SeeThrough Parts To Layers

The See-through pack's default exit is a PSD you take somewhere else. But if you just want to fiddle with the result - nudge the hair forward, drop a layer's opacity, change a blend mode - leaving ComfyUI to open another program is a hassle. SeeThrough Parts To Layers is the node that keeps you in the graph: it converts the pack's custom SEETHROUGH_PARTS data into ComfyUI's native LAYERS document format, which the built-in layer editor and the Create Layered Image node understand. One tiny adapter node, and suddenly your decomposed character is editable right there on the canvas.

It's the least glamorous node in the pack and it does exactly one job. The conversion is mechanical: it takes each part from Post Process - already split into left/right eyes, front/back hair, and so on - and hands it to the layer editor as a raster layer with its canvas position (x, y), its name, and a z_index that's assigned from the part's depth_median. That last bit matters: because the stacking order is derived from the depth values computed earlier in the pipeline, the layer stack in the editor is already in correct draw order. You don't have to re-sort anything; the stuff that's meant to be behind is behind.

The input:

• parts - the parts output from SeeThrough Post Process. Same input the Save PSD node takes, which tells you this node is a fork in the road rather than an extra step: either route goes straight from Post Process.

The output: layers (type LAYERS, the core ComfyUI type) - wire it into Create Layered Image (the node that renders the stack back into an IMAGE) and you can adjust position, opacity, and blend modes per layer in the editor, or route the document wherever else you use LAYERS. The example workflow in the repo (workflows/seethrough-layer-editor.json) shows the intended wiring.

Installing. It ships in the ComfyUI-See-through pack - Manager (search "See-through") or:

cd ComfyUI/custom_nodes
git clone https://github.com/jtydhr88/ComfyUI-See-through.git
cd ComfyUI-See-through
pip install -r requirements.txt

Restart and it appears under the SeeThrough category. No extra models or downloads; the heavy lifting already happened upstream.

Where people get burned: the two things that trip people up are both expectations, not bugs. First, this node doesn't give you the depth channels - the depth data is folded into the stacking order and then dropped, so if you need per-layer depth maps, that's the Save PSD route (it exports *_depth.png files). Second, the editor it feeds is ComfyUI's own layer editor, not a Photoshop clone: it's great for position, opacity, and blend modes, but it won't do mesh warping or masking the way Live2D tooling would. For that, export the PSD. The right mental model: this node is for quick iterative tweaks while you're still in ComfyUI; Save PSD is for taking the character somewhere serious.