
Retro-Futurist World
A panoramic environment study with architecture, landscape, water, and small human figures.
Choose Flare for everyday image creation or Sunburst when editing precision matters most. Both support text prompts and reference images, with output up to 4K. Explore examples and start with 10 free credits.



Result

Cinematic City
Create a cinematic wide photograph of a woman seen from behind, sitting on a cracked rooftop ledge overlooking a fictional city at night. Distant skyscrapers glow with orange flames beneath towering smoke clouds. Wet concrete reflects the firelight. Center the figure against a deep teal skyline, with shallow depth of field, subtle film grain, dramatic atmospheric lighting, and no text.
Select an example to fill the prompt. Generation starts only when you choose Generate.
Explore nine visual directions, from panoramic scenes and painted cityscapes to poster systems, invitations, diagrams, and mosaics.

A panoramic environment study with architecture, landscape, water, and small human figures.

Nine coordinated posters combining typography, illustration, and a consistent visual language.

A vintage poster layout combining bold display type, a black cat, illustrated stickers, and bilingual details.

An impressionist street scene with layered depth, foliage, buildings, water, and a distant bridge.

An ornate invitation composition combining a monogram, decorative borders, illustration, and styled props.

A cinematic wide scene built around inverted scale, deep blue atmosphere, and a floating figure.

Eight related stamp designs with distinct landmarks, lettering, color palettes, and vintage print texture.

A presentation-ready diagram that organizes a scientific process into a clear four-part sequence.

