirty or damage its brand identity.
Background: Include real-life context without letting clutter overpower the subject. Suitable objects include keyboards, books, cosmetics, clothing, delivery boxes, sticky notes, tableware, and children’s toys.
Prohibit commercial studio lighting, perfect symmetry, overly retouched skin, stiff model poses, fake customer-review text, fabricated product effects or results, changes to the product packaging or logo, identical compositions across all six photos, cheap filter aesthetics, or obvious AI plastic skin.
The final images should look like the same ordinary user naturally recorded moments of using the product on a phone in different everyday settings. When shared publicly, clearly label AI-generated, AI-assisted, or synthetic content according to platform rules; do not present it as a real customer’s photo.
品牌与标志（2条）
26. 品牌视觉套件：4张连续品牌提案图
提示词
Using the brand information below, create a polished set of “4 brand-visual proposal images” with one cohesive visual system.
Important: These are not a four-panel collage or one combined image. Generate four separate images in sequence. All four must belong to the same brand system, but each has a different role and must not be a mere variation of the others.
Brand information
Brand name: 天马奶茶
English supporting name: TIANMA TEA
Brand type: youthful Eastern tea drinks / contemporary light milk tea.
Positioning: Make Eastern tea drinks lighter, fresher, and more suitable for young people’s everyday check-ins and sharing.
Slogan: 这一杯，刚好是喜欢的样子。
Core keywords: contemporary Chinese style, light, refreshing, youthful, Eastern aesthetics, tea aroma, milk fragrance, fruit aroma, everyday comfort, photo-friendly, translucent, tactile tableware.
Target audience: people aged 18–35 who enjoy milk tea, taking photos, polished delivery experiences, light socializing, visual design, and brand experiences.
Overall direction: youthful, fresh, translucent, restrained, lightly Eastern, spacious, tactile, with layered products. This is neither an old-fashioned traditional tea shop nor a cheap viral milk-tea store; make it a younger contemporary Chinese tea brand.
Main colors: off-white, tea-leaf green, creamy white, misty blue-green, oolong gray, amber tea brown.
Supporting colors: pale apricot, light wood brown, cool gray, soft bamboo green.
Accent colors: small touches of peach pink, lemon yellow, and grapefruit orange, used only to brighten selected products.
Supporting graphics: water ripples, tea-leaf outlines, seals, small tags, Eastern line motifs, cup silhouettes, arcs of tea, and icons of pearls / coconut jelly / lemon slices / tea leaves.
Tableware: clear takeaway cups, frosted cups, cold-brew bottles, white porcelain cups, glass cups, wooden trays, paper bags, seals, cup sleeves, and coasters.
Core products: 1. Signature Tianma milk tea with green tea; 2. light oolong milk tea; 3. jasmine light milk tea; 4. lemon iced tea; 5. peach oolong; 6. brown-sugar pearl fresh milk.
Unified requirements
1. All four images must belong to one brand system. Keep the brand name, palette, typographic character, supporting graphics, packaging logic, and product logic consistent.
2. Use a unified 4:5 vertical format, suitable for social media and brand proposals.
3. Generate each image separately. No four-panel layout or combined image.
4. Make every image polished, informative, well designed, and proposal-ready.
5. Avoid overly historical Chinese styling, old-established-brand aesthetics, or a cheap dessert-shop look. Emphasize contemporary Chinese style, youthfulness, light tea, and freshness.
6. Chinese headlines must be clear; English may appear only as supporting text. Give the layout a strong design sensibility.
7. Products must look appetizing and believable, with clear tea layers, refined cups, and consistent packaging.
8. Generate in sequence: image 1 → image 2 → image 3 → image 4.
IMAGE 1 | BRAND HERO BOARD / Brand cover
Task: Create the cover image for the proposal set. The first impression should be youthful, contemporary Chinese, refreshing, tea-forward, and clearly branded.
Layout: At the top, show the brand name “天马奶茶,” supporting name “TIANMA TEA,” and slogan “这一杯，刚好是喜欢的样子。” In the center, feature a cup of “招牌天马奶绿” as the hero product, in a clear or frosted cup, showing the green-tea milk base, milk cap or light-milk layer, condensation on the cup, and translucent quality. Around it, arrange a restrained selection of tea leaves, milky elements, lemon slices, oolong leaves, a wooden tray, small seals, and Eastern-inspired light and shadow. At the bottom, show small cards for four featured products: 招牌天马奶绿, 轻乌龙奶茶, 茉莉轻乳茶, 桃桃乌龙. Add selling-point icons for refreshing tea, quality ingredients, light Eastern drinks, and photo-friendly presentation.
