GPT Image 2 提示词指南：80+ 示例与 API 技巧

如果你想获得更好的 GPT Image 2 结果，不要从松散的风格词开始，而要从结构化 prompt 开始。最稳定的写法是：先说明图片任务，定义主体，锁定需要出现的精确文字，描述构图和光线，再写清哪些内容不能改变。这个结构适用于海报、产品图、UI mockup、信息图、角色设定图、图片编辑和图生视频第一帧。

本指南提供 8 个创意角度下的 80 个可复制 GPT Image 2 prompt，并包含可复用 prompt 模板、API 价格说明、设置建议、常见失败点，以及把通过审核的静态图转成视频的 PixVerse 工作流。它面向需要可用图片结果的创作者、营销人员、设计师和开发者，而不只是一次性生成好看的图片。

OpenAI 于 2026 年 4 月 21 日发布了 ChatGPT Images 2.0。许多创作者也会用 GPT Image 2、gpt-image-2 或 ChatGPT Images 2.0 来搜索同一套图像生成体验。我们在发布周进行了首次测试，并在 2026 年 6 月 4 日根据 OpenAI 官方 prompting guide 和 API 价格页重新检查了本文。

GPT Image 2 一览

GPT Image 2 是什么？

GPT Image 2 是 OpenAI ChatGPT Images 2.0 图像生成体验的创作者常用简称。它真正的优势不是泛泛的“好看图片”，而是结构化视觉任务：可读文字、清晰版式、产品级构图，以及后续可编辑的参考图。

GPT Image 2 最适合什么？

优先用于文字较多的海报、产品广告、UI mockup、信息图、角色设定图、教学图和可编辑视觉 brief。图像越依赖版式、标签、层级和 prompt 遵循度，结构化 GPT Image 2 prompt 就越有价值。

GPT Image 2 仍然容易在哪些地方出错？

精确复现品牌 Logo、专有字体、极小法律文字，以及速度比精度更重要的大批量草图仍需谨慎。如果素材必须保留真实 Logo、授权字体、产品标签或合规文本，生成后要安排人工检查，必要时再做合成。

GPT Image 2 可以在 ChatGPT 和 API 中使用吗？

对于 ChatGPT 访问，OpenAI Help 显示 ChatGPT Images 2.0 面向所有级别开放；images with thinking 面向 Plus、Pro 和 Business。对于 API 工作流，OpenAI 列出了 GPT-Image-2 的图像输入、缓存图像输入、图像输出、文本输入和缓存文本输入 token 价格。

PixVerse 如何接入 GPT Image 2 工作流？

本文提供同一组 8 个实用创意角度下的 80 个 prompt。如果静态图需要成为视频源，可以在 PixVerse 中生成或导入图片，必要时比较不同图像模型，然后用图生视频把通过审核的画面转成动态内容。

如何写出真正有效的 GPT Image 2 Prompt

好的 GPT Image 2 prompt 不只是描述一张图，而是说明这张图要完成什么任务。社交广告、产品抠图、信息图、UI 屏幕和视频起始帧，写法都应该不同。

一个可靠的起手模板如下：

Create [type of image] for [use case].
Main subject: [specific subject and visible details].
Exact text, if any: "[copy that must appear]".
Composition: [framing, layout, negative space, subject placement].
Style and lighting: [visual language, medium, mood, light direction].
Constraints: [what must not change, no extra words, no watermark].
Output format: [aspect ratio, transparent background, video-ready frame].

技巧 1：先说明任务，再说风格

先写输出类型：海报、产品广告、App 界面、角色设定图、教学图、编辑任务，或图生视频第一帧。GPT Image 2 先理解成功标准，才更容易遵循你的细节。

较弱 prompt：

A cool futuristic speaker, cinematic, high detail.

更好的 prompt：

Create a premium product ad for a matte black wireless speaker. The image should work as a 16:9 campaign banner, with the product on the right, a short headline on the left, clean negative space, and sharp product edges.

第二个 prompt 不只是要求好看，而是明确了布局、层级和可用性。

技巧 2：把文字当作需要锁定的素材

如果图中需要文字，把文字放进引号，并说明如何渲染。不要只说“加一句口号”，除非你希望模型自己编文案。

可以使用这个写法：

Headline: “SOUND YOU CAN FEEL”. Render the headline verbatim. No extra words, no duplicate text, no fake logo. Bold white sans-serif type, left side of the composition, readable from a distance.

长文案建议拆成多行。如果结果拼错了，减少文字量，放大字号，并加入更严格的“只使用精确文字”约束。

技巧 3：给模型一个镜头和版式

GPT Image 2 能理解构图，但你需要明确写出镜头距离、角度、主体位置、留白和宽高比。

常用短语：

• Close-up for product texture, hands, faces, materials, labels.
• Wide shot for environments, story scenes, city posters, and video-ready frames.
• Top-down for food, desk scenes, flat-lays, packaging kits.
• Left third / right third for ad layouts with text and product balance.
• Clean grid for UI mockups, character sheets, diagrams, and infographics.

