Easy Apply Fooocus Inpaint

Fooocus's inpainting was quietly one of the best things about it - good enough that the patch got lifted out of Fooocus and ported into ComfyUI as a node pack that people still maintain. easy applyFooocusInpaint is Easy-Use's one-node wrapper for that patch. It takes your SDXL model and a latent, applies the Fooocus inpaint head and patch weights, and hands back a model that inpaints noticeably cleaner than a raw checkpoint with a mask bolted on. If you're doing traditional mask-based inpainting on SDXL, this is a real upgrade for basically free.

Why this exists

Standard inpainting just masks a region and runs img2img - the model treats the masked area as "generate here," and you get seams and color drift. Dedicated inpainting patches understand the mask as a condition and blend the edges properly. Fooocus shipped its own tuned inpaint model doing exactly that, and because no purpose-trained open inpainting checkpoint has really shipped since late 2024, the older SDXL inpainting tooling is still what a lot of people reach for. This node makes the Fooocus version a two-wire drop-in. It replaced the pack's earlier easy fooocusInpaintLoader, so if a tutorial mentions that older node, this is its successor.

The inputs and outputs that matter

Everything here is required and there are only four:

• model - your SDXL checkpoint's model. SDXL only - the Fooocus inpaint weights are SDXL, so don't hand it an SD1.5 or Flux model.
• latent - the latent you're inpainting into (an encoded image with its inpaint region). The patch needs this to know what it's working on.
• head - the inpaint head weights. There's one option, fooocus_inpaint_head; just select it.
• patch - the actual inpaint patch, and the one real choice you make. inpaint_v26 (1.32GB) is the newest and the one to use; inpaint_v25 (2.58GB) and plain inpaint (1.32GB) are older revisions kept for compatibility.

The single output is a patched model. Wire it into your sampler in place of the original model, keep the rest of your inpainting graph as-is, and sample.

Installing it and the model files

Install the pack via ComfyUI Manager (search ComfyUI Easy Use, restart) or:

cd ComfyUI/custom_nodes
git clone https://github.com/yolain/ComfyUI-Easy-Use

Install requirements and restart. The head and patch dropdowns reference the Fooocus inpaint weight files - the head plus one of the patch files (the 1.3–2.6GB downloads listed above). If the dropdowns are populated, you're set; if they're empty, the weights aren't in the folder the loader scans, which is the usual "why won't it work" answer.

Where it bites

Historically the big one was that applying Fooocus inpaint broke your LoRAs - after patching the model, LoRAs would silently stop working. That was a genuine bug and it was fixed back in v1.2.2, so on any current version you're clear, but it's worth knowing if you're on an ancient install and your LoRAs vanished right after this node. Beyond that: keep everything SDXL, feed a properly encoded inpaint latent (not an empty one), and remember this is a mask-based inpainting tool - if you actually want whole-image editing where the model rewrites everything, that's a different job entirely and this isn't it.