GPTImage 2.5
Generation and editing
IMAGE CREATION WORKSPACE
GPT Image 2.5 Generator & Editor
GPT Image 2.5 is live — released September 8, 2026. Generate or edit an image right here in your browser.
Browse creative prompts and visual examples. Create and edit with GPT Image 2.5 or GPT Image 2.
Image studio
Fast everyday image creation.
Quality and image count are handled automatically for GPT Image 2.5.
Shortcut ⌘ / Ctrl + Enter
Preview
Yuzu soda campaign
Create a square premium beverage campaign photo: a single acid-lime aluminum soda can floating diagonally, label facing camera with bold black text 'YUZU SODA'. Surround it with a sculptural arc of clear sparkling water, translucent ice and three juicy citrus slices, frozen in high-speed motion. Electric cobalt-blue studio background, crisp rim lighting, realistic condensation, metallic reflections and sharp citrus pulp. Keep the complete can centered within the middle 70% for safe cropping. Striking depth, premium ecommerce art direction. No extra text, logos, borders or watermark.
Try an example
Automatic or custom dimensions/ PNG, JPEG or WebP
1:3–3:1aspect ratio range
Preset and custom dimensions
6Image 2.5 quality options
Automatic through Maximum
32,000prompt characters
Space for a detailed creative brief
Product definition
What is GPTImageGen25?
GPTImageGen25 is an independent GPT Image 2.5 generator and editor. Create product photos, posters, portraits, and campaign visuals from a prompt or reference image. Choose GPT Image 2.5 or GPT Image 2, adapt an example to your brief, and preview and download the result.
Looking for model availability? Read our GPT Image 2.5 release status for the dated review, or consult OpenAI’s image generation documentation for official capabilities.
For a product listing, specify the package shape, label, surface, and lighting. For a campaign, define the mood, audience, and space needed for a headline. For an edit, name the change and the details that must remain consistent. Keep the rest of the brief stable while refining one instruction at a time, so you can judge what improved. This product, also called ImageGen25, is not an official OpenAI service.
02 / Image examples
From prompt to art direction.
Explore product campaigns, fashion editorials, beauty launches, technology, travel illustration, and architectural spaces.
Yuzu soda campaign
Frozen splashes · electric color · product storytelling
Chrome fashion editorial
Chrome fabric · bold portrait · vermilion backdrop
Rose serum campaign
Translucent glass · soft reflections · beauty campaign
Chrome audio launch
Liquid chrome · cyan light · futuristic product launch
Mediterranean summer
Screen-print texture · bold type · illustrated escape
Sunlit architectural retreat
Warm travertine · sculptural furniture · natural light
Workflow guides
Start with the job to be done.
Use a focused guide for editing, product photography, poster design, or a controlled Image 2 comparison.
01
Edit an existing image
Change one visible detail while preserving identity, framing, materials, and everything else that matters.
Open guide02
Create product photography
Direct the product, surface, camera, light, reflections, negative space, and exclusions as one studio brief.
Open guide03
Design text-led posters
Treat exact copy, typography, hierarchy, spacing, and color as explicit parts of the generated composition.
Open guide04
Compare Image 2 and 2.5
Check current model access and learn how to compare composition, text, materials, and editing with a consistent brief.
Open guide03 / FeaturesMore control.
More control.
Less prompt wrestling.
The interface keeps the creative brief and essential output decisions together, so iteration stays readable.
Prompt-aware composition
Direct the subject, camera, layout, light, materials, and mood in one natural-language brief.
01Natural image editing
Add a public image link and describe what should change. List the product, framing, colors, or other details that must stay consistent.
02Useful output controls
Choose automatic or custom image dimensions, six quality options for GPT Image 2.5, and transparent backgrounds.
03“Acid-lime soda can, suspended citrus and ice, a frozen arc of sparkling water, electric cobalt blue, crisp studio lighting.”
04 / Natural direction
Write a brief, not a bag of keywords.
Start with the subject and intended use, then describe the composition, light, and materials. A concrete brief gives you specific details to evaluate in the result and a clear next step when something needs changing.
