GPT Image 图片提示词与视频模板

Turn my uploaded image into a standalone sticker suitable for an animated reaction. Automatically identify the main subject: For a person, preserve recognizable facial features, hairstyle, clothing, and expression. For an animal, preserve its breed characteristics, fur color, markings, and expression. For an object, preserve its distinctive silhouette, structure, and main colors. If the image mainly depicts a scene, turn the entire scene into a sticker with a simple outline. Use a cartoon sticker style with subtle depth, clear contours, rounded forms, and a touch of humor. Do not give different subjects the same generic face. Do not add limbs to objects that do not have them. Add a clean white die-cut border around the sticker. Center the subject with ample space on every side, without cropping it. Generate only one sticker, not a grid. Add no text, and do not show the sticker applied to another object. Use a square 1:1 canvas. Output a PNG with a true alpha-transparent background; do not draw a checkerboard into the image.

Create a Meme sticker pack based on the images attached, remixing with 😎😛💕🚀🥳. Use exaggerated internet-reaction expressions, including crying, confusion, shock, smugness, side-eye, and deadpan disbelief, with awkward poses, low-fi cutout textures, and absurd humor. Create a single square(1:1) transparent sticker sheet with nine distinct stickers arranged in a 3×3 grid, each showing a different expression, pose, or reaction. Separate the stickers with wide, fully transparent gaps. No background, shadows, or overlapping elements.

Turn the main person, animal, or cartoon character in my uploaded image into a personalized chat sticker pack. First identify the subject's most recognizable features, then translate them consistently into a simple, expressive, slightly exaggerated hand-drawn sticker style. Preserve facial proportions, hairstyle or fur, distinctive markings, clothing, and main colors. Do not turn the subject into a generic character. Generate one neatly arranged 3×3 grid. All nine cells must depict the same character with a consistent appearance and art style. From left to right, top to bottom, show: Laughing happily; utterly confused; completely speechless. Suddenly shocked; upset and about to cry; secretly smug. Extremely sleepy; puffed up with anger; expressing affection. Each cell must have a clearly different expression and pose that remains readable at a small size. Movements must suit the subject's anatomy; do not attach a realistic human body to an animal. Leave empty space around each sticker. Stickers must not touch. Do not draw dividing lines or add text. Use a pure white background and a square 1:1 canvas.

〈主体〉在〈场景与环境〉中〈主要动作或事件〉。 画面呈现〈视觉风格〉。 镜头采用〈景别、机位、运镜或切镜〉。 声音包括〈对白、环境声、音效或音乐〉。

Use the image generation tool and my single uploaded image as the character reference to generate a six-frame action sequence sheet for a looping GIF. This step must generate new action drawings. Do not substitute Python translations, rotations, scaling, or local stretching of the original image for drawing new poses. CHARACTER AND ACTION Identify the main person, animal, or cartoon character. Preserve its recognizable facial proportions, hairstyle or fur, markings, clothing, accessories, and main colors. If the source is already a sticker or illustration, keep its style. If it is a photograph, convert it consistently into a simple, expressive hand-drawn sticker style. Create only one reaction or action: [Enter an action here, such as laughing happily. If left blank, choose one suitable action based on the uploaded image.] Develop this one action throughout; do not draw six unrelated expressions. SIX CONSECUTIVE STAGES Read left to right, top to bottom: Frame 1: Starting pose, before the action fully develops. Frame 2: Anticipation begins; the face and relevant limbs start to change. Frame 3: The action develops and the expression clearly intensifies. Frame 4: Peak action, with the emotion most clearly expressed. Frame 5: The action subsides and the character begins to recover. Frame 6: Near the starting pose, ready to reconnect smoothly to frame 1. Design the mouth, cheeks, eyebrows, eyes, and body pose in each frame for the chosen action. At least one facial feature must show a clear, continuous change, supported by body movement: for example, a mouth gradually opening and cheeks rising during laughter, or an opening mouth and dropping jaw during surprise. Redraw the same character's facial features and limbs at different moments as needed. Preserve identity and design without freezing the eyes, mouth, and pose. If the character wears sunglasses, keep them and express emotion through the mouth, cheeks, chin, and body; do not invent visible eyes through the lenses. Animal movements must respect their anatomy. FRAME CONSISTENCY Keep the camera, viewpoint, lighting, art style, and base character size fixed. Use the same composition coordinates and body anchor position in every cell. Local contours may change with the action, but do not resize the character in each frame to fill the cell. Clothing, accessories, limb count, and fur markings must not change without reason. Existing props and decorations may move naturally with the action, but must not randomly appear or disappear. SHEET LAYOUT Use 3 columns × 2 rows, with an overall aspect ratio of 3:2 and six equal square cells. Use a regular invisible grid, with no dividing lines, outer border, text, or frame numbers. Each cell contains one complete character, including its props and decorations. Leave approximately 15% safety space around it; nothing may cross into a neighboring cell. Use a true alpha-transparent background, without white panels or a painted checkerboard. Preserve the sticker's white outline and white details within the subject. DELIVERABLE Generate the six-frame action sheet directly. Do not make the GIF yet.

