> For the complete documentation index, see [llms.txt](https://xfigura.gitbook.io/xfigura-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://xfigura.gitbook.io/xfigura-docs/models/image-models/gpt-image-2.5.md).

# GPT Image 2.5

Image generation and editing with Flare for speed and Sunburst for fine detail and precise changes.

GPT Image 2.5 is OpenAI's image generation and editing model. It comes in two variants, **Flare** and **Sunburst**, with different balances of speed and detail.

In xFigura, GPT Image 2.5 uses **Sunburst by default**. Check the model and variant shown in your workflow before running a job.

## Flare and Sunburst

| Variant      | Strengths                                                                           | Use it for                                                           | Trade-off                   |
| ------------ | ----------------------------------------------------------------------------------- | -------------------------------------------------------------------- | --------------------------- |
| **Flare**    | Fast generation, natural lighting, rich textures, and complex layouts               | Everyday creative work, concept exploration, and repeated iterations | Prioritizes turnaround time |
| **Sunburst** | Intricate detail and tightly scoped edits that preserve the subject and composition | Detail-sensitive assets, close-up views, and precise revisions       | Takes longer to generate    |

Flare's [generation endpoint](https://fal.ai/models/openai/gpt-image-2.5/flare/text-to-image) and [editing endpoint](https://fal.ai/models/openai/gpt-image-2.5/flare/edit) are designed for general-purpose work. Sunburst's [generation endpoint](https://fal.ai/models/openai/gpt-image-2.5/sunburst/text-to-image) and [editing endpoint](https://fal.ai/models/openai/gpt-image-2.5/sunburst/edit) focus on extra fidelity and precise control.

## Example uses

* Explore several compositions before choosing a direction
* Develop product, interior, and campaign images with clearly described materials and lighting
* Create posters or layouts with exact copy and a defined visual hierarchy
* Revise an existing image while preserving the elements you identify as fixed

## Inputs and outputs

|                   | Supported                                     |
| ----------------- | --------------------------------------------- |
| Inputs            | Text prompts and reference images for editing |
| Output            | Image                                         |
| xFigura workflows | Image generation and editing                  |

## xFigura controls

* **Aspect ratio:** With an image input, **Auto** follows the input dimensions. The Resolution control is hidden in this mode
* **Resolution:** Choose a specific aspect ratio to show the Resolution dropdown. **Max** depends on the selected ratio; it should not be read as the same dimensions for every format
* **Quality:** Quality is a separate control. Review it alongside resolution and the run estimate before generating

## Provider capabilities

The [fal model reference](https://fal.ai/gpt-image-2.5) documents output up to 3840 pixels on the long edge, up to 16 reference images for editing, and up to four outputs per request. It also lists six quality settings: auto, low, medium, high, xhigh, and max. Transparent output uses PNG or WebP.

Use the options offered in your xFigura workflow. Provider limits describe the underlying API and may exceed the controls available in a particular node or mode.

## Limitations

* Review generated copy, factual labels, dimensions, and technical geometry
* Inspect preserved areas after an edit, especially across several revisions
* Fine detail can take longer, and higher resolution or quality can increase processing time and credit use

## xFigura tip

Separate the brief into three parts: what to create, what must stay fixed, and what may change. For a localized revision, describe one change at a time. For example: "Change the sofa fabric to forest-green velvet. Keep the furniture positions, camera angle, artwork, and lighting unchanged."

## Credits and timing

Credit use and processing time depend on the mode, output settings, and current configuration. Review the estimate shown in xFigura before running the job.

## Continue

[Compare image models](/xfigura-docs/models/image-models.md)