wtf is image-prompt-corrector?

Image Prompt Corrector is a desktop application, built with Python and Qt, that helps you turn a rough or messy image generation prompt into a clean, faithful one for Krea 2 or FLUX.2 Klein 9B. Instead of calling out to a cloud AI service, it talks to a chat model you already have running locally through LM Studio or Ollama, and uses that model to clean up spelling, tighten wording, and reorganize your prompt so that details about an object stay grouped with its own attributes and actions. The app is organized into separate modes: a main Prompt Corrector for general image prompts, a Comic Story mode, a Meme Creator with its own caption and layout tools, a Model Chat mode for talking directly to your local model, and a Workbench that closes the loop by letting you review the actual generated image against your prompt. Several correction styles are offered, ranging from an Exact mode that prioritizes staying faithful to your original wording, to Explore and Improve modes that allow more creative rewriting. It includes a number of built in checks meant to catch problems before you spend time generating an image, such as verifying object counts, catching contradictions or unclear actions, translating vague or slang phrasing into concrete visual language, and turning abstract feelings into concrete visual cues like posture, lighting, and framing. There is also an optional feature for grounded web verification and reference image analysis, plus batch correction from a CSV file. To run it you need Python 3.10 or newer, the PySide6 library, and either LM Studio with its local server turned on or a running Ollama installation with a compatible model downloaded. After installing the Python dependencies in a virtual environment, you launch the GUI with a provided shell script or by running the main Python file directly. The full README is longer than what was shown.