GPT IMAGE 2 IMAGE CREATION

GPT Image 2 AI Image Generator

GPT Image 2 is OpenAI’s image-generation and editing model. Create images from text or work from reference photos to make posters, product visuals, portraits and new visual styles.

gpt-image-2
GPT IMAGE 2 EXAMPLES

Images Created with GPT Image 2

See how GPT Image 2 handles a portrait restyle, an outfit breakdown and a text-led poster. Each example includes its complete prompt and any reference images used.

WeShop result: Neon Sculpture Style Transfer
Reference 1

Neon Sculpture Portrait

A portrait becomes a softly glowing sculpture, with neon gradients, rounded forms and translucent-looking surfaces.

Please transform the input image into a soft focus, neon gradient sculptural style. Do not copy specific specific characters, crowds, compositions, or poses from any reference, but follow this visual language: Style & Texture: Create a soft, dreamy, surreal 3D illustration and digital sculpture. The subject should look like translucent plastic, frosted glass, glowing rubber, or soft ceramic sculpture wrapped in colored mist. Smooth rounded edges, moderately simplified details, with a hazy, low-contrast, non-realistic artistic effect. Color & Light: Use highly saturated but softly diffused neon gradients (fluorescent pink, coral orange, lemon yellow, mint green, cyan, violet, pale grayish-blue). Colors should blend seamlessly like thermal imaging, colored mist, or dream halos. Shadows must be bluish-purple, cyan, or deep magenta—NO pure black hard shadows. Use strong inner glow, backlight, rim light, and ambient color reflections to make the subject glow from within. Include soft focus, slight mist, fine grain, and gradient noise. Subject Treatment: Retain the identity, pose, proportions, and clothing outlines of the input image, but transform them into a minimalist, rounded, dreamy neon sculpture. Soften and weaken facial features to create a quiet, ethereal, detached expression. The face should look like a soft mask covered by a glowing halo. Composition: Clean background (solid color, soft gradient, misty or large color blocks). Quiet, surreal, fairytale-like but slightly alienating atmosphere with a floating feeling. Keywords: soft neon gradient, dreamy 3D illustration, translucent sculpture, frosted glass texture, glowing rubber, pastel fluorescence, diffused light, inner glow, soft focus, blurred facial features, surreal portrait, minimal details, smooth rounded forms, luminous mist, thermal color gradient, ethereal atmosphere, clean composition, high-end digital art.
Use This Prompt
WeShop result: Reference-to-Editorial Breakdown
Reference 1

Portrait-to-Outfit Breakdown

One person photo becomes a presentation board combining the full outfit, individual garments, expression close-ups and material details.

You are an expert in ultra-detailed, photorealistic character breakdown sheets. Create a highly detailed character reference sheet based on the uploaded person photo. Preserve 100% of the person’s original facial features, identity, hairstyle, clothing, body proportions, and realistic textures. Do not redesign, beautify, or alter the person. TASK: CENTER: Place the full-body photorealistic image of the uploaded person prominently in the center. Keep the person’s original appearance, outfit, proportions, and visual identity exactly consistent with the reference photo. SURROUNDING ELEMENTS: • Outfit Breakdown: Display each clothing item separately, from outer layers to inner layers. Preserve the exact original design, silhouette, colors, materials, patterns, stitching, accessories, and fabric textures. Present each item as realistic product photography. • Expression Variations: Create 3–4 close-up portraits of the same person showing different natural expressions, such as calm, shy, surprised, and focused. The face, hairstyle, skin tone, makeup, and identity must remain fully consistent with the uploaded person. • Material Close-ups: Include realistic macro photography of key details such as fabric texture, stitching, skin texture, hair strands, accessories, buttons, hardware, and other visible materials from the original photo. • Personal Items: Arrange realistic everyday objects that naturally match the person’s fashion style and overall character, such as bag contents, cosmetics, diary, phone, accessories, or other appropriate personal belongings. Present them as premium editorial flat-lay photography. VISUAL GUIDELINES: Layout: Use a clean white background. Position the full-body person as the main visual focus in the center. Arrange all outfit pieces, facial close-ups, material details, and personal items evenly around the central figure. Use thin, refined editorial-style lines to connect relevant details to the person. Style: Hyper-realistic professional photography only. Use natural studio lighting, realistic skin, authentic fabric textures, accurate material reflections, subtle shadows, and premium fashion-editorial photography quality. Overall aesthetic: Clean, sophisticated, modern fashion magazine character analysis board with a New York editorial atmosphere. Maintain a balanced layout, generous white space, and refined visual hierarchy. TEXT: Display the main title: “Fashion Blogger – Photo Graph” Include one short English style/persona judgment: “Polished designer aesthetic, perfect for city commuting → New York” STRICT RULES: The final image MUST look like real photography. NO illustration, anime, cartoon, sketch, painting, CGI, or obvious 3D-rendered appearance. Preserve the uploaded person’s identity, facial structure, facial features, hairstyle, skin tone, clothing, body proportions, and overall appearance as accurately as possible. Do not change, redesign, beautify, replace, or reinterpret the person’s original features. Do not randomly change the clothing design, colors, materials, patterns, or accessories. All facial close-ups must clearly represent the same person. All clothing breakdowns and material close-ups must correspond accurately to the clothing worn by the central person. All surrounding elements must have authentic photographic textures, realistic lighting, and consistent visual quality. Maintain strong visual consistency between the central full-body character and every surrounding detail image.
Use This Prompt
WeShop result: Typography-Led Jazz Poster