技巧 4：编辑任务用三句话写清楚

编辑类 prompt 最好拆成三部分：要改什么、哪些内容锁定不变、如何保持物理真实。

Replace the parked car with a vintage bicycle.
Preserve the house, fence, driveway, landscaping, lighting direction, camera angle, and time of day exactly.
Match the bicycle scale, contact shadow, and perspective to the existing scene.

这种写法比“让它更好看”更可靠，因为它告诉 GPT Image 2 哪里可以发挥，哪里不能动。

技巧 5：图片要转视频时，提前写运动线索

如果静态图后续会进入 PixVerse 图生视频，请为画面写出纵深和运动准备。可以要求前景、中景、背景、清晰主体轮廓，以及一个可见运动线索：尘土、布料、头发、雨、反射、车辆移动、产品旋转或镜头推进路径。

不要只写：

An astronaut in a desert.

可以改成：

A cinematic first frame for an image-to-video clip: a lone astronaut standing at the edge of a glowing desert crater at dawn, cape and dust ready to move in the wind, strong foreground silhouette, clear depth layers, and warm horizon light.

我们如何测试 GPT Image 2

我们围绕人像、文字密集海报、产品构图、角色设定图、UI mockup 和实验性叙事画面测试 GPT Image 2。目标不是做实验室式跑分，而是判断设计师、营销人员或创作者能否只做轻微修改就使用结果。

结论很清楚：GPT Image 2 更奖励精确 brief，而不是关键词堆砌。当 prompt 说明任务和成功标准时，模型更容易保持结构。

8 个 GPT Image 2 Prompt 角度：80 个可复制示例

下面每个角度包含 10 条 prompt。每组第 1 条最适合作为配图示例，因为它最能体现该角度要测试的能力；其余 prompt 可直接复制、改写和测试。

为便于跨模型复制，并与现有英文示例图保持一致，以下 prompt 示例保留英文；说明、技巧和 FAQ 已本地化。

1. 写实与电影感场景

适合人像、编辑视觉、生活方式场景和需要光影真实感的氛围画面。

Prompt 1:

Generate a cinematic portrait of a solitary figure standing in an intense orange-to-red gradient environment. Strong silhouette lighting from behind, deep shadow contrast, reflective glossy floor mirroring the figure. Symmetrical composition, minimal set design, no background clutter. The mood is contemplative and powerful, like a still from a science-fiction film. Aspect ratio 16:9.

Prompt 2:

A candid photorealistic street scene in Seoul after rain. A florist closes a small shop at blue hour, wet pavement reflections, warm shop light, tired natural posture, 50mm documentary feel, realistic skin texture, no glamour pose, no watermark. Aspect ratio 3:2.

Prompt 3:

Close-up of weathered hands repairing an old film camera on a scratched wooden desk. Window light from camera-left, visible dust, brass and black leather texture, shallow depth of field, quiet workshop mood, photorealistic, no text overlay. Aspect ratio 4:3.

Prompt 4:

A quiet overnight train platform in northern Europe during light snow. One traveler in a long coat stands under a warm station lamp, breath visible in cold air, train windows glowing in the background, cinematic realism, restrained color palette, 35mm documentary feel, no text. Aspect ratio 16:9.

Prompt 5:

A top-down editorial food photograph of handmade noodles on a dark ceramic plate, steam rising, chopsticks resting at an angle, worn wooden table, soft side light, realistic oil sheen and texture, no branding, no text overlay. Aspect ratio 4:5.

Prompt 6:

A realistic documentary-style portrait of a ceramic artist trimming a clay bowl on a pottery wheel. Medium close-up, hands and spinning clay in sharp focus, apron with natural stains, soft workshop window light, shelves of unfinished bowls in the background, honest texture, no glamour retouching, no text. Aspect ratio 3:2.

Prompt 7:

A wide cinematic still of a small mountain town after a summer storm. Mist rises from dark green pine trees, warm lights appear in cottage windows, wet road reflections lead toward the center, one person walking with an umbrella in the distance, natural scale, realistic atmosphere, no text. Aspect ratio 16:9.

Prompt 8:

A close-up photorealistic shot of a vintage wristwatch resting on a folded linen cloth. Visible brushed metal, tiny scratches on the case, readable but fictional watch face markings, soft directional morning light, shallow depth of field, refined editorial product-photo mood, no real brand logos. Aspect ratio 4:5.

Prompt 9:

A candid indoor scene of a small architecture studio late at night. Two designers review foam models and printed floor plans under a desk lamp, coffee cups nearby, realistic shadows, practical workspace clutter, calm focused mood, 35mm film look, no text overlay. Aspect ratio 16:9.

Prompt 10:

A natural fashion editorial image of a model in a simple cream coat standing near a subway entrance at dusk. Streetlights beginning to glow, muted city background, realistic fabric folds, relaxed posture, eye-level framing, subtle film grain, no visible brand names, no text. Aspect ratio 2:3.

