Fusion 360 Image Import

If you're using ComfyUI to generate textures or concept art for CAD models, the pain point is always the same: getting an image out of Fusion 360 and into the graph. MKRFusion360ImageImport is the bridge node that turns a Fusion 360 render, viewport, or texture capture - delivered as JSON payload - into a proper ComfyUI IMAGE (plus an optional MASK), so you can process it, img2img it, or generate a matching texture.

It's the "read" half of the pack's Fusion 360 addon lane. The companion add-on scaffold (shipped in addons/fusion360/ as a Fusion 360 Python add-in) exports view/material/body snapshots and hands them to ComfyUI; this node is what catches that packet on the ComfyUI side. If you haven't installed the Fusion 360 add-in, you can still use the node - the payload is just JSON with a path to an image, and you can paste it by hand.

How it works

The node is a thin, purpose-built subclass of the pack's shared host-image import base, so the behavior is identical across the Blender/Maya/Fusion 360 image-import family. You give it:

• payload_json (required) - the JSON packet from the add-on (or hand-built). It carries the asset path - the node's docs call it out as the thing that matters: "No image asset path found in payload_json" is the error you'll see if it's missing.
• preferred_slot (optional) - a slot hint for which capture to load when the payload lists several candidates. Empty string = default.

It resolves the best candidate path from the payload, loads it as an IMAGE (and builds a MASK if the source has alpha), and hands back:

• image (IMAGE) - wire this into your sampler/VAE/img2img.
• mask (MASK) - alpha/selection, if the capture has one.
• payload_json (STRING) - the resolved packet, with the chosen path.
• asset_path (STRING) - the actual file path that was loaded. Handy for downstream logic.
• summary_json (STRING) - host, path, candidate count, dimensions, warnings.

The summary_json width/height and candidate count are your sanity check: if you expected the 4K render but got a viewport thumbnail, the candidate selection picked the wrong slot - that's what preferred_slot is for.

Installing

Two parts. The ComfyUI side is just the pack:

cd ComfyUI/custom_nodes
git clone https://github.com/criskb/MKRShift_Nodes

Restart ComfyUI (or ComfyUI Manager, search "MKRShift_Nodes"). The Fusion 360 side is the add-in scaffold under addons/fusion360/ in the repo - a Fusion 360 Python add-in you load inside Fusion. You need both halves for the full round trip; the node alone just parses JSON.

The honest take

This is squarely a CAD-bridge tool for people doing texture/material handoff with Fusion 360. It's a clean, minimal node - parse payload, resolve path, load image, summarize - and its value is entirely in the pipeline it's part of, not in anything fancy it does on its own. There's no community signal for it yet; expect to read the add-on scaffold to see exactly what payload fields it emits. If you're not doing Fusion 360 handoff, this node has nothing for you - but if you are, it's the glue that makes the round trip possible without hand-copying file paths.