Use the main product in my uploaded image as the sole product reference to generate a high-quality base image for a subsequent animated advertisement. Strictly preserve the product's silhouette, proportions, structure, colors, materials, logo, brand name, label text, packaging information, and every recognizable detail. Do not redesign the product or add text, selling points, prices, specifications, decorative elements, or accessories absent from the source. Use high-end commercial product photography. Place the product centrally, facing the camera, with a camera as steady as if mounted on a tripod. Maintain natural perspective and realistic proportions, with generous safety space around the product. Use a simple, neutral, unobtrusive studio background. Choose off-white, light gray, warm gray, or another suitable neutral based on the product's colors. Avoid complex scenes, props, or obvious textures. Use soft, restrained, even baseline studio lighting. Do not add a visible light sweep, strong strip highlights, dramatic shadows, or localized overexposure at this stage. Keep exposure balanced and render glass, metal, plastic, fabric, leather, and liquids realistically. Metal may have natural reflections; frosted materials should remain soft; transparent or translucent materials should retain realistic light transmission. Do not change the product's original material to make it look more luxurious. This image will be a completely fixed base frame for the subsequent animation. Prioritize accurate, stable product structure, logo, label, and text. The result should be clean, restrained, and refined, like a finished studio product photograph that only needs moving light added over the fixed image.

Edit image 1 by replacing its main product with the product in image 2. Strictly separate the two references' roles: Image 1 determines only the advertising composition, background, lighting direction, decorative elements, and text layout outside the product. Image 2 determines only the new product's actual shape, structure, materials, colors, logo, and packaging information. Remove the old product from image 1 and place the product from image 2 naturally in the same area. Preserve the new product's true width-to-height proportions. Do not squash, stretch, or restructure it to fit the old silhouette. You may resize the whole product proportionally to fit the layout. Change only the area occupied by the new product and the contact shadows, occlusion, and local reflections necessarily affected by the replacement. Do not remake the entire poster. Keep all headings, text content, fonts, font sizes, positions, background textures, decorative elements, and canvas dimensions outside the product unchanged. Do not blend the two product designs. Retain none of the old product's cap, handle, label, or other parts. Do not add accessories absent from image 2, guess illegible packaging text, or change the product's colors to match the background. Output only one complete poster with the product replaced.

@图片1用于〈主体〉的〈外貌、服装、结构或材质〉，不采用〈无关内容〉。 @视频1用于〈动作、运镜或节奏〉，不采用〈无关内容〉。 @音频1用于〈角色或声音类型〉的〈音色、台词、环境声或音乐〉。 〈主体〉在〈场景〉中完成〈主要动作或事件〉。 画面呈现〈视觉风格〉，镜头采用〈镜头表达〉。

Dress the person in image 1 in the top from image 2, producing a realistic virtual try-on image. Image 1 determines only the person's identity, body proportions, pose, camera angle, and background. Image 2 determines only the top's style, color, fabric, cut, patterns, and details. Preserve the person's face, hairstyle, skin tone, body shape, and original pose. Replace only the top. Do not change shoulder width, waist size, or body proportions to make the clothing look better. Accurately preserve the neckline, sleeve shape, placket, pockets, buttons, and all other visible construction details. Keep existing prints, embroidery, logos, and text in their corresponding positions on the garment. They may bend and wrinkle naturally with the fabric, but must not be rewritten, mirrored, added, removed, or moved elsewhere. Let the fabric drape plausibly over the shoulders, arms, and torso. Preserve its apparent thickness and surface texture. Do not turn knitwear into smooth plastic or structured fabric into thin, clingy cloth. Show the front view. Do not invent a back design absent from the reference. Preserve the original lighting and environment so the replacement clothing blends naturally into the photograph. Output one try-on image, without added text or a before-and-after comparison.