观察重点: 检查光线方向、反射或阴影是否可信，人物姿态是否自然。如果画面过度精修，加入更多纪实细节，减少空泛的高级感描述。

2. 精准文字海报与字体设计

当 prompt 把文字当作设计要求，而不是装饰元素时，GPT Image 2 表现最好。

Prompt 11:

A striking Spring 2026 city poster for New York with a bold contemporary design and an elegant celebratory mood. Clean off-white textured background with generous negative space. A miniature kayaker paddles across a narrow ribbon of reflective water in the lower-right corner. The wake sweeps upward in a dynamic calligraphic curve, gradually transforming into the Hudson River and then into a dreamlike hand-painted panorama of Manhattan. Inside the flowing river-shaped composition: the Empire State Building, Brooklyn Bridge, Central Park canopy, One World Trade Center, brownstone rooftops, yellow cabs, harbor ferries, and the Statue of Liberty in soft distance. Soft morning fog, golden spring light, subtle accents in navy and gold. Elegant typography in the lower left reads “SPRING 2026” with a vertical slogan “NEW YORK - A CITY OF BRIDGES, DREAMS, AND REINVENTION”. Text must be sharp and beautifully composed. No extra words. Premium graphic design, aspect ratio 9:16.

Prompt 12:

Create a vertical launch poster for a fictional design conference called “FRAME 2026”. Large headline: “FRAME 2026”. Subtitle: “DESIGNING WITH MACHINE IMAGINATION”. Clean Swiss grid, off-white background, black typography, one red geometric accent, generous negative space, perfectly legible text, no extra words, no watermark. Aspect ratio 9:16.

Prompt 13:

Create a minimalist album cover titled “SOFT SIGNALS”. Artist name: “MIRA VALE”. Centered typography, muted blue paper texture, small silver line illustration of a radio tower, elegant spacing, no extra text, no logo, aspect ratio 1:1.

Prompt 14:

Create a bookstore window poster reading “READ MORE SLOWLY” in large serif type. Smaller line: “SPRING READING WEEK”. Warm evening street reflections in the glass, cream paper texture, readable typography, no extra words, no watermark. Aspect ratio 4:5.

Prompt 15:

Create a museum exhibition poster titled “OBJECTS OF TOMORROW”. Subtitle: “A DESIGN HISTORY OF 2026”. Black text on off-white paper, one abstract chrome object in the center, clean modernist layout, exact readable text only, no fake logos. Aspect ratio 9:16.

Prompt 16:

Create a vertical music festival poster with the exact headline “AFTERLIGHT SESSIONS”. Smaller text: “JUNE 12-14”. Use a deep navy background, one glowing circular stage light, elegant condensed sans-serif typography, balanced negative space, exact text only, no extra words, no watermark. Aspect ratio 9:16.

Prompt 17:

Create a clean cafe menu board titled “MORNING MENU”. Include exactly four items: “ESPRESSO”, “MATCHA LATTE”, “CARDAMOM BUN”, “COLD BREW”. Warm cream background, black serif type, simple divider lines, readable from a distance, no prices, no extra items. Aspect ratio 4:5.

Prompt 18:

Create a square social campaign graphic for a fictional running club. Main text: “RUN THE RIVER”. Secondary line: “SATURDAY 7 AM”. Bold kinetic typography, abstract river line, bright green and black palette, clear hierarchy, no extra text, no real logos. Aspect ratio 1:1.

Prompt 19:

Create a book cover for a fictional novel titled “THE QUIET MACHINE”. Author name: “ELENA ROWE”. Minimalist cover with a small silver mechanical bird silhouette, matte black background, refined typography, exact text only, no publisher logos, no extra copy. Aspect ratio 2:3.

Prompt 20:

Create a classroom poster titled “ASK BETTER QUESTIONS”. Include three short lines: “Observe”, “Explain”, “Test”. Friendly editorial design, soft yellow background, simple line icons, high contrast readable text, no extra words, no watermark. Aspect ratio 4:5.

观察重点: 先检查每个字母是否可读。如果模型添加额外文字，请重申“exact text only”，并把每行文字单独写出。

3. 产品摄影与广告创意

适合广告视觉、主视觉、电商图、社交广告和产品故事。

Prompt 21:

A premium product ad for a matte black wireless speaker on a concrete plinth. Headline: “SOUND YOU CAN FEEL”. Product on the right, bold white type on the left, dramatic rim light, clean shadow, luxury tech campaign style, sharp product edges, no fake brand logo, no watermark. Aspect ratio 16:9.

Prompt 22:

Editorial skincare serum photo on frosted glass. A translucent bottle with a simple label reading “LUMA SERUM”, soft diffused light, pale green background, high-end beauty campaign style, label text sharp, clean reflection, no extra props, aspect ratio 4:5.

