curl-X POST 'https://pollo.ai/api/platform/v1/generation/openai/gpt-image-2-5/image'\-H'Content-Type: application/json'\-H'x-api-key: YOUR_API_KEY'\-d'{ "input": { "prompt": "A cinematic shot of a golden retriever running through a field of sunflowers at sunset, warm rim light, shallow depth of field", "mode": "flare", "aspectRatio": "auto", "resolution": "1K", "quality": "low", "background": "auto" }}'
선택적으로, 작업 성공 또는 실패 시 콜백을 수신하기 위해 webhookUrl을 포함할 수 있습니다.
선택적 요청 필드
{"input":{"prompt":"A cinematic shot of a golden retriever running through a field of sunflowers at sunset, warm rim light, shallow depth of field","mode":"flare","aspectRatio":"auto","resolution":"1K","quality":"low","background":"auto"},"webhookUrl":"https://example.com/webhooks/pollo"}
개요
선택한 엔드포인트 및 모드에 대한 입력 필드입니다.
필드
유형
필수의
화면 설명
prompt
string
예
Text prompt describing the content to generate.
mode
string
아니요
Generation mode of the model (e.g. a faster or higher-quality tier).
aspectRatio
string
아니요
Output aspect ratio, as width:height (e.g. 16:9).
resolution
string
아니요
Output resolution of the generated media (e.g. 720p, 1080p, 2K, 4K). Values are case-insensitive.
quality
string
아니요
Output quality tier of the generation.
background
string
아니요
Background of the generated image: auto lets the model decide, transparent produces an alpha background (PNG), opaque forces a solid background.
Change the part that needs work while keeping the subject recognizable. GPT Image 2.5 combines image generation with editing guided by detailed natural-language instructions.
Run GPT Image 2.5 on Pollo API to add visual creation and revision to design tools, asset workflows, or applications that turn customer briefs into images.
Key Features of GPT Image 2.5 API
Targeted Image Revision
Natural-language edits let you describe the intended change and identify what should remain. This supports focused revisions rather than requiring a complete visual restart.
Subject Preservation
The model family improves subject preservation during editing. Reference-led workflows can retain recognizable people or objects while exploring a different treatment for the surrounding image.
Detailed Brief Interpretation
Instructions can combine subject placement, visual treatment, and required content. Give the image a defined purpose, then describe the relationships between its most important elements.
Integrated Typography
Text can be generated within the image composition, making headlines and labels part of the visual brief. Specify exact wording and review the final spelling.
Transparent Asset Creation
Transparent backgrounds support isolated visual assets for later composition. Define the subject’s silhouette clearly and inspect fine edges before placing the result into another design.
Use Cases of GPT Image 2.5 API
Focused Asset Revisions: Change a selected color, material, or background while explicitly protecting the approved subject and composition.
Catalog Scene Variants: Place a referenced object in a new visual setting, checking its geometry and identifying details against the original.
Transparent App Illustrations: Create isolated illustrated objects that can be placed over different interface backgrounds.
Campaign Copy Layouts: Combine a supplied headline with a defined visual hierarchy, then inspect the wording before export.
Reference-Based Character Art: Explore new situations for a character while retaining the key appearance details from a reference image.
Design Review Iterations: Turn specific feedback, such as “move the focal point left” or “soften the background,” into a new visual candidate.
How to Use GPT Image 2.5 API
Get an API Key: Create a Pollo API account and generate your API key from the developer dashboard.
Choose the Model: Send requests to the GPT Image 2.5 endpoint at /generation/openai/gpt-image-2-5/image.
Add Your Inputs: Supply a prompt and any supported source images. Choose qualityand output settings like background; use transparent when an isolated asset is required.
Generate and Retrieve: Submit the task, poll the returned task status, and download the finished content once it succeeds.
Prompting Best Practices for GPT Image 2.5 API
Separate the visual goal from the revision constraints. For an edit, identify the exact change and the elements whose appearance, placement, or wording must remain.
Protected Details: Identify the particular silhouette, expression, label, or arrangement that makes the source asset recognizable.
Limited Revision Scope: Resolve one important design issue per edit so you can judge the improvement without unrelated changes.
Text Placement: Provide both the required words and their position; ask for a clear hierarchy rather than simply requesting attractive typography.
Asset Boundaries: For a cutout, describe a complete, unobstructed subject with space around the silhouette.
Variant Selection: Flare prioritizes speed; Sunburst prioritizes demanding quality requirements. Compare available options on your own workload rather than assuming one always wins.
Example Prompts
A Controlled Material Change
“Edit the supplied desk organizer image. Change only its outer shell from glossy red plastic to matte sage-green ceramic. Keep the compartments, proportions, camera angle, desk surface, and object placement unchanged. Preserve the existing side lighting, with subtle ceramic texture and no new text or decoration.”
An Isolated Learning Asset
“Create a friendly illustrated yellow submarine for a children’s vocabulary app, shown in three-quarter view. Use rounded forms, two blue windows, and a small orange propeller. Center the complete object with generous space around it. Fully transparent background, crisp edges, no sea, bubbles, letters, or shadows.”
A Community Repair Poster
“Design a flat digital poster for a community repair event. Show a neatly arranged lamp, bicycle wheel, and stitched backpack around the central headline ‘FIX IT TOGETHER’. Place ‘Saturday, 10 AM–2 PM’ below it. Use warm cream, cobalt, and orange, generous margins, and no additional wording.”
GPT Image 2.5 vs Qwen Image 3.0 Pro vs Qwen Image 3.0
Workflow Factor
GPT Image 2.5
Qwen Image 3.0 Pro
Qwen Image 3.0
Generation and editing
Both available
Both available
Both available
Documented emphasis
Instruction following and image revision
Refined details and complex compositions
Text rendering and structured compositions
Distinctive evaluation
Preservation across a sequence of edits
Fine details in a dense design
Clarity of an everyday information layout
Suggested first brief
A narrowly defined source-image edit
A detailed editorial page
A labeled visual guide
Why Choose GPT Image 2.5 API?
GPT Image 2.5 is worth testing when useful output depends on both creating the right image and revising it without losing the subject’s defining details.
Pollo API lets your application combine image generation and editing with other creative models through a common platform, supported by documentation for integration and task retrieval.
Try GPT Image 2.5 on Pollo API with an existing asset and one specific revision to evaluate how well the result preserves your original intent.
GPT Image 2.5 API FAQs
What is GPT Image 2.5?
GPT Image 2.5 is OpenAI’s image generation model family. It turns natural-language prompts into images, with strengths in detailed instruction following, text rendering, and visual composition.
How do I generate images with GPT Image 2.5 API?
Create a Pollo API key, select GPT Image 2.5, and submit your prompt with the desired output settings. Retrieve the generated image once the task completes.
Can GPT Image 2.5 API generate text inside images?
Yes. GPT Image 2.5 can generate headlines, labels, and other text within images. Include the exact wording and placement in your prompt, then review the output for accuracy.
What types of images can I generate with GPT Image 2.5 API?
You can generate product visuals, posters, illustrations, interface concepts, and other prompt-based images. Specify the intended format and visual requirements to guide the composition toward your use case.
Why run GPT Image 2.5 through Pollo API?
Pollo API lets you access GPT Image 2.5 and over 300 other models through just one integration, backed by full documentation, task tracking, and webhooks — all at a lower cost than what comparable providers charge.