Jazz Event Poster

Four text blocks separate the headline, venue, time and footer, paired with a saxophone silhouette and geometric color shapes.

Create a polished vertical 3:4 poster for a fictional jazz event. Use a warm ivory background, bold vermilion geometric shapes, a single black saxophone silhouette, and a carefully aligned editorial grid. Include exactly these four text lines, with no other words: 'MIDNIGHT JAZZ' as the largest headline, 'LIVE AT RIVER HALL' as the secondary line, 'FRIDAY 8 PM' as the small event detail, and 'ONE NIGHT ONLY' as a small bottom label. Make all four lines legible and spelled exactly. Keep a clear hierarchy between headline, illustration, details, and footer. Flat graphic design, crisp edges, restrained paper texture, no border or watermark.
Use This Prompt
WHAT IS GPT IMAGE 2?

What Can You Create with GPT Image 2?

Use text and reference images for graphic layouts, photo-based edits and style transformations.

Text and Graphic Layouts

Combine headlines, supporting copy, illustrations and color shapes for posters and infographics. The jazz example shows a clear hierarchy between its title and event details.

Reference-Based Editing and Layouts

Use an existing photo to establish a person, outfit or object, then change its presentation or arrange it in a new layout. The outfit example expands one person photo into a board of garments and details.

Changes in Style and Material

Specify colors, lighting, materials and a visual treatment to restyle an image. The neon example combines a recognizable portrait outline with a luminous sculptural finish.

WORKFLOWS

Generate from Text or Edit a Reference

Start a new image from a written description, or use reference images to supply a person, product, scene or layout.

Create from Text

Describe the subject, setting, style and layout when you do not need to retain an existing image. For text in the image, supply the exact copy and where it should appear.

Edit with Reference Images

Upload the image you want to edit and explain what to keep and what to change. With multiple references, identify which supplies the person, scene or composition.

PROMPT GUIDE

GPT Image 2 Prompt Techniques

These six examples show how to assign reference roles, describe materials, separate subject and layout, position lights, and specify text. Read each complete prompt with its inputs and result, then adapt the content to your own task.

WeShop result: Give Each Reference a Defined Role
Reference 1
Reference 2

Define Each Reference’s Role

Give each reference a clear job. Here, the first image supplies the desert environment and plants; the second supplies the court layout and camera angle. The prompt combines them into a new scene.