A wide decorative composition translating Earth, planets, stars, and a spiral galaxy into tile art.
Choose the model, add up to 16 references, set fifteen aspect ratios and five quality tiers, then generate at 1K, 2K, or 4K.
Describe the subject, setting, composition, and wording. GPT Image 2.5 creates a new image from that prompt.
Switch to image-edit mode when a visual should guide the result. State what changes and what remains.
Upload up to 16 references. They can show object, character, color, or composition cues.
Choose from fifteen ratios, including square, landscape, portrait, panoramic, and extra-tall canvases. Set the canvas before generating.
Choose Low, Medium, High, XHigh, or Max. Each step adds detail, generation time, and credit cost.
Choose 1K for drafts, 2K for detailed review, or 4K for the largest output. Credits follow model, resolution, and quality.
Review the completed image in the preview panel. Download it when the composition and details meet your needs.
Use the GPT Image 2.5 AI Image Generator to compare an input with a targeted edit. Drag the handle to inspect the change.
Upload a reference image, describe one required change, and list the details to preserve. Review the subject, pose, background, and composition before downloading the result.
Open Image Edit Mode
GPT Image 2.5
InputSelect Flare or Sunburst, use up to 16 references, choose five quality tiers, and generate at 1K, 2K, or 4K.
Choose GPT Image 2.5 Flare or Sunburst before starting a text-to-image or reference-editing task.
Add visual sources for the subject, object, pose, color, background, or composition in Image Edit mode.
Select Low, Medium, High, XHigh, or Max according to the detail level and credit cost needed for the image.
Choose the output resolution before generation and review the calculated credits for the selected combination.
Make posters, product images, infographics, interface concepts, social images, and character references.
Combine a subject, hierarchy, and requested wording for event, cover, or display posters.
Place a product in a scene, change the background, or test packaging while references guide object details.
Build an illustrated explainer with labels and callouts. Verify facts and numbers before publishing.
Explore a screen or dashboard as a visual reference, not finished code.
Choose a feed, story, or landscape ratio and describe space around the subject.
Create pose, clothing, or scene references. Upload visuals to guide specific features.
Create drafts, scenes, and references.
Designers can test composition, color, and format before refinement. References keep variations tied to the brief.
Content creators can prepare covers, explainers, thumbnails, and character references, then review them before download.
Store owners can explore backgrounds, packaging, seasonal scenes, and compositions. They should verify labels and details.
Small teams can use one prompt and shared references as a starting point for later production.
Compare GPT Image 2.5, Seedream 5.0 Pro and Lite, and GPT Image 2 by editing tools, reference limits, resolutions, and output controls.
| Comparison point | GPT Image 2.5 | Seedream 5.0 | GPT Image 2 |
|---|---|---|---|
| Primary purpose | Still-image creation and reference editing. | Pro: design and interactive editing. Lite: creation with visual reasoning. | General creation and editing. |
| Image creation | Written prompts with Flare or Sunburst. | Text prompts, reference blending, and related image sets. | Text instructions. |
| Existing image editing | Image Edit mode with text instructions. | Pro: annotation-guided edits and layer separation. Lite: local edits and style transfer. | Image input and output. |
| Reference images | Up to 16. | Pro: up to 10. Lite: up to 14. | Supported; limit not stated. |
| Output controls | Model, 15 ratios, five quality tiers, resolution, PNG, JPEG, WebP. | Resolution, custom dimensions, image count, PNG, JPEG. | Flexible sizes and quality. |
| Resolution | 1K, 2K, or 4K. | Pro: 1K or 2K. Lite: 2K, 3K, or 4K. | Flexible sizes; no single maximum stated. |
| Aspect ratios | 15 presets, including 1:1, 16:9, 21:9, and 9:21. | Custom width-to-height ratios from 1:16 to 16:1, within pixel limits. | Flexible sizes; full ratio list not stated. |
| Suitable outputs | Posters, products, infographics, interfaces, social images, characters. | Infographics, multilingual layouts, portraits, storyboards, and coordinated image sets. | General images and edits. |
| Typical users | Designers, creators, stores, teams. | Designers, illustrators, educators, and visual production teams. | General image creators and editors. |
| Payment model | One-time credits and a $0 starter option. | Usage-based image billing; consumer plans depend on the application. | Depends on the product used. |
| Best fit | 16-reference edits with selectable quality and one-time credits. | Pro: annotated design revisions. Lite: search-informed images and related variations. | Official GPT Image 2 access. |
Start with your task: everyday image creation or precision-focused editing. Both models support creation and editing; compare their credit costs before generating.
Choose Flare for everyday image creation, from product scenes to social images. It also supports reference image editing, with five quality tiers and 1K, 2K, or 4K output. At 2K Low and Medium, it uses fewer credits than Sunburst.
1K Low or Medium: 1 credit
2K Low: 1 credit
2K Medium: 2 credits
Choose Sunburst when editing precision is the priority, such as changing an object while preserving the surrounding scene. It also creates images from text prompts, with the same five quality tiers and 1K, 2K, or 4K output. Review the result against your references before downloading.
1K Low or Medium: 1 credit
2K Low: 2 credits
2K Medium: 3 credits
Choose a model, describe or upload up to 16 references, set ratio, quality, resolution, and format, then generate and download.
Choose Flare or Sunburst, describe the subject or requested edit, and upload up to 16 references when needed.
Select one of fifteen ratios, one of five quality tiers, 1K, 2K, or 4K output, and PNG, JPEG, or WebP. Review the calculated credit cost.
Generate when enough credits are available. Check the preview, then download the finished file.
Start with 10 credits for $0 or choose a larger one-time pack. The generator calculates cost from the selected model, quality, and resolution.
Answers about credits, references, editing, output settings, pricing, and comparisons.
GPT Image 2.5 has a $0 Free Starter option with 10 credits. You must sign in before the generator starts a task, and the required credits depend on the selected model, resolution, and quality. A 1K Low image costs one credit with either Flare or Sunburst, while higher quality and larger output use more. The free allocation is useful for testing the prompt, model, aspect ratio, preview, and download flow. If you need more generations, the page lists one-time packs with 150, 540, or 1,100 credits. Every pricing card is presented as a one-time choice.
Yes, GPT Image 2.5 includes an image-edit mode for work based on uploaded visuals. Switch from text-to-image mode, add the reference material, and write a prompt that explains both the requested change and the details that should remain. This works for background changes, visual variations, composition studies, product-scene ideas, and character references. The result is a newly generated image, so inspect labels, proportions, and small details before using it. You can review the output in the preview panel and download it when the edit matches the direction you want.
You can upload up to 16 reference images in the current GPT Image 2.5 workspace. A reference can communicate an object, subject, color direction, clothing detail, pose, background, or composition that would be difficult to explain with text alone. More references do not guarantee a better result, so include only images that have a clear role in the request. Use the prompt to identify what should be taken from each reference and what may change. After generation, compare the preview with the important source details before downloading the final file.
GPT Image 2.5 supports 1K, 2K, and 4K output with Low, Medium, High, XHigh, and Max quality. It provides fifteen aspect ratios: 1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9, 2:1, 1:2, 3:1, 1:3, and 9:21, plus PNG, JPEG, and WebP output. Choose the ratio before generating because the canvas affects composition. Use lower settings for early drafts and higher settings for closer detail review. The workspace calculates credits from the selected model, resolution, and quality.
Choose Flare for everyday image creation or Sunburst when editing precision matters most. These are starting points, not restrictions: both models support text-to-image creation and reference image editing, with five quality tiers and 1K, 2K, or 4K output. Their listed credit costs differ at 2K Low and Medium: Flare uses 1 and 2 credits, while Sunburst uses 2 and 3 credits. Compare results using the same prompt and references to decide which suits your task. The Generate button shows the selected cost before you start.
Choose Flare or Sunburst, start with a prompt or up to 16 references, select the output settings, and review the result before downloading it.