PoseEditor (🚫Deprecated)

Short version up front: this node is dead. It lives under Easy-Use's own "🚫 Deprecated" category in the current build, its schema shows exactly one input and zero outputs, and that's not a documentation gap - a pose editor with no output is a shell that can't actually feed a pose into anything. If you found this page because a workflow you downloaded uses it, the fix is to swap it out, not to fight it.

What it used to be. PoseEditor was one of ComfyUI-Easy-Use's original nodes, present since v1.0.0, built on top of the poseNode from AlekPet's ComfyUI_Custom_Nodes_AlekPet pack. The idea was solid: an in-canvas OpenPose skeleton editor, where you drag joints around directly on the node to pose a stick figure, and it renders that pose out as an image you could feed into a ControlNet - no separate app, no external pose reference photo, just click-and-drag inside ComfyUI. For a specific kind of precise pose control, that's genuinely useful, which is why it made the cut in v1.0.0 in the first place.

Why it's gone. The current schema takes one required image input (a STRING, likely a base64 or file reference the old JS widget used to populate) and produces nothing. Whatever wiring used to connect the frontend editor to a usable pose output has been stripped or disabled in the pack's current state, and the maintainer has flagged it as deprecated rather than quietly leaving it broken - which is the honest way to handle a feature that no longer works.

Install. If you're pulling the pack for its other nodes, install is the same as always - ComfyUI Manager (search "ComfyUI Easy Use") or:

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

then restart. There's nothing pose-editor-specific to install; you just won't get a working pose editor out of it.

What to use instead. For drawing a pose by hand inside ComfyUI, ComfyUI's own frontend has gained native pose-editing capability, and AlekPet's ComfyUI_Custom_Nodes_AlekPet pack (the original source this node was built from) is still maintained separately if you want the source implementation directly. For the more common case - you have a reference photo and want its pose, not a hand-drawn one - the standard move is a DWPose or OpenPose ControlNet preprocessor run over your reference image, which gives you a real, photo-accurate skeleton instead of an approximated one anyway.

Common issues. The main failure mode is confusion, not a crash: the node loads fine, you can wire an image into it, and then you sit there wondering why nothing comes out the other side. That's expected given zero declared outputs - there's nothing to wire downstream from this node in its current state. Don't spend time debugging it; route around it.