Prompt 23:

Square social ad for a durable travel bottle on a mountain trail at golden hour. Tagline: “BUILT FOR THE LONG WAY”. Product clearly visible in the foreground, natural hand grip, warm sunlight, crisp readable text in lower third, no extra words, aspect ratio 1:1.

Prompt 24:

A clean e-commerce product photo of wireless headphones on a pure white background. Straight-on angle, crisp silhouette, subtle contact shadow, visible ear cushion texture, no text, no logo, no props, high-resolution product photography. Aspect ratio 1:1.

Prompt 25:

A billboard-style campaign visual for a ceramic coffee cup. Headline: “MORNINGS, REHEATED”. Product large in the foreground, warm kitchen window light, soft steam, bold readable type in upper left, no extra copy, no watermark. Aspect ratio 16:9.

Prompt 26:

A premium ecommerce hero image of a minimalist hiking backpack on a stone ledge. Product centered, front pocket and straps visible, soft alpine morning light, clean shadow, no person, no logo, no text overlay, realistic nylon texture and zipper details. Aspect ratio 1:1.

Prompt 27:

A polished skincare campaign image for a frosted glass moisturizer jar. Headline: “CALM IN A JAR”. Product in lower right, pale blue background, soft water reflections, crisp label area with no fake brand, elegant white typography, no extra words. Aspect ratio 4:5.

Prompt 28:

A cinematic product photo of matte white wireless earbuds in an open charging case. Dark charcoal background, thin rim light, subtle reflection underneath, clean negative space for a campaign headline, no logo, no text, sharp product edges. Aspect ratio 16:9.

Prompt 29:

A square snack packaging mockup for a fictional granola brand called “NOVA OATS”. Show one pouch standing upright on a light wood surface, label text sharp, oats and dried fruit around the base, warm natural light, premium but approachable packaging design, no extra brands. Aspect ratio 1:1.

Prompt 30:

A luxury jewelry product shot of a silver ring with a small blue stone on a dark velvet surface. Macro detail, realistic metal reflections, soft spotlight from upper left, clean shadow, no hands, no text, no watermark, product clearly separated from background. Aspect ratio 4:5.

观察重点: 产品必须是视觉主角。如果模型擅自发明包装细节，使用参考图时加入“preserve the input product exactly”。

4. 信息图与教学视觉

适合图解、流程图、教学视觉、博客配图和需要清晰标签的图形。

Prompt 31:

Create a clean infographic titled “HOW IMAGE PROMPTS WORK”. Five labeled steps: “Scene”, “Subject”, “Text”, “Composition”, “Constraints”. Flat editorial icons, arrows between steps, high contrast, white background, readable sans-serif labels, consistent spacing, no extra text, no watermark. Aspect ratio 16:9.

Prompt 32:

Educational diagram showing the layers of a camera lens. Include labeled parts: “Front Element”, “Aperture”, “Focus Group”, “Image Sensor”. Clean cutaway illustration, white background, textbook style, clear leader lines, readable labels, no decorative clutter. Aspect ratio 16:9.

Prompt 33:

Comparison infographic titled “POSTER PROMPT VS PRODUCT PROMPT”. Two columns, six rows, concise labels, neutral background, black text, blue accent lines, professional blog graphic style, all copy readable, no extra text. Aspect ratio 16:9.

Prompt 34:

Create a step-by-step instructional visual titled “HOW TO MAKE COLD BREW”. Five illustrated steps with short labels: “Grind”, “Steep”, “Filter”, “Pour”, “Serve”. Warm earth tones, clear arrows, consistent icon style, readable text, no extra words. Aspect ratio 16:9.

Prompt 35:

Create a clean comparison chart titled “AI IMAGE WORKFLOW”. Three columns: “Draft”, “Refine”, “Animate”. Use simple icons, short labels, high contrast, generous spacing, white background, professional blog graphic style, all text readable. Aspect ratio 16:9.

Prompt 36:

Create a clean timeline infographic titled “FROM PROMPT TO POSTER”. Five stages: “Brief”, “Layout”, “Text”, “Review”, “Export”. Horizontal flow, simple numbered circles, blue and black accent palette, high contrast labels, no extra text, no watermark. Aspect ratio 16:9.

Prompt 37:

Create an educational diagram titled “REFERENCE IMAGE ROLES”. Three labeled cards: “Subject”, “Style”, “Background”. Show simple image thumbnails, arrows into one final output frame, clear labels, white background, consistent spacing, no extra text. Aspect ratio 16:9.

Prompt 38:

Create a decision tree titled “WHICH IMAGE PROMPT?”. Branches: “Text”, “Product”, “Scene”, “Edit”. Use clean boxes and arrows, readable sans-serif typography, minimal gray background, one green accent color, no extra words, no decorative clutter. Aspect ratio 16:9.