Place my uploaded transparent or translucent product into a new commercial photography scene. Preserve the product's silhouette, proportions, thickness characteristics, bottle neck or opening structure, label position, and readable packaging text. If it contains liquid, preserve the liquid's color and fill level. Use a dark gray studio background and a dark, low-reflectivity tabletop. Place a large, soft light source on one side and a restrained warm rim light behind the product. Pay particular attention to the product's optical relationship with the new environment: Transparent regions must reveal the new background, without retaining colors or objects from the original background. Render visually plausible transmission and refraction based on the product's shape. Surface reflections must match the new light sources. Keep a natural contact shadow beneath the product. Reflections, illumination, and the background seen through the product may change, but the product's structure, color, label, and contents must not. Do not turn transparent glass into opaque plastic, frosted material into perfectly clear glass, or add chrome effects where none existed. Preserve the original camera angle; do not show an unreferenced back view. Generate only one 3:4 portrait product photograph, with no added props or copy.

Using my two uploaded images, generate a natural, believable photograph of a real person presenting a product. Image 1 is the person reference and determines identity, facial features, hairstyle, skin tone, and clothing. Image 2 is the product reference and determines shape, structure, colors, materials, logo, and packaging information. Have the person in image 1 naturally hold the product in image 2 as if presenting it to the camera. Place the product in front of the chest or slightly forward beside the body, with its main display surface facing the camera and without blocking the face. Choose a plausible grip for the product's shape. Fingers should naturally wrap around, support, or pinch it, with correct occlusion, contact shadows, and weight support. Hands must not intersect the product, and the product must not float above the palm. Preserve the person's recognizable facial proportions, apparent age, hairstyle, and skin detail. Do not replace the face or change the body shape. The arms and hands may change pose as needed to hold the product. Do not redesign the product. Preserve its silhouette, cap or other structure, label position, main colors, and readable text. Do not mirror the logo or add accessories. Light the person and product consistently to avoid a pasted-on appearance. Keep the background from the person's photograph, simplifying distracting elements where appropriate. Use a 3:4 portrait composition with both person and product clearly visible. Add no headline, slogan, or customer testimonial.

【人物】 〈人物A〉对应@图片1，只采用外貌、发型和服装。 〈人物B〉对应@图片2，只采用外貌、发型和服装。 人物的外貌、动作、站位和台词不互相交换。 【道具】 〈道具A〉对应@图片3，只属于〈人物A〉。 【场景】 〈场景A〉参考@图片4，只采用空间、材质和光线，不采用图中人物。 【动作与声音】 @视频1用于〈动作〉，不采用视频中的人物和场景。 @音频1用于〈角色〉的音色和指定台词。 场景一｜〈场景名〉 使用：〈人物、道具、场景与动作素材〉。 事件：〈一个主要事件〉。 结束时：〈可直接观察的稳定状态〉。 场景二｜〈场景名〉 使用：〈本场景需要的素材〉。 事件：〈一个主要事件〉。 结束时：〈可直接观察的稳定状态〉。

Combine my uploaded product photographs into one group product image with a unified photographic style. Treat the main product in each reference image as a distinct object that must not be confused with another. Every product must appear exactly once in the final image: none omitted, duplicated, or merged. Accurately preserve each product's own shape, structure, colors, materials, logo, labels, and readable text. Do not mix caps, color schemes, packaging patterns, or labels between products. Arrange all products on the same studio tabletop in a simple, layered composition. Slight front-to-back overlap is allowed, but the main display surfaces and branding must remain clearly visible. Retain the angles already shown in the reference images as much as possible. Do not invent back views that lack a reference. Unify lighting direction, color temperature, sharpness, and contact shadows so the products look as though they were photographed together. Avoid a collage in which each product retains different lighting or a different background. Use a simple neutral background that complements the group without competing with it. Do not change product colors to suit the background. Add no unprovided products, gifts, gift boxes, shopping bags, prices, or promotional text. Generate one square 1:1 group product image, keeping all products completely inside the frame.