Make this feel like a complete brand cover, not an ordinary milk-tea poster. Convey proposal-level design, appetite appeal, and youthful Eastern character.
IMAGE 2 | VISUAL SYSTEM BOARD / Brand identity archive
Task: Show how the brand identity is constructed. Make it more systematic than image 1 while retaining the fresh, youthful, contemporary Chinese character.
Layout: At the top, show the brand name, English name, and slogan, with a representative milk-tea image on the right. In modular sections, show: the brand marks (Chinese “天马奶茶,” English “TIANMA TEA,” a simplified seal, or supporting graphic); palette (off-white, tea-leaf green, creamy white, misty blue-green, oolong gray, amber tea brown, pale apricot); type system (Chinese headline type, body type, supporting English type); icons (tea leaf, pearls, coconut jelly, milk cap, lemon slice, peach, cup); labels (Eastern light drinks, refreshing milk tea, quality ingredients, light tea, youthful check-in, daily cup); and supporting graphics (water ripples, tea arcs, small tags, cup silhouettes, Eastern line structures). At the bottom, show six product cards: 招牌天马奶绿, 轻乌龙奶茶, 茉莉轻乳茶, 柠檬冰茶, 桃桃乌龙, 黑糖珍珠鲜奶. Each card has a small image, icon, and short description.
Do not make this an ordinary VI manual. Make it a “milk-tea brand visual-system board”: information-rich but orderly, suitable for a proposal.
IMAGE 3 | PRODUCT & PACKAGING BOARD / Product and packaging applications
Task: Show how the brand becomes a sellable tea-drink system.
Layout: At the top, display four key products—招牌天马奶绿, 轻乌龙奶茶, 桃桃乌龙, 黑糖珍珠鲜奶—in clear cups, frosted cups, cold-brew bottles, or other refined vessels. Make the tea layers clear and appetizing. In the middle, show packaging extensions: takeaway cups, sleeves, seals, paper bags, shopping bags, menu cards, order stickers, delivery labels, and two-cup sharing packs, all following the same brand system. At the bottom, show six complete product cards: 招牌天马奶绿, 轻乌龙奶茶, 茉莉轻乳茶, 柠檬冰茶, 桃桃乌龙, 黑糖珍珠鲜奶, with selling points such as refreshing tea, quality ingredients, lightness, daily recommendations, and check-in friendliness.
It must feel genuinely sellable. Do not merely pile up products; show packaging, system, brand identity, and commercial readiness.
IMAGE 4 | CAMPAIGN & SPACE BOARD / Campaign and in-scene applications
Task: Show the brand’s communication and spatial experience in real settings.
Layout: Large scene on the left: a youthful contemporary Chinese milk-tea shop with the storefront sign “天马奶茶” and supporting English “TIANMA TEA.” Keep it clean, fresh, and lightly Eastern—not old-fashioned. Upper right: a mall pop-up or small tea stand with brand posters, product display, a small menu, and packaging. Middle right: two campaign posters, one featuring 招牌天马奶绿 and the other 桃桃乌龙 or 黑糖珍珠鲜奶. Poster copy may include “这一杯，刚好是喜欢的样子。” and “喝一口，刚好轻松下来。” Lower left: simulate 3–4 vertical social-media images, such as a new-product recommendation, store check-in, today’s recommendation, and ingredient showcase. Lower right: show brand touchpoints—paper bags, napkins, coasters, seals, small cards, and delivery stickers. At the bottom, add application tags: contemporary Chinese tea, youthful check-ins, light Eastern drinks, everyday comfort, quality ingredients. Optionally add the emotional line “今天，也值得来一杯。”
Make this more narrative and situational than image 3. It should be neither a cheap milk-tea store nor a cold, aloof café, but a youthful, fresh, shareable contemporary Chinese tea brand.
Execution sequence: First understand and unify the brand’s visual DNA. Then generate image 1, the brand cover. While maintaining the same brand system, generate image 2, the visual identity archive; image 3, product and packaging applications; and image 4, campaign and real-world applications. Do not combine the four images, make a four-panel collage, or repeat the same content. These are four consecutive proposal images for one brand, not four different projects.
