Jiema GPT Image 2 Generator

Problem

When engineers need posters, covers, product concepts, or social media images, a direct API call often misses use case, aspect ratio, text, and negative constraints, causing rework. This skill turns a short image request into a structured prompt and calls Jiema GPT Image through an OpenAI-compatible Images API.

How it works

It infers prompt, size, n, output directory, and filename, asking only when missing details would break the result. Default size is 1024x1024; use 1536x1024 for horizontal banners and 1024x1536 for vertical posters. It expands simple requests to cover subject, composition, style, text, negative constraints, and quality requirements. If the request can go in very different directions, it offers 2-3 options instead of guessing blindly. After generation, it decodes b64_json or downloads a url, saves the file with a clear name, and verifies it with file and ls -lh. If the API fails, it can check the base URL, size, text length, and style description, then retry with common sizes.

Boundaries

It only generates new images and does not crop, compress, or remove backgrounds. For highly accurate text, generate a clean background first and add text later. API keys should not be written into scripts, logs, prompts, or shared artifacts.