Change only: [the intended change]. Must preserve: [identity, composition, object positions, text, colors, and other essential details]. Allowed changes that follow from the edit: [necessary contact shadows, reflections, and occlusion].

Use image 1 only to establish the subject's identity and appearance. Use image 2 only as a reference for colors, brushwork, and material texture. Use image 3 only as a reference for composition and the subject's size within the frame. Do not bring the person from image 2 or the objects from image 3 into the final image.

【生成目标】 生成一段〈视频类型〉。核心主体是〈主体〉，主要事件是〈故事概要〉。 【阶段一】 开始时：〈人物、道具和场景的初始状态〉。 主要事件：〈一个主要动作或事件〉。 结束时：〈人物位置、道具归属或画面状态〉。 【阶段二】 承接上一阶段：〈需要保持的状态〉。 主要事件：〈一个主要动作或事件〉。 结束时：〈可观察状态〉。 【阶段三】 主要事件：〈收束事件〉。 结束时：〈最终画面状态〉。 【保持一致】 保持〈人物身份、数量、服装、道具归属、空间方向和声音关系〉稳定。

The image must contain the following text exactly as written, with each item appearing only once: "[HEADLINE]" "[SUBTITLE]" Place the headline at [POSITION] and the subtitle at [POSITION]. Do not add any other text.

Use a restrained color palette and a clear visual hierarchy. Preserve realistic material details and reduce decoration. Leave a large area of quiet negative space around the subject. Use soft side lighting to reveal surface contours, avoiding oversharpening and exaggerated highlights.

First identify the clearest relationship within the theme, then translate it into a visible visual event. Use only one central metaphor per image. Let the subject, action, and spatial relationships carry the meaning. Do not crowd the image with multiple symbols or rely on explanatory text to make it understandable.

【编辑目标】 编辑@视频1，在〈全片或明确时间段〉对〈画面对象、区域或声音类别〉进行〈增加、移除、替换或调整〉。 【原视频职责】 @视频1是唯一编辑母版，负责〈人物、场景、动作、构图、镜头、遮挡关系、声音和事件顺序〉。 【目标素材职责】 @图片1或@音频1用于〈目标对象或声音的指定属性〉。 【编辑范围】 只处理〈对象、区域、时间段或声音类别〉。 【保持内容】 保持@视频1中的〈不应变化的画面、动作、声音和时间关系〉。

Frame [person] from the chest up against a warm gray backdrop. Place a large softbox at camera left and a white reflector opposite. Use an 85mm portrait perspective. Keep pores visible, eyes sharp, and wardrobe understated.

Prepare a square profile portrait of [person] for [profession]. Leave room around the head for a circular crop. Choose a relaxed expression and indirect window light. Avoid beauty filters and distracting accessories.

Photograph [model] wearing [outfit] beside a sculptural concrete wall. Let the garment silhouette lead the composition. Use a low camera position, a restrained palette, and crisp directional light. Include believable fabric folds.

【编辑目标】 编辑@视频1，只把〈原对象〉修改为〈目标对象〉。 【原视频职责】 @视频1是唯一编辑母版，负责原始场景、机位、镜头运动、动作轨迹、遮挡关系和事件顺序。 【目标素材职责】 @图片1用于〈目标对象〉的〈外观、结构或材质〉，不采用〈无关背景、人物或构图〉。 【编辑对象与范围】 只修改〈明确对象和区域〉。全片目标对象数量为〈数量〉。不修改〈需要保持的内容〉。 【时间线继承】 〈目标对象〉继承〈原对象〉每次出现、运动、遮挡和离开的时点、持续时长、路径与速度变化。 除以上修改外，@视频1中的人物、道具、场景、镜头切换和事件顺序保持原样。