What is gpt-image-2-unlocked-api-toolkit?

VisionCraft AI is a wrapper and orchestration layer built on top of OpenAI's GPT-Image-2 model and the Claude API, aimed at generating and editing images from text prompts. Instead of sending a prompt straight to one API, it expands short prompts into more detailed descriptions, can fall back from OpenAI to Claude or a local model if needed, and tries to keep a consistent visual style across multiple generations from the same session. Features described include locking an artistic style across a batch of generations, scaling image detail based on how many tokens are budgeted, editing an already generated image using a natural language instruction, generating multiple style variations within one theme, blending two or more styles together, keeping a version history of past generations, and creating up to 25 images at once with progressive refinement. The interface is described as working on mobile, tablet, desktop, and terminal, with prompts accepted in 47 languages while output stays consistent. It also lists a self hosted deployment option, audit logging of each generation, and no user tracking. To install it, you need Python 3.11 or newer, at least 8 GB of RAM, with 16 GB recommended for batch generation, and 500 MB of disk space, plus an internet connection for API calls. Setup involves installing dependencies with pip, adding your OpenAI and Claude API keys to an environment file, and starting the tool with a serve command. Configuration is also possible through a YAML profile file that sets the default and fallback engine, style anchor text, resolution tier, and export formats like PNG, JPG, or WebP. The project is licensed under the MIT License, which allows commercial use, modification, distribution, and private use. The README notes it is not affiliated with OpenAI or Anthropic and that users are responsible for following the terms of service of whichever API they connect to.