Use the two uploaded images as complementary visual references. REFERENCE ROLES: Image 1 is the environment and botanical reference. Use Image 1 to guide: - the open desert atmosphere - Joshua tree shapes and branching structure - dry natural terrain - scattered desert vegetation - spacious outdoor mood - warm arid landscape identity Image 2 is the composition and tennis-court structure reference. Use Image 2 to guide: - the camera angle - the overall court placement - court proportions - baseline and service-line geometry - the central net position - surrounding open space - the relationship between the court and the landscape Do not copy either image literally. Combine the environmental character of Image 1 with the tennis-court composition of Image 2 to create one coherent new scene. FINAL SCENE: Create a dreamy pastel tennis court located in a surreal desert landscape. The tennis court should use a refined pastel palette with: - soft blush pink outer areas - lavender and pale violet playing surfaces - clean white court lines - a subtle warm peach transition around the edges - a pale pink tennis net with realistic mesh texture Preserve believable tennis-court geometry. Keep the sidelines, baselines, service boxes, center marks, net height and court perspective visually coherent. ENVIRONMENT: Surround the court with rolling cream-colored and pale peach sand dunes. Add several Joshua trees inspired by Image 1, along with tall columnar cacti, round barrel cacti, desert succulents and small flowering plants. Scatter clusters of pink, coral, orange, yellow and lavender desert flowers around the edges of the court. Keep the central playing area clean and unobstructed. Place a simple metal bench near the foreground edge of the court and a minimal referee chair near one side of the net. The furniture should match the scene’s clean editorial styling without blocking the court. LIGHTING AND ATMOSPHERE: Use bright, soft natural daylight under a clear pale-blue sky. Add a few soft white clouds. Use gentle shadows, warm highlights and subtle atmospheric depth across the sand dunes. The mood should feel optimistic, imaginative, premium and calm. STYLE: High-end editorial landscape photography blended with polished surreal advertising art. Pastel color grading, clean commercial composition, realistic materials, precise perspective, crisp court details and refined botanical textures. The result should feel suitable for a sports-fashion campaign, lifestyle brand advertisement, social media post, website hero image or creative editorial. COMPOSITION: Use a wide landscape composition. Keep the tennis court as the central visual focus. Use desert plants and flowers to frame the foreground and side edges. Leave the sky and distant dunes relatively clean to prevent visual clutter. The scene must feel like one naturally designed environment, not a collage of two separate photographs. IMPORTANT: Do not add people or players. Do not add text, logos, signs, watermarks, brand names or promotional badges. Do not create extra tennis nets. Do not distort the court lines or perspective. Do not place cacti, flowers or furniture inside the active playing area. Do not reproduce the exact background from either reference image. Generate a new, coherent pastel desert tennis environment inspired by both references. High-resolution commercial quality, realistic lighting, precise geometry, premium AI-generated campaign visual.
WeShop result: Describe the Change in Material and Style
Reference 1

Describe Shape and Material Changes

Turn the reference building and seating into a souvenir magnet with a white outline, shallow relief and a soft shadow. This run used 3:4, 1K and low quality, producing 768×1024; the original prompt asks for horizontal 3:2. Choose the ratio separately in the form.

Please generate a minimalist fridge magnet photography poster based on the real photo I provided. The overall style should reference "architectural fridge magnets" or "minimalist collage". 【Layout Structure】Horizontal composition, 3:2 ratio. Minimalist solid color background. 【Fridge Magnet Icon】Extract the most recognizable main subject from the provided photo (e.g., the dining table and chairs) and convert it into a minimalist "fridge magnet-style icon". Requirements: 1. Retain the core outline and iconic features of the subject. 2. Simple, clean shape, resembling a cute travel souvenir fridge magnet. 3. Add a slight 3D pop-up effect and a subtle drop shadow. 4. The edges must be clear, with a white or light-colored sticker-like border/stroke. 5. Moderately simplify the details; do not make the drawing too complex or hyper-realistic. 6. Place the icon in the upper center part of the image. 7. The icon should be relatively small and delicate, leaving plenty of negative space around it. 【Background Color】Extract the most prominent, memorable, and saturated color from the provided photo for the background. The background must be a solid, flat color—saturated, clean, and giving a travel postcard vibe. Do NOT use gradients, and do NOT add any complex patterns or background clutter.
WeShop result: Separate the Subject from the Layout
Reference 1
Reference 2
Reference 3

Describe Subject and Background Separately

Specify the outfit details first, then the burgundy background, sparse white notes and arrows. This run used 3:4, 1K and low quality and produced 768×1024, rather than the approximately 2:3 requested in its prompt.

Use my uploaded image as the main person or product asset to create a vertical, vintage fashion collage poster. Carefully cut out the subject and place it in the center of the composition. Preserve its recognizable clothing, pose, silhouette, proportions, colors, and important details without changing its identity. Present the subject as a clean magazine-style cutout layered directly onto the background. If the uploaded image features a person, the head may be cropped out or extend beyond the top edge so the visual focus remains on the outfit and body shape. If the subject is a product, keep it complete, clear, and easily recognizable. Use a solid deep burgundy, dark brick-red, or muted oxidized-red background that fills the entire canvas. Do not add a white border, white margins, a paper frame, or empty white space around the poster. Add only a few white handwritten English words, curved arrows, and casual fashion-sketch doodles on the left and right sides of the subject. The arrows may point toward clothing pieces, accessories, or other important details. The writing should look naturally hand-drawn, slightly irregular, and loosely arranged, like annotations in a fashion designer’s sketchbook. Keep these elements sparse and make sure they do not cover the subject. The overall visual language should resemble an independent fashion magazine, a clothing-design scrapbook, or an experimental late-20th-century fashion advertisement. Keep the layout minimal and clear, with one central cutout subject against the red background. Use an approximately 2:3 vertical aspect ratio. The subject should occupy around 75%–85% of the poster’s height and be positioned centrally, slightly toward the bottom. Keep the final image sharp and clearly defined. Do not add torn-paper edges, rough cutout edges, glue marks, tape, folds, stains, fading, film grain, scanning noise, blur, distressed textures, decorative badges, realistic scenery, complex props, modern interface elements, neon lighting, 3D rendering, or overly vivid colors.
WeShop result: Describe Materials through Lighting

