GPT Image 2.5: Flare vs Sunburst, Sketch Input and Prompt Patterns
Choose Flare when speed matters and Sunburst when quality requirements dominate. Attach references with explicit roles and name what must change or stay fixed. In the API, set technical controls separately from the prose. Chat drawing and annotation features are interface tools, not portable API syntax.
Model capabilities and interface features
- Two API models. Flare (speed) and Sunburst (precision).
- Sketch reference. Attach the sketch image and describe its layout role. Drawing directly in a chat depends on the product interface.
- Templates and shared prompts. Reusable briefs for a series, shareable with a team.
- Focused edits. Name the region to change and the features to preserve. Use masks or annotations only when your interface supports them; inspect the output.
- Reference preservation. A real face or product from a photo survives edits more reliably than in GPT Image 2.
- Transparent backgrounds with explicit background and file-format controls.
Which model for which job
| Task | Model | Why |
|---|---|---|
| Drafts, thumbnails, dozens of variations | Flare | Speed-oriented starting point; measure total cost |
| Final hero image, packaging, poster | Sunburst | Precision, text fidelity, detail |
| Keep a real person's likeness from a photo | Sunburst | Quality-oriented; inspect reference fidelity |
| Layout from a rough drawing | Flare + sketch reference, then Sunburst | Use the sketch as a layout reference; inspect the result |
| Transparent-background cut-outs | Either | Native transparent output |
| Small focused edit ("only the sofa") | Flare with a focused edit instruction | Name the target region and review the result |
Settings and acceptance checks
Set model, size, quality and background in the API or provider controls. GPT Image 2.5 supports low, medium, high, xhigh, max and auto quality. For transparency set background to transparent and choose PNG or WebP. Identify each reference by role, say exactly what to change and what to preserve, and make one focused revision at a time. Check the result at full size for identity, spelling, geometry and transparent edges before accepting it.
Prompt patterns that work
- Generation: "A single photorealistic product photograph of a matte black ceramic mug on a light oak table, soft north-window light from the left, 50 mm lens at eye level, shallow depth of field, no text, no reflections of a person."
- Edit with a reference photo: "Use the attached photo for the person's face, hair and body only — do not copy its pose, background or lighting. Place her in a sunlit café, three-quarter view, natural skin texture, keep the face exactly."
- Sketch-driven layout: "Follow the attached sketch for the placement: headline top-left, product centre-right, a small badge bottom-left. Render as a clean e-commerce banner, 16:9, brand colours navy and cream."
- Focused edit: comment on the sofa and write "replace with a linen sectional in warm grey, same size and position, keep everything else".
- Text in the image: keep copy short, put it in quotes, verify at full size — inspect every word and typeset critical copy manually if necessary.
What to drop from your old prompts
Negative-prompt fields (write constraints instead), quality tag soups ("8k, masterpiece, trending on artstation"), and unclear positional paragraphs. Keep one subject, one light, one mood, and one clear instruction per edit.
Pricing and labeling
API cost depends on input and output tokens, model, quality, size and any partial images. Use OpenAI’s current image-cost calculator and include inputs and retries. A shared quality label does not imply equal token use, quality or latency across models. OpenAI API
Sources checked · 2026-09-30
FAQ
What is GPT Image 2.5?
OpenAI offers GPT Image 2.5 Flare for speed and GPT Image 2.5 Sunburst for quality. Both support image generation and editing. Inspect reference faces, product details and text rather than treating preservation as guaranteed.
Flare or Sunburst?
Choose according to the job. Compare the same prompt, references, size and explicit quality setting. Keep the model that meets your acceptance criteria, then reduce settings only if quality remains acceptable.
Can I use a sketch through the API?
Yes: attach the sketch as an image and explain that it controls layout. A label such as @Sketch in a prompt is only a label; it does not attach an image or set an API parameter. Chat-interface drawing and annotation controls depend on the product you use.
Which settings belong outside the prompt?
Set model, size, quality and background in the API or provider controls. GPT Image 2.5 supports low, medium, high, xhigh, max and auto quality. For transparency set background to transparent and choose PNG or WebP.
How is it priced?
API cost depends on input and output tokens, model, quality, size and any partial images. Use OpenAI’s current image-cost calculator and include inputs and retries. A shared quality label does not imply equal token use, quality or latency across models.
What is a reliable editing prompt?
Identify each reference by role, say exactly what to change and what to preserve, and make one focused revision at a time. Check the result at full size for identity, spelling, geometry and transparent edges before accepting it.
Want GPT Image 2.5 briefs written from a few clicks? GoldenPrompts builds literal, single-instruction photo prompts tuned for GPT Image, Nano Banana Pro and Seedream 5.0 — reference roles included. Free to start: 24 hours of everything, no card.