Skip to main content
OpenAI GPT Image 2.5 (gpt-image-2.5) is OpenAI’s newest image model family, available in ComfyUI through Partner Nodes on the OpenAI GPT Image 2.5 node. It comes in two variants tuned for different stages of the creative process:
  • GPT Image 2.5 Flare (gpt-image-2.5-flare): fast image generation and editing optimized for rapid prototyping and creative iteration. It delivers quick previews for concept exploration, mood boards, poster drafts, character sketches, and scene mockups, balancing solid visual quality with low latency for high-volume trial-and-error workflows.
  • GPT Image 2.5 Sunburst (gpt-image-2.5-sunburst): high-fidelity image generation and editing for polished, detail-rich visuals. It preserves subject identity, complex materials, lighting consistency, and scene structure across iterations, making it suitable for concept refinement, product visuals, character scenes, cinematic backgrounds, and finished artwork.
Both variants support text-to-image and image editing with reference images (up to 16), an optional mask for inpainting, and five quality levels: low, medium, high, xhigh, and max. The template library currently ships with four example workflows:
  • Flare: Text to Image. Fast poster generation from a text prompt
  • Flare: Image Edit. Outfit and hairstyle swap while keeping the scene fixed
  • Sunburst: Text to Image. High-fidelity editorial fashion poster with rendered typography
  • Sunburst: Image Edit. High-fidelity hairstyle and makeup edit with subject consistency
To use the Partner Nodes, you need to ensure that you are logged in properly and using a permitted network environment. Please refer to the Partner Nodes Overview section of the documentation to understand the specific requirements for using the Partner Nodes.
Make sure your ComfyUI is updated.Workflows in this guide can be found in the Workflow Templates. If you can’t find them in the template, your ComfyUI may be outdated.If nodes are missing when loading a workflow, possible reasons:
  1. You are not using the latest ComfyUI version (Nightly version)
  2. Some nodes failed to import at startup

​
Getting started

  1. Update ComfyUI to the latest version. The node and templates require ComfyUI 0.35.0 or later
  2. In the Node Library, search for OpenAI GPT Image 2.5 and add the node
  3. Set the model field to gpt-image-2.5-flare or gpt-image-2.5-sunburst

​
Node settings

The OpenAI GPT Image 2.5 node calls OpenAI’s image API synchronously and returns the generated image. The options change with the selected model:
  • model: gpt-image-2.5-flare (fast iteration) or gpt-image-2.5-sunburst (high fidelity)
  • prompt: Text description of the image to generate, or the edit instruction when reference images are connected
  • size: auto, a preset such as 1024x1024 or 3840x2160, or Custom
  • custom_width / custom_height: Used only when size is Custom. From 480 to 3840, in steps of 16
  • background: auto, opaque, or transparent (both 2.5 variants support transparent backgrounds)
  • quality: low, medium, high, xhigh, or max. Affects cost and generation time
  • images: Optional reference image(s) for image editing, up to 16
  • mask: Optional mask for inpainting (white areas are replaced). Requires exactly one reference image
  • n: How many images to generate (1 to 8)
  • seed: Kept for API compatibility; the backend does not use it yet

​
GPT Image 2.5 Flare: Text to Image

Generate a poster from a text prompt with Flare’s fast generation. The template prompt produces a 1:1 square racing-car poster with the text “GPT IMAGE 2.5” projected onto the car body: a good example of asking for rendered text, which the GPT Image family handles reliably.

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “GPT Image 2.5 Flare” in Template Library
GPT Image 2.5 Flare Text to Image workflow preview No API node workflows need model downloads: the generation runs on OpenAI’s infrastructure. Connect the node’s image output to a Save Image node and run the workflow. Tips:
  1. Iterate with low quality first: The template ships with quality set to low, which is cheap and fast. Raise it only for the final run
  2. Ask for text explicitly: Quote the exact text in your prompt (for example “the words GPT IMAGE 2.5”) and state where it should appear
  3. Exclude unwanted elements: The template prompt ends with a list of exclusions (“No sidebar, no extra UI panels…”), which helps keep posters clean
  4. Use Custom sizes: Pick Custom under size to set an exact width and height in 16px steps, up to 3840

​
GPT Image 2.5 Flare: Image Edit

Edit an input image with Flare. The template demonstrates an outfit and hairstyle swap, but the workflow is general-purpose image editing: character appearance changes (outfit, hairstyle, accessories, expression), object replacement, background swaps, color adjustments, and any local modification where you want a fast preview. Good for iterating on edits before committing to a final high-quality run.