Final goal: Output four separate yet highly consistent “天马奶茶” brand-visual proposal images. Image 1 establishes the direction; image 2 explains the system; image 3 shows product applications; image 4 shows communication and real-world applications. The set should have unified brand DNA, clearly differentiated page roles, polished aesthetics, commercial proposal quality, presentation value, portfolio quality, and be ready for concept presentation.
27. 物字共生：器物结构融进文字的标志
提示词
Design a “form-and-letter integrated” logo for the exact text [2–4 Chinese characters / a short English word or phrase], combining it with an object [item / tool / instrument / food / animal].
First analyze the object’s most distinctive structures: its silhouette, handle, opening, joint, button, base, tip, texture, or motion path. Map those structures into the letterforms: the object’s main body becomes a primary stroke; handles or openings become negative space in the glyph; joints become turning points; functional parts become dots, ticks, or decorative details.
Keep the text content, order, and character count completely accurate. Do not simply place the object next to the text or let an illustration obscure the text. The final mark must satisfy all three: (1) the text remains legible when reduced; (2) the object remains recognizable without color; and (3) the text and object form one inseparable whole.
Create one complete logo, centered on a solid-color background, in black-and-white vector style, with clean edges, consistent stroke weights, and necessary negative space. No poster scene, 3D model, mockup, shadow, gradient, complex texture, or extra text.
风格化重绘（3条）
28. 儿童蜡笔画风重绘
提示词
Redraw the uploaded image as a childlike crayon illustration.
Give it the hand-drawn quality of a 10-year-old child: naive, imperfect, slightly wobbly lines; simplified shapes and slightly exaggerated proportions; bright, lively crayon colors; color areas may spill outside the outlines; a clean white-paper or sketchbook-paper background; and playful additions such as stars, clouds, candy, or a little sun.
It must look like a real child’s drawing, not an illustration-software effect.
Avoid watercolor texture, realistic lighting, delicate gradients, and photographic style.
29. 照片续写：真实照片＋手绘手稿
提示词
Using Image 1 as the sole content reference, create one standalone “photo continuation” visual in a 3:4 aspect ratio.
Preserve approximately the top 50% of Image 1 as a real photograph. Strictly maintain the subject’s face, hairstyle, clothing, pose, structure, scene perspective, and lighting direction.
Starting from a natural boundary around the middle of the image—such as a horizon, table edge, hem, road, or building base—gradually transition the real image into warm-white handmade paper.
The photo’s contours, road, body, or object edges must cross the boundary accurately and continue across the paper as [colored-pencil / crayon / watercolor / pen] sketches.
Continue the hand-drawn area into [a scene that has not happened yet / a person’s wish / a making process / spatial structure / a travel route]. Keep paper fibers, pencil test lines, layered crayon, slight erasures, tape, and irregular handmade edges clearly visible.
Derive 3–5 hand-drawn colors from the real photo. The photograph and drawing must share the same subject, perspective, and spatial direction; do not collage two unrelated images.
Prohibit a hard straight split, before-and-after comparison, grid, framed collage, duplicate people, face changes, broken perspective between top and bottom, excessive text, or unrelated decoration.
30. 刺绣自生长：针线正在绣出图案
提示词
Create a single 3:4 vertical “embroidery growing into existence” visual for [a city landmark / Chinese character / brand graphic / character / food].
Stretch warm ivory natural linen in a round wooden embroidery hoop and photograph it from a three-quarter overhead macro view. The upper part of the image shows [the completed motif]; the middle is the active stitching boundary; the lower part retains a broad area of untouched blank fabric.
A real metal embroidery needle is emerging from the fabric. Colored silk thread passes through a clearly visible needle hole and tightening loop, then continues along a connected path. Each needle pass gradually forms outlines, color blocks, and texture; areas the needle has not reached must not show a completed motif.
Choose stitches according to the structure: long-and-short stitch, satin stitch, chain stitch, and French knots. Use chain stitch for outlines, satin stitch with consistent direction for the main color fields, and French knots for tiny highlights. Every thread must connect to a real stitch point and have believable tension, thickness, occlusion, and contact shadows.
Keep the motif essentially flat like embroidery; only allow slight relief from local knots. Do not turn buildings, people, or flowers into 3D models rising from the fabric.
适用模型/参数：可配合视频：固定机位，让针线从空白布面完成整个图案