Prompt 39:

Create a safety checklist infographic titled “BEFORE YOU GENERATE”. Four checks: “Rights”, “Privacy”, “Text”, “Brand”. Use simple check icons, concise labels, white background, professional SaaS help-center style, high contrast, no extra copy. Aspect ratio 4:5.

Prompt 40:

Create a visual explainer titled “IMAGE EDITING PROMPT”. Three stacked rows: “Change”, “Preserve”, “Match”. Include tiny example icons for each row, clean leader lines, readable labels, restrained colors, no extra text, no watermark. Aspect ratio 16:9.

观察重点: 先检查标签文字。如果画面好看但文字错误，这张图就不可用。复杂图表应减少标签数量后重试。

5. 角色设计与设定图

角色设定图能把身份、服装、配色和表情压缩到一张可复用参考图中。

Prompt 41:

Create a professional character reference sheet for an original fantasy RPG character: a young female mage with silver hair and violet eyes, wearing an ornate dark cloak with glowing rune patterns. Include on a clean white background: a three-view turnaround showing front, side, and back; facial expression variations showing neutral, smiling, angry, and surprised; detailed breakdowns of costume and equipment pieces; a color palette swatch row; and brief world-building notes in clean typography. Organized grid layout, concept art style, high resolution. Aspect ratio 16:9.

Prompt 42:

Create a sci-fi courier character sheet for an original character named “NOVA”. Include front, side, and back views, four facial expressions, jacket and backpack callouts, color palette swatches, clean white background, readable labels, consistent face and jacket across all views. Aspect ratio 16:9.

Prompt 43:

Create a children’s book character sheet for a small forest helper in a green raincoat. Include expression row, prop row, walking pose, waving pose, color palette, simple readable notes, soft illustration style, no extra characters. Aspect ratio 16:9.

Prompt 44:

Create a cyberpunk detective character sheet for an original character named “REI”. Include front view, side view, back view, three expressions, trench coat callouts, device props, neon color palette, clean labels, consistent face and hairstyle. Aspect ratio 16:9.

Prompt 45:

Create a mascot reference sheet for a friendly robot baker. Include full-body pose, three facial display expressions, apron details, pastry props, color palette swatches, simple turnaround, clean white background, readable labels. Aspect ratio 16:9.

Prompt 46:

Create a mobile game character sheet for an original desert scout named “KAI”. Include front, side, and back views, three action poses, scarf and utility-belt callouts, color palette swatches, readable labels, consistent face and outfit, clean off-white background. Aspect ratio 16:9.

Prompt 47:

Create a cozy fantasy village merchant character sheet for an original character named “MARN”. Include full-body front view, side view, prop row with lantern and ledger, four expression studies, fabric texture callouts, warm color palette, clean grid layout, readable notes. Aspect ratio 16:9.

Prompt 48:

Create a sci-fi maintenance drone design sheet. Include top, side, and front views, small detail panels for sensors, landing feet, tool arm, battery pack, and warning lights. Clean technical concept-art layout, neutral background, readable labels, consistent industrial design. Aspect ratio 16:9.

Prompt 49:

Create a children’s animation character sheet for an original classroom inventor named “MILO”. Include one standing pose, one thinking pose, one excited pose, expression row, backpack and notebook props, bright but restrained palette, readable labels, no extra characters. Aspect ratio 16:9.

Prompt 50:

Create a tactical costume reference sheet for an original cyberpunk courier. Include front, back, and side views, jacket callouts, shoe detail, messenger bag detail, color swatches, three silhouette poses, crisp label text, consistent hairstyle and face across views. Aspect ratio 16:9.

观察重点: 同一张脸、服装和配色应在多个视角中保持一致。如果侧面图改变了服装，加入更强的 preserve 约束。

6. UI 与 App Mockup

适合 App 概念、仪表盘、社交主页和可用于团队讨论的产品界面。

Prompt 51:

A hyper-realistic iPhone screenshot of a fictional Instagram profile page for Leonardo da Vinci, username @davinci_official, as if he were a modern influencer in 2026. Profile photo is a Renaissance self-portrait in a circle crop. Bio reads: “Artist, Engineer, Inventor | Currently dissecting things | DM for commissions”. The grid shows 9 posts: the Mona Lisa reframed as a mirror selfie, a helicopter sketch captioned “just dropped my new drone design”, an anatomy study posted as a gym progress photo, The Last Supper staged as a dinner party group shot, and other creative anachronistic mashups. Follower count: 12.4M. Story highlights labeled Sketches, Inventions, and Florence Life. Complete iOS status bar with carrier text reading “Renaissance 5G”, battery icon, and current time. Dark mode UI throughout. Photorealistic screenshot quality, aspect ratio 9:16.

Prompt 52:

A realistic mobile onboarding screen for a fictional habit app called “LUMA”. Headline: “BUILD BETTER DAYS”. Buttons: “Start now” and “View demo”. Clean iOS-style layout, soft white background, blue accent, readable UI text, shown straight-on inside a phone frame. Aspect ratio 9:16.

