Illustration drafts
Generate a few concept images for an article or a slide deck; pick 1792×1024 for landscape and refine the one you like afterwards.
Loading…
Guide
AI Image Generator sends a text prompt to the OpenAI-compatible image endpoint you configured ({Base URL}/images/generations) and shows the returned images in a grid, ready to download as PNG. It asks for responseformat: b64json first, so the image arrives as base64 and the download does not depend on a third-party URL; if the provider rejects that parameter, the tool retries once without it.
Updated 2026-09-093 min read
AI Image Generator sends a text prompt to the OpenAI-compatible image endpoint you configured ({Base URL}/images/generations) and shows the returned images in a grid, ready to download as PNG. It asks for response_format: b64_json first, so the image arrives as base64 and the download does not depend on a third-party URL; if the provider rejects that parameter, the tool retries once without it.
It suits anyone who already has an account with OpenAI, Zhipu AI, SiliconFlow or another provider that exposes an image endpoint. Note that text-only services such as DeepSeek have no image endpoint, and the page says plainly that image generation is not supported.
| Input | Output | Notes |
|---|---|---|
Prompt + 1024×1024 + 2 images, server supports b64_json |
A grid of 2 images, downloaded as local PNGs |
The image data comes back with the response, with no extra request to an external URL |
Server returns only a remote URL and blocks cross-origin downloads |
“The image server does not allow direct downloads. The image was opened in a new tab — right-click and Save image as.” |
The fallback when the download fails |
Base URL points at DeepSeek |
This endpoint does not support image generation (/images/generations is missing)… |
A dedicated message for 404/405 |
/images/generations; text-only model services answer 404/405 and are reported as unsupported.Generate a few concept images for an article or a slide deck; pick 1792×1024 for landscape and refine the one you like afterwards.
Enter dall-e-3 and gpt-image-1 in turn with the same prompt to compare style and cost.
Generate square 512×512 or 1024×1024 images and post-process them with local tools such as Image Rounded Corners.
That is settled by your agreement with the provider you use. This site claims no rights and stores no images. Before commercial use, check the provider's policy and avoid content that infringes anyone's likeness or trademarks.
When the provider returns only a remote image URL and that server does not allow cross-origin requests, the browser cannot save the file directly. The tool opens it in a new tab instead, where you can right-click and save it.
Your provider bills by model and size; this site charges nothing and never sees the bill. Check the provider's documentation for current prices.
The prompt and the API key go from your browser straight to the endpoint you configured, and this site is not involved. Generated images come back to the browser as base64 or a URL and are not uploaded here. The key is kept in memory only by default. Do not describe content that involves other people's privacy or is unlawful.
Updated 2026-09-09
Generate images from text (DALL·E-compatible API) in several sizes and counts
AI image generation needs an OpenAI-compatible endpoint
Fill in the endpoint URL (Base URL) and the API Key on the Settings page.
The provider must support a DALL·E-compatible /images/generations endpoint
(OpenAI, 智谱, 硅基流动 and others support it; some text-only services such as DeepSeek do not support image generation yet).
The key is kept only for this session (cleared on refresh) and is never uploaded to any server; you can enable "Remember the key on this device" in Settings to keep it (stored in plain text — do not enable it on a shared device)