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “GPT Image 2.5 Flare” in Template Library
GPT Image 2.5 Flare Image Edit workflow preview Upload this file to the Load Image node:

Input Image: outfit_swap_base_gpt_image25.png

The example input image, or use your own.
Tips:
  1. State what to keep: The template prompt opens with “Keep the background environment… completely unchanged” before describing the edit. This anchoring is what keeps the rest of the frame stable, and it works for any edit type: describe the scene, pose, and lighting you want preserved, then the change you want made
  2. Swap any appearance attribute: The same prompt structure covers outfit changes, hairstyle changes, accessories, color schemes, or a full character redesign while keeping pose and scene
  3. One change per run: Describe a single edit (or a tightly related set) rather than a list of unrelated changes. Run several iterations instead
  4. Mask for local edits: Connect a mask to limit the edit to specific regions; white areas are replaced. A mask requires exactly one reference image
  5. Iterate in Flare, finish in Sunburst: Use Flare to explore edit options cheaply, then re-run the winning prompt with gpt-image-2.5-sunburst for the final render

​
GPT Image 2.5 Sunburst: Text to Image

Generate a high-fidelity editorial poster with Sunburst. The template prompt is a long, structured editorial prompt: scene, subject, oversized typography, scattered graphic elements, caption block, and a full palette/render specification. Use it as a scaffold for your own poster prompts.

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “GPT Image 2.5 Sunburst” in Template Library
GPT Image 2.5 Sunburst Text to Image workflow preview Tips:
  1. Structure long prompts by layer: The template prompt describes one layer at a time (background scene, subject, typography, small graphic elements, caption, then global style). This keeps every element where you asked for it
  2. Pin text placement: Say what the text is, its font style, its color, and what it should overlap (“letters weaving through the upper half of the frame, some behind the biker, some in front”)
  3. Declare the palette: Listing the exact palette (“burnt orange, black, desert beige and electric blue”) plus the render style keeps the whole image coherent
  4. Raise quality for finals: Sunburst at high or above is where its fidelity advantage shows; keep exploration on low

​
GPT Image 2.5 Sunburst: Image Edit

Edit an input image with Sunburst’s high fidelity. The template demonstrates a hairstyle and makeup change, but the workflow is general-purpose image editing: character appearance replacement (outfit, hairstyle, accessories), object insertion or removal, style conversion, and precise partial modifications. Use it for any edit where identity preservation and detail fidelity matter.

Run on Comfy Cloud

Open in Comfy Cloud

Download Workflow

Download JSON or search “GPT Image 2.5 Sunburst” in Template Library
GPT Image 2.5 Sunburst Image Edit workflow preview Upload this file to the Load Image node:

Input Image: hairstyle_makeup_base_gpt_image25.png

The example input image, or use your own.
Tips:
  1. Keep-list before change-list: The template prompt names everything to preserve first (“Keep head, face, body, flowers, sweater, background, lighting unchanged”), then the single change. Follow that order for any edit: appearance swaps, object edits, background replacements, or style conversions
  2. Character appearance replacement: To give a character a new outfit or look, enumerate the identity anchors you want preserved (face, body proportions, pose, lighting), then describe the new appearance precisely
  3. Strong subject consistency: Sunburst keeps identity across sequential edits, which makes it suitable for multi-step edit chains and frame-to-frame generation with the same character
  4. Transparent backgrounds: Both variants can return transparent backgrounds; useful for compositing the result into another scene downstream