Prompt 53:

Desktop SaaS dashboard for an e-commerce analytics tool. Left sidebar, top KPI cards for Revenue, Orders, Conversion Rate, a line chart, and a top-products table. Clean white interface, realistic spacing, readable labels, no real brand names. Aspect ratio 16:9.

Prompt 54:

A realistic mobile weather app screen for a fictional app called “SKYLINE”. Current city: “Lisbon”. Headline temperature: “22C”. Cards for Wind, Humidity, UV, and Sunset. Calm blue interface, readable labels, iPhone frame, no real app branding. Aspect ratio 9:16.

Prompt 55:

A restaurant booking app screen showing a reservation confirmation. Restaurant name: “North Table”. Date: “June 18”. Time: “7:30 PM”. Party size: “4 guests”. Warm editorial food photo at top, clean CTA button reading “Add to calendar”, readable UI text. Aspect ratio 9:16.

Prompt 56:

A realistic desktop analytics dashboard for a fictional creator studio. Left navigation, top cards for Views, Watch Time, Revenue, and New Followers, a line chart, and a campaign table. Clean white UI, blue accent, readable labels, practical spacing, no real brand names. Aspect ratio 16:9.

Prompt 57:

A mobile checkout screen for a fictional outdoor gear shop called “TrailCart”. Show product thumbnail, quantity stepper, shipping address card, discount field, total price, and a CTA button reading “Place order”. Modern iOS style, readable UI text, no real logos. Aspect ratio 9:16.

Prompt 58:

A tablet UI mockup for a prompt library app. Show tabs labeled “Posters”, “Products”, “UI”, and “Edits”. Main panel includes three prompt cards with short preview text, copy buttons, and category chips. Clean interface, high legibility, no real brand names. Aspect ratio 4:3.

Prompt 59:

A SaaS settings screen for a fictional AI image tool. Sections labeled “Model”, “Quality”, “Aspect Ratio”, “Reference Images”, and “Safety”. Use toggles, dropdowns, sliders, and a clear Save button. Quiet professional UI, readable labels, no decorative clutter. Aspect ratio 16:9.

Prompt 60:

A mobile travel itinerary app screen for a fictional trip to Kyoto. Header reads “Kyoto Weekend”. Cards for “Day 1”, “Day 2”, “Temple Walk”, and “Dinner”. Soft neutral UI, realistic spacing, small map preview, readable text, no real app branding. Aspect ratio 9:16.

观察重点: 界面应像真实产品，而不是装饰海报。重点检查导航、按钮文字、图标间距和信息层级。

7. 叙事分镜与实验艺术

短叙事 prompt 可以测试 GPT Image 2 对视觉笑点、多格故事和场景内小文字的理解。

Prompt 61:

Inside a museum exhibit titled “Ancient Technology: The Desktop Era”, a programmer in a glass display case is live-demonstrating coding on a CRT monitor while amazed schoolchildren press their faces against the glass. The exhibit placard reads: “Homo Developerus (c. 2005) - Primitive human using keyboard-based input devices.” A second display case nearby shows a physical book labeled “Stack Overflow - Print Edition, Vol. 1 of 4,827”. 2D cartoon illustration style, warm museum lighting, humorous and nostalgic tone. Aspect ratio 16:9.

Prompt 62:

A four-panel comic strip titled “MORNING ROUTINE”. Panel 1: alarm goes off. Panel 2: character makes coffee. Panel 3: character sits down to work. Panel 4: character is already asleep at the desk. Warm minimal illustration style, expressive character, readable title, no extra text. Aspect ratio 16:9.

Prompt 63:

A single editorial illustration for an article about creative automation. A designer and an AI assistant arrange paper storyboards on a large table, soft studio light, subtle humor, modern magazine illustration style, no visible brand logos, no text. Aspect ratio 3:2.

Prompt 64:

A newspaper front-page style illustration titled “THE MORNING HERALD”. Main headline: “CITY APPROVES ROOFTOP GARDENS”. Two-column layout, one photorealistic city council photo area, classic broadsheet design, readable masthead and headline, no extra article text. Aspect ratio 4:5.

Prompt 65:

A two-panel comic about a robot learning to paint. Panel 1: the robot carefully studies a blank canvas. Panel 2: the robot proudly shows a messy but charming painting. Warm studio lighting, expressive body language, no speech bubbles, simple title: “FIRST ATTEMPT”. Aspect ratio 16:9.

Prompt 66:

A three-panel editorial comic titled “THE DEADLINE”. Panel 1: a designer calmly opens a blank file. Panel 2: the clock jumps forward and sticky notes cover the desk. Panel 3: the designer presents a polished poster with surprised relief. Minimal expressive illustration style, readable title, no speech bubbles. Aspect ratio 16:9.

Prompt 67:

A surreal magazine illustration about creative focus: a person sits at a small desk floating in a quiet library of glowing windows, each window showing a different unfinished idea. Soft cinematic lighting, thoughtful mood, clean composition, no visible brand logos, no text. Aspect ratio 3:2.

