Flux Inpaint and Outpaint Workflow — Flux Inpaint Outpaint

Flux Inpaint and Outpaint Workflow is a ComfyUI workflow for inpainting and outpainting, using an image and mask information and providing image previews and saved images.

The graph includes LoadImage for an image, with ImageToMask and MaskComposite for mask handling.

PreviewImage and SaveImage are listed for image previews and saved images.

The instructions place FLUX.1-Turbo-Alpha.safetensors in models/loras/flux and FLUX.1-dev-Controlnet-Inpainting-Beta-fp8.safetensors in models/controlnet.

For a 16GB VRAM setup, the instructions say to rename diffusion_pytorch_model.safetensors and convert it to FP8 with Kijai’s script.

The listed model files are ae.safetensors; FLUX.1-dev-Controlnet-Inpainting-Beta-fp8.safetensors; FLUX.1-Turbo-Alpha.safetensors; flux1-dev-fp8-e4m3fn.safetensors; t5xxl_fp8_e4m3fn_scaled.safetensors; and ViT-L-14-BEST-smooth-GmP-TE-only-HF-format.safetensors.

The listed add-ons are ComfyUI-KJNodes and comfyui_segment_anything.

The author describes alimama-creative/FLUX.1-dev-Controlnet-Inpainting-Beta and alimama-creative/FLUX.1-Turbo-Alpha as the model pair for eight-step inpainting and outpainting in one workflow.

The author lists automatic segmentation as an option for systems with more than 16GB VRAM.

Before setup, verify the diffusion model filename: the instructions name flux1-dev-fp8-e4m3fb.safetensors, while the listed model file is flux1-dev-fp8-e4m3fn.safetensors.

Vorschlag · nicht geprüft

The author says this workflow is no longer state of the art and will not be updated; for current inpainting or outpainting, they point to Flux.1 Fill and the official ComfyUI workflows.

Vorschlag