Describe How Materials Catch Light

Go beyond “premium product photo.” This prompt names brushed metal, fabric cushions and a clear glass block, then specifies a broad softbox reflection on the metal. Each surface has a concrete visual requirement.

A photorealistic commercial product photograph of one unbranded pair of brushed-silver wireless over-ear headphones. The headphones rest on a clear rectangular glass block against a seamless pale-cyan background. Show fine metal brushing, soft charcoal fabric ear cushions, a realistic transparent block, a clean contact shadow, and a broad softbox reflection across the metal. Use an eye-level three-quarter camera angle and a restrained monochrome palette. One pair of headphones only, no people, letters, logos, packaging, border, or watermark. Vertical 3:4 composition with the whole product visible.
WeShop result: Organize Foreground, Background, and Light

Position the Subject, Background and Lights

Place the globe in front, lightning behind to the left, and the candle on the right. This run used 4:3, 2K and medium quality: 2048×1536 original, 1280×960 preview. Its original prompt asks for 3:4.

Create a finished 3:4 photorealistic editorial image of exactly one clear glass snow globe resting on a dark walnut desk during a thunderstorm. Inside the globe is one miniature Victorian greenhouse containing exactly three red orchids and one small brass telescope. Place a sharp lightning bolt in the upper-left background and one lit ivory candle on the desk to the right of the globe. Use realistic glass reflections and refractions, cool blue storm light from the left, warm candlelight from the right, shallow depth of field, and detailed natural materials. Keep the entire globe visible. No people, no extra flowers, no extra globes, no text, no logo, no watermark.
WeShop result: Specify Panel Order and Exact Labels

Set Panel Order and Exact Copy

For a multi-panel design, specify each panel’s content, position and words. This prompt orders GRIND, BREW and POUR and assigns an illustration and number to each.

Create a clean flat-vector educational infographic in a vertical 3:4 layout with exactly three equal horizontal panels stacked from top to bottom. Panel 1 shows a simple hand grinder and coffee beans, with the exact heading 'GRIND'. Panel 2 shows a pour-over cone over a glass server and a kettle pouring water, with the exact heading 'BREW'. Panel 3 shows a single ceramic cup filled with coffee, with the exact heading 'POUR'. Place the number 1, 2, or 3 beside each corresponding heading. Use teal, coral, cream, and dark-navy colors, consistent line weights, generous spacing, and a left-to-right illustration-and-label arrangement within every panel. Keep the three panels distinct and correctly ordered. No extra text, border, photorealistic imagery, or watermark.
HOW TO USE

How to Use GPT Image 2 on WeShop AI

Describe the image you want, add reference images if needed, and choose your generation settings in the form above.

  1. Describe the Image You Want

    Specify the subject, setting, style and composition. List any required text exactly, line by line. To edit an existing image, upload it and explain what to keep and what to change.

  2. Choose Your Output Settings

    Select an aspect ratio, size and quality for your intended use. Check the image quantity and points required, then click Generate.

  3. Review and Refine the Result

    Open the result and check the text, people or product details against your brief. To refine it, adjust the relevant part of your prompt and generate another version.

GPT Image 2 Tutorials

Follow complete workflows for structured prompts, portrait posters, hand-drawn photo annotations and product-on-model images.

COMPARE GPT IMAGE MODELS

GPT Image 2 vs Flare vs Sunburst

All three models accept text and image inputs. OpenAI positions Flare for fast everyday generation and Sunburst for precise editing. GPT Image 2 supports the text-led designs and reference edits shown here. Compare their current WeShop options below.

