mflux

Why we included this project

Running modern image generation on a Mac usually means renting cloud GPUs. MFLUX sidesteps that by reimplementing current text-to-image and image-to-image models natively in Apple's MLX framework, so generation runs locally on the Mac's unified memory. It covers Z-Image, FLUX, Qwen Image, Ideogram, and FIBO, with quantization, LoRA fine-tuning, in-context editing, ControlNet conditioning, inpainting, and upscaling available across the supported models. A command-line interface and a small Python API make it easy to script batch jobs or wire into an existing pipeline. For teams that want offline generation, privacy, or predictable local costs, this is a practical base for image workflows on Apple hardware.