Prompt 68:

A four-panel storyboard for a product launch teaser. Panel 1: closed box on a table. Panel 2: light leaking from the box. Panel 3: hands lifting the lid. Panel 4: glowing product silhouette revealed. No readable brand, no dialogue, cinematic lighting, clear panel borders. Aspect ratio 16:9.

Prompt 69:

A humorous museum diorama titled “THE FIRST GROUP CHAT”. Show ancient-looking figures gathered around stone tablets with message bubbles carved above them, warm museum lighting, playful editorial illustration, readable title only, no extra text. Aspect ratio 16:9.

Prompt 70:

A split-screen narrative poster showing “before” and “after” creative iteration. Left side: messy sketch wall and rough notes. Right side: clean polished campaign board. Modern editorial illustration, strong contrast, no logos, no extra words beyond “BEFORE” and “AFTER”. Aspect ratio 16:9.

观察重点: 场景应该靠视觉本身传达想法。如果笑点完全依赖文字，简化设定并让动作更清楚。

8. 图片编辑、多参考图与视频起始帧

这一类让 GPT Image 2 不只是首轮生成器，还能用于抠图、换装、换背景、多参考图编辑和视频起始帧。

Prompt 71:

Create a cinematic first frame for an image-to-video clip: a lone astronaut standing at the edge of a glowing desert crater at dawn, cape and dust ready to move in the wind, strong foreground silhouette, clear depth layers, warm horizon light, no text, no watermark. Aspect ratio 16:9.

Prompt 72:

Use Image 1 as the product photo and Image 2 as the background style reference. Place the product into the scene from Image 2. Preserve the product shape, label text, proportions, color, and material exactly. Match lighting, scale, shadow, and perspective. Do not restyle the product. No extra logos or watermark.

Prompt 73:

Remove the background from the input product image. Output a transparent background with crisp silhouette, clean edges, no halos, no fringing. Preserve bottle geometry, cap shape, label text, label colors, and print sharpness exactly. Do not change proportions.

Prompt 74:

Change only the weather and lighting in the input image. Make the scene look like a winter evening with light snowfall. Preserve the people, buildings, signs, camera angle, object placement, and composition exactly. Keep all readable text unchanged.

Prompt 75:

Image 1 is the person to preserve. Image 2 is the jacket reference. Image 3 is the boots reference. Dress the person from Image 1 using the clothing from Images 2 and 3. Preserve face, body shape, pose, hands, background, camera angle, and lighting exactly. Replace only the clothing.

Prompt 76:

Use the input product photo as the locked subject. Place the product on a clean marble bathroom counter with soft morning window light. Preserve product shape, label text, cap color, proportions, and material exactly. Match contact shadow, scale, and perspective. Do not add extra labels, logos, or props.

Prompt 77:

Create a cinematic first frame for an image-to-video clip: a glass perfume bottle standing on wet black stone as a thin ribbon of mist moves behind it. Product centered, strong silhouette, clear foreground and background depth, no hands, no text, no watermark. Aspect ratio 16:9.

Prompt 78:

Edit the input portrait by changing only the background to a clean editorial studio backdrop in warm gray. Preserve the face, hair, clothing, pose, skin tone, camera angle, lighting direction, and expression exactly. Match the new background shadow and depth naturally.

Prompt 79:

Use Image 1 as the room photo and Image 2 as the wall-art reference. Add the artwork from Image 2 to the empty wall in Image 1. Preserve furniture, floor, window light, camera angle, color balance, and room layout exactly. Match frame scale, perspective, and wall shadow.

Prompt 80:

Create a video-ready first frame for a product reveal: a closed matte black box on a table, thin blue light leaking from the seam, dust particles visible in the beam, camera positioned low and close, strong depth layers, empty space for motion, no text, no logo. Aspect ratio 16:9.

观察重点: 编辑任务成功的前提是锁定细节不被破坏。视频起始帧则要检查主体分离、画面纵深和运动空间。

常见 GPT Image 2 Prompt 错误

• 想要准确文字，却没有给出精确文案。 如果图片需要文字，请写出准确文案和位置。
• 一个 prompt 塞入过多细节。 先生成核心场景，再一次只修改一个变量。
• 编辑时忘记写不变量。 明确哪些内容必须保持不变：身份、背景、姿势、光线、产品形状、标签文字或镜头角度。
• 用空泛美学词替代功能要求。 “Beautiful”不能保证标签可读。请使用“sharp label text”“clean kerning”“readable from a distance”。
• 跳过宽高比。 好看的方图未必适合竖版广告或视频封面。
• 把 Logo 当作普通文字。 GPT Image 2 可以设计 Logo 概念，但精确品牌标志通常应使用官方素材后期合成。

GPT Image 2 API 与价格说明

OpenAI 的 API pricing page 以 token 方式列出 GPT-Image-2 价格。截至 2026 年 6 月 4 日，页面列出的价格如下：

