Use GPT Image 2.5 Sunburst for prompt-based creation or reference-guided image editing.
Gallery images illustrate prompt ideas. Generate your own result to evaluate this model.
Resolution pricing
Playground Examples
A refined editorial still-life photograph of a sculptural pale yellow ceramic teapot with two matching cups on a sage green stone table, sunlit apricot curtain, lemons and a linen napkin, warm soft light, realistic ceramic glaze, rich tactile detail, understated premium campaign image. Landscape 16:9. No text, logos or watermark.
A photorealistic editorial photograph of a tranquil sunken courtyard with curved peach plaster walls, pale limestone steps, a shallow reflecting pool, one sculptural desert tree and soft golden sunlight, carefully balanced spatial composition, exquisite material details, contemporary desert architecture. Landscape 16:9. No people, no text, logos or watermark.
A luminous hand-painted illustration of a traveling greenhouse boat drifting on an emerald mountain lake at sunrise, small wooden cabin, abundant flowering plants, golden mist, delicate reflections, sophisticated storybook gouache textures in peach, jade and butter yellow. Landscape 16:9. No text, logos or watermark.
Similar Models
High-fidelity image generation and reference-guided editing with photorealistic detail, typography, and flexible framing.
ByteDance image generation and editing for cinematic visuals, product shots, precise local edits, and reference-guided creative work.
ByteDance image generation and editing for reference consistency, multi-image composition, typography, and polished visual creatives.
Model overview
Create one image from a written brief, or combine your instructions with source images. GPT Image 2.5 Sunburst offers text-to-image and image-edit workflows on the same page, with 1K, 2K, 4K resolution tiers.
Describe the subject, material, background, lighting, and camera angle. Product concepts, still lifes, and campaign imagery are useful starting points for exploring your visual direction.
Choose 1:1, 3:4, 9:16, 4:3, 16:9 for both generation and editing. State the viewpoint, subject placement, and background relationships in your prompt; the resolution tier is selected separately.
The image-edit workflow accepts 1–16 reference images. Explain the role of each reference and specify which subject details, colors, or layout elements should remain.
Model characteristics
Use Text to Image when starting from a prompt. Select Image Edit when source or reference images should guide the result.
List the desired changes and the elements to preserve. Give each reference a clear purpose instead of relying on the upload alone.
Each task requests one image. Choose the aspect ratio and one of the 1K, 2K, 4K resolution tiers; review the generated result before use.
Submit a task with your API key, retain data.task_id, and poll the shared task endpoint or supply a top-level webhook for completion.
FAQ
Text to Image needs a nonblank prompt. Image Edit also requires 1–16 reference images.
Upload JPG, JPEG, PNG, or WebP images. API requests use public HTTP(S) URLs.
Frame options are 1:1, 3:4, 9:16, 4:3, 16:9. Resolution tiers are 1K, 2K, 4K. Defaults are 1:1 and 1K.
Poll the task endpoint until it succeeds and read data.output_urls, or use the top-level webhook callback.
