MFLUX
Local AI image generation toolkit runs on Apple Silicon Macs with MLX, model-specific editing and LoRA fine-tuning under the MIT license.
MFLUX is a Python toolkit for creators and developers who want to generate and edit images on an Apple Silicon Mac. It provides a command-line interface for working with models locally.
- Image generation. Choose models for different visual tasks, including Krea 2 for varied styles, Ideogram 4 for typography, and Ming-Image for posters, cards and UI graphics with RGBA output.
- Image editing. Supported models offer image-to-image generation, inpainting, multi-image editing and virtual try-on, with ControlNet and depth guidance available for specific models.
- LoRA customization. Apply multiple LoRA adapters or fine-tune supported models, including Z-Image and FLUX.2.
- Upscaling and depth maps. SeedVR2 enlarges images, while Apple's Depth Pro extracts depth maps for depth-guided work.
- Structured prompts. FIBO accepts JSON-based prompts and includes tools for prompt inspiration and refinement; metadata export and reuse help carry generation settings between runs.
MFLUX uses native Apple MLX implementations and supports quantization and loading local model files. Selected models support float16 computation on Apple M1-family GPUs. Generation runs on your Mac, but editing and training capabilities vary by model; Qwen Image is a larger, slower option. The toolkit is open source under the MIT license.