实际生成成本取决于 prompt 长度、参考图、输出尺寸、缓存、质量设置和访问方式。如果使用 ChatGPT 而不是 API，套餐限制和 API token 价格是两套规则。构建重复生成或批量工作流时，也建议阅读 OpenAI 的 GPT Image Generation Models Prompting Guide。

对团队来说，问题不只是“GPT Image 2 能不能生成静态图”，还包括“静态图通过后下一步怎么做”。如果想比较第一版图片质量，可以看我们的 GPT Image 2 vs Nano Banana 2 同 prompt 测试。如果需要从终端或 AI agent 自动生成图片和视频，PixVerse CLI 指南 覆盖了命令行工作流。

在 PixVerse 中从图片到视频

生成一张强静态图只是第一步。许多工作流真正变慢的地方，是把角色图或产品海报下载下来，再上传到另一个工具，并希望视频模型不要扭曲设计。把最好的 GPT Image 2 输出当作视频源帧：它应该有清晰主体轮廓、明确景深层次，以及一个视频模型可以动起来的可见运动线索。

2026 年 4 月 22 日，PixVerse 将 GPT Image 2 作为文生图选项上线，与 Nano Banana 2、Seedream 和 HappyHorse 1.0 并列。你可以在应用内选择 GPT Image 2，生成图片后直接转成视频。

如果你正在用同一 brief 比较 OpenAI 和 Google 图像模型，请阅读我们的 GPT Image 2 vs Nano Banana 2 对比。如果你的 prompt 工作主要面向视频，best image-to-video AI tool guide 解释了静态图完成后如何选择运动工作流。

PixVerse 为创作者和团队提供更完整的生产管线：

• 文生图 使用 GPT Image 2、Nano Banana 2、Seedream 等模型，让你按任务选择模型。
• 图生视频 把通过审核的静态图转成动态内容，并保留角色和构图控制。
• 文生视频 使用 PixVerse V6 或电影感 C1 model 生成视频。
• 原生音频生成 为视频工作流添加音效和对白。

常见问题

GPT Image 2 和 ChatGPT Images 2.0 是一回事吗？

从搜索意图看，是的。许多用户用 GPT Image 2、gpt-image-2 和 ChatGPT Images 2.0 指代 OpenAI 新版 ChatGPT 图像生成体验。

最好的 GPT Image 2 prompt 结构是什么？

建议使用：任务、主体、精确文字、构图、风格、限制、宽高比，以及如果要转视频则加入运动线索。

如何让 GPT Image 2 正确拼写图中文字？

把精确文字放在引号里，说明位置，并加入“render the text verbatim”“no extra words”等约束。长文案应减少文字量或拆成更大的海报布局。

GPT Image 2 可以编辑已有图片吗？

可以。ChatGPT Images 支持通过选区或文字指令编辑生成图和上传图。最稳妥的写法是先说要改什么，再列出哪些内容必须不变。

GPT Image 2 支持透明背景吗？

可以为产品抠图和透明背景素材编写 prompt，但最终结果取决于界面、格式和设置。产品图建议强调清晰边缘、无毛边、保持几何形状和标签文字。

GPT Image 2 能使用多张参考图吗？

可以。多参考图适合用一张图定义主体，另一张定义风格、服装、背景或光线。请明确标注 Image 1、Image 2，并说明哪些内容转移、哪些必须保持。

GPT Image 2 免费吗？

OpenAI Help 显示 ChatGPT Images 2.0 面向所有级别开放，但不同套餐的配额、速度和 thinking-mode 访问并不相同。API 或 PixVerse 使用前请确认当前价格和积分规则。

GPT Image 2 API 多少钱？

OpenAI 价格页列出了 GPT-Image-2 的图像输入、缓存图像输入、图像输出、文本输入和缓存文本输入 token 价格。截至 2026 年 5 月 18 日，图像输出标价为每 100 万 token 30 美元。

GPT Image 2 和 Midjourney 相比如何？

Midjourney 的艺术风格控制和社区生态很强。GPT Image 2 更适合需要可读文字、UI 标签、图解、编辑指令和生产版式的结构化输出。

GPT Image 2 比 Nano Banana 2 更好吗？

取决于任务。GPT Image 2 强在文字版式、提示词遵循和可编辑视觉 brief；Nano Banana 2 可能更适合快速图像探索或特定视觉风格。

我可以把 GPT Image 2 输出转成视频吗？

可以。在 PixVerse 中，GPT Image 2 生成的图片可以直接进入图生视频工作流。你也可以上传外部生成的 GPT Image 2 图片。

应该先尝试哪些 GPT Image 2 prompt？

先从成功标准很清楚的 prompt 开始：精准文字海报、产品广告、UI mockup、信息图和图片编辑。这些类别方便你检查 GPT Image 2 是否真正遵循 brief，因为可以直接验证拼写、层级、产品保留、版式和锁定细节。