DimensionGPT Image 2GPT Image 2.5 FlareGPT Image 2.5 Sunburst
Official positioningImage generation and reference-based editingFast everyday image generationPrecise editing and demanding visual briefs
Inputs in WeShopText; optional reference imagesText; optional reference imagesText; optional reference images
Quality options in WeShopLow, Medium, HighLow, Medium, High, XHigh, MaxLow, Medium, High, XHigh, Max
Size labels in WeShop1K, 2K, 4K1K, 2K, 4K1K, 2K, 4K
Aspect-ratio choices in WeShopAuto and 14 ratios, including 3:4 and 4:3The same choices, plus 4:5 and 5:4The same choices, plus 4:5 and 5:4
Official positioning
GPT Image 2
Image generation and reference-based editing
GPT Image 2.5 Flare
Fast everyday image generation
GPT Image 2.5 Sunburst
Precise editing and demanding visual briefs
Inputs in WeShop
GPT Image 2
Text; optional reference images
GPT Image 2.5 Flare
Text; optional reference images
GPT Image 2.5 Sunburst
Text; optional reference images
Quality options in WeShop
GPT Image 2
Low, Medium, High
GPT Image 2.5 Flare
Low, Medium, High, XHigh, Max
GPT Image 2.5 Sunburst
Low, Medium, High, XHigh, Max
Size labels in WeShop
GPT Image 2
1K, 2K, 4K
GPT Image 2.5 Flare
1K, 2K, 4K
GPT Image 2.5 Sunburst
1K, 2K, 4K
Aspect-ratio choices in WeShop
GPT Image 2
Auto and 14 ratios, including 3:4 and 4:3
GPT Image 2.5 Flare
The same choices, plus 4:5 and 5:4
GPT Image 2.5 Sunburst
The same choices, plus 4:5 and 5:4

Choose GPT Image 2 When

Choose GPT Image 2 when your image needs text, subjects and a planned layout, or when you want to edit an existing reference. It supports tasks such as event posters, product visuals, portrait edits and style transformations.

Choose GPT Image 2 When

Choose GPT Image 2.5 Flare When

Choose Flare when speed is a priority for everyday image creation, whether starting from text or a reference image. OpenAI positions it for fast, high-quality everyday generation.

Choose GPT Image 2.5 Flare When

Choose GPT Image 2.5 Sunburst When

Consider Sunburst when your editing task calls for precise changes and close attention to detail. OpenAI positions it for workflows where editing precision matters most.

Choose GPT Image 2.5 Sunburst When
MORE IMAGE MODELS

Explore More Reference-Editing Models

Explore Nano Banana and Qwen Image Edit for other reference-image editing approaches and examples.

FAQ

GPT Image 2 FAQ

What is GPT Image 2?

GPT Image 2 is OpenAI’s image-generation and editing model. It creates images from text and works with uploaded images for edits and style transformations.

What can I make with GPT Image 2?

Use it for text-led posters, infographics, product visuals, portrait edits and style transformations. Specify the copy and layout for graphic work, or provide a reference image and describe what should stay and what should change.

Do I have to upload an image?

No. You can create a new image from text alone. Upload references when you want to use an existing person, product or scene. With multiple images, explain the visual role of each one.

Which sizes, quality levels and aspect ratios are available?

WeShop offers 1K, 2K and 4K size labels; Low, Medium and High quality; and Auto plus 14 fixed aspect ratios, including 1:1, 3:4, 4:3, 16:9 and 9:16. Select the ratio in the form. Check the downloaded file for its actual pixel dimensions rather than relying on the prompt or size label.

What should I include in a GPT Image 2 prompt?

Start with the subject and intended image, then add the setting, composition, style, lighting and materials. List exact text and its position when needed. For reference edits, identify each image’s role and what to retain or change. Replace example subjects, products and copy with your own brief.

How do I choose between GPT Image 2, Flare and Sunburst?

GPT Image 2 supports text-led design, image generation and reference editing. Consider Flare when everyday generation speed is a priority, or Sunburst when precise editing matters most. All three accept text and images; see the comparison table for their quality and aspect-ratio options in WeShop.

How do I check access and points?

Check whether your WeShop plan includes this model and review the points estimate before generating. Quality, size and image quantity can affect the points required.

What details should I check in the result?

Check spelling and text placement, faces and hands, product shapes, and any details you asked to retain. Reference edits may differ from the original. Before printing or using a product image, check the file dimensions and compare the result with your source material or real item.

Can I use an example prompt directly?

Use it as a starting point, replacing its subject, copy and scene with your own. Reference-based examples also need your own uploaded images; copying a prompt does not attach the example’s files. Confirm the form’s aspect ratio, size and quality before generating.