- Define one clear subject and action.
- Describe the environment and composition.
- Specify light, lens, color, and material cues.
- End with text, layout, or exclusion constraints.
For the soda campaign, the important instructions are the can’s angle, readable label, frozen splash, and blue background. To create a calmer catalogue image, remove the splash and citrus, straighten the can, and request a plain backdrop. Keep the same product description while changing the scene. This makes the next result easier to compare with the first.
Open the prompt library05 / Model guideRead the model comparison
GPT Image 2.5 or GPT Image 2?
GPT Image 2.5 introduces Flare for fast generation and Sunburst for precise editing. Read the official capability comparison, including quality settings and transparent output, alongside the options currently offered in this workspace.
The studio includes GPT Image 2.5 Flare, GPT Image 2.5 Sunburst, and GPT Image 2 controls. Choose automatic, preset, or custom dimensions; six quality options for GPT Image 2.5; transparent or opaque backgrounds; PNG, JPEG, or WebP; and 1–10 images per request.
OpenAI launch announcement06 / How it works
A direct path from idea to image.
01
Describe
Write a visual brief with the subject, scene, style, composition, light, and constraints.
02
Configure
Select GPT Image 2.5 or GPT Image 2 and Create, then choose an aspect ratio, resolution, and quality level.
03
Generate
Create the result, review it at full size, and refine one instruction at a time.
07 / FAQ
Questions, answered directly.
Practical answers about getting started, reference images, choosing output settings, and refining the result.
OpenAI positions Flare for rapid iteration and Sunburst for detailed creative work that needs tighter editing control and longer generation times. OpenAI reports 50% lower latency for Flare than GPT Image 2; this is a provider claim, not a GPTImageGen25 benchmark.
OpenAI released GPT Image 2.5 on September 8, 2026. The API offers Flare for fast everyday generation and Sunburst for precise editing. ChatGPT Images 2.5 also began rolling out across ChatGPT, ChatGPT Work, and Codex.
GPT Image 2.5 is available right here in this workspace, and OpenAI is rolling it out across ChatGPT and its API. If you don't see it in your ChatGPT account yet, you can use it here right away.
Create product photos, campaign concepts, editorial portraits, illustrated posters, and interior design studies. Start from a prompt or use reference images for an edit. The examples above show different visual directions; select one in the studio to load its prompt, then replace the subject, colors, and composition with your own brief. Generated results can be previewed and downloaded.
Select Create and GPT Image 2.5, then describe the subject, scene, style, composition, lighting, and constraints in up to 32,000 characters. Choose an aspect ratio, resolution, and quality level before selecting Generate image. Sign in when prompted; the workspace saves your prompt and settings for your return. Review the result, change one instruction if needed, and download the version you want to keep.
GPT Image 2.5 and GPT Image 2 are both available for image generation and editing. GPT Image 2.5 is selected by default; choose either model in the studio. Sign-in and account credits are required.
Choose Edit and add one to ten publicly accessible image links. Describe a specific change, such as replacing the background with a blue studio, and state what must stay unchanged: the product silhouette, label, camera angle, and crop. For multiple references, identify them by their numbered order. Use clear, uncropped source images when small product details matter, and check the result before using it in a listing.
Choose 1:1 for square product tiles, 4:3 for landscape content, 3:4 for portrait layouts, 16:9 for widescreen visuals, or 9:16 for vertical creative. Choose automatic, preset, or custom pixel dimensions, with six quality options for GPT Image 2.5. Set the shape before generating so the composition has room for the intended placement. Inspect small text and edges in the downloaded file, not only the fitted preview.
Put essential requirements first. Quote short headlines or label text exactly, specify where they belong, and ask for no additional wording. Describe visible materials, reflections, colors, and camera angles instead of relying on words such as premium or beautiful. If a result misses a detail, revise that instruction while keeping the rest of the brief stable. Check spelling, logos, and product accuracy before publishing.
No. GPTImageGen25 is an independent product and is not affiliated with, endorsed by, or operated by OpenAI.
IMAGE CREATION STUDIO