# A standalone command line tool to run Z-Image Turbo

- A Python CLI program that generates images using the local Z-Image
  Turbo model stack referenced by:
  `~/programs/ComfyUI/user/default/workflows/Z-image Turbo.json`
- The program must not depend on ComfyUI at runtime.
- The ComfyUI workflow is only a reference for model filenames, LoRA
  configuration, sampler defaults, image size, prompt fields, and
  generation parameters.
- Do not download model files or any other files from the internet.
  Assume all model files are stored locally.
- Use uv to manage virtual env and dependencies.
- The CLI should support prompt, negative prompt, seed, width, height,
  batch size, steps, cfg, output path, and device options.
- The first implementation milestone is proving a single image can be
  generated from local weights without importing ComfyUI.