Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsUse OpenAI’s Images API when your application needs a direct image-generation or image-edit request: send a prompt, select gpt-image-1, and, for edits, include the source image. The API returns image data encoded in base64, which your application must decode and save. For a conversational workflow with iterative or multi-step image work, consider the Responses API image-generation tool instead. OpenAI’s current guide foregrounds newer GPT Image model variants, so confirm that gpt-image-1 and each option in your request remain available before adapting an older example.
Choose the API surface for your workflow
The two API routes serve different application patterns; their request formats and controls are not interchangeable.
| API surface | Best fit | What to check |
|---|---|---|
| Images API | A direct generation or edit request when your application already has the prompt and any image inputs. | Use the generation or edit operation appropriate to the task, and verify which parameters are supported by gpt-image-1. OpenAI’s image-generation guide |
| Responses API image-generation tool | Image creation within a conversation, iterative editing, or a larger multi-step interaction. | Follow the Responses API tool workflow and check its model-specific controls. OpenAI’s image-generation guide |
For a single request from an application that already knows what it wants, the Images API is the more direct route. If the image request is one stage in an exchange where later steps depend on earlier results, the Responses API tool is designed for that conversational pattern. OpenAI documents both approaches; choose based on how your app manages the interaction, not on an assumption that one request can be copied unchanged into the other.
Generate an image with gpt-image-1
Send a generation request to the Images API with gpt-image-1 explicitly selected as the model and a prompt describing the intended image. OpenAI’s current guide covers image generation and configurable output options, but its examples feature newer model identifiers. Check the current generation endpoint schema and confirm that the model and each parameter you plan to use are supported before putting an older gpt-image-1 example into production. See the current image-generation guide.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →#1 Best Overall
Write prompts to make the desired result clear: describe the subject, setting, composition, and visual style that matter to your application. Treat the model identifier and output settings as separate choices. Output size, quality, background, format, and compression may be configurable, but availability varies by model and request path; do not assume a setting documented for a newer model applies to gpt-image-1.
Edit an existing image
For a direct edit, use the Images API edit operation and provide the source image together with a prompt that states what should change. The edit route can use one or multiple reference images. OpenAI’s current edit reference lists up to 16 images, with PNG, WEBP, or JPG files under 50 MB each; confirm the limits for the selected model and request path before relying on them. See the edit endpoint reference.
Be specific about which parts should remain unchanged as well as what to alter. When a localized change is needed, provide a mask so the request identifies the intended edit region.
Use a mask to localize an edit
In the documented file-mask path, transparent areas in the mask indicate where the image should be edited. The reference specifies a PNG mask under 4 MB, with the same dimensions as its source image. Ensure the mask transparency and alignment match the region you intend to change; otherwise, the mask may not describe that region as expected. Check these requirements against the exact model and endpoint path in use. OpenAI’s edit reference
Distinguish input fidelity from output quality
For gpt-image-1, input_fidelity controls how much effort the model puts into matching features and style from a reference image. The documented values are low and high, with low as the default. It is not the same setting as output quality, which concerns the generated result. Select each independently according to the role of the reference and the output you need. Edit endpoint parameter reference
Parameter support is model-specific: the current reference says input_fidelity does not apply to gpt-image-2. That difference is a reason to avoid carrying parameters across model changes without checking the relevant endpoint documentation.
Decode and save the returned image
GPT Image responses return image data as base64 rather than the URL-style response associated with some older image endpoints. Your application needs to read the returned data, decode the base64 into bytes, and write those bytes to a file or pass them to the next stage of your workflow. GPT Image outputs support PNG, JPEG, and WEBP; compression is documented for JPEG and WEBP. Check the response structure and supported output options for the model and endpoint you select. OpenAI’s edit reference
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Check pricing and access before budgeting
OpenAI’s April 2025 launch announcement listed gpt-image-1 rates of $5 per million text-input tokens, $10 per million image-input tokens, and $40 per million image-output tokens. It estimated square-image costs at approximately $0.02 for low, $0.07 for medium, and $0.19 for high quality at launch. These are historical figures, not verified October 2026 rates; actual cost depends on token use and request details. Check the current pricing information before estimating a live workload. OpenAI’s 2025 launch announcement
Free tools Windows power users keep installed
One-click scans. No signup required.
The available documentation does not establish current gpt-image-1-specific pricing, geographic availability, or retirement timing. Confirm model availability and account requirements in OpenAI’s current developer documentation and console before committing to it. The launch announcement also said some developers might need organization verification; treat that as launch-era guidance, not a guarantee of today’s requirements.
Practical checks before shipping
- Confirm
gpt-image-1is currently available to your account and that the endpoint and parameters match that model. - For edits, validate each image’s format and size, and ensure any mask is a matching-dimension PNG with transparency in the intended edit area.
- Handle the response as base64 image data: decode it before saving or forwarding the output.
- Choose the API surface to fit the interaction: direct Images API requests for straightforward generation or edits, and the Responses API tool for conversational, iterative, or multi-step flows.
- Budget using current pricing rather than launch-era examples.
OpenAI’s launch announcement described its API data practice at that time this way: “By default, we never train on customer API data, and all image inputs and outputs remain subject to our API usage policies.” That is a dated statement from the 2025 announcement, not a substitute for reviewing current terms and policies. Read the announcement.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
