I’m building a paint app where ComfyUI workflows live as layers — pose a 3D doll, hit generate, and the character follows the pose
retained text
Hi all, I'm the solo developer of LoomStroke, a desktop paint app with ComfyUI integration. I wanted to share a short demo of the workflow I've been working on.
1. Place a 3D drawing doll on the canvas and pose it by dragging its joints.
2. Pick a ComfyUI workflow I registered in the app and drop it onto the layer stack as an "AI effect" layer.
3. Connect two layers to the workflow's image inputs: the character illustration as image1 and the 3D doll as image2.
4. Type the prompt ("Change the pose in image1 to match the pose in image2") and press Generate. This queues the job on my local ComfyUI.
5. The result comes back as a layer, with the character in the doll's pose.
**Why I built it this way:**
- The workflow sits in the layer stack like a filter, so the source layers stay untouched. I can re-pose the doll and generate again.
- Any workflow exported in API format can be registered. The app builds a parameter form from the nodes, so I don't have to switch to the browser for every tweak.
- Layers are wired straight into the workflow's image inputs, so there is no export → upload → download → import loop.
It's still under active development and there are rough edges. I'd love feedback from people who use ComfyUI every day, especially on which workflows you'd want to drive from a canvas like this.
https://loomstroke.com/
https://github.com/NakamuraShippo/LoomStroke
I’ve entered the AICU Awards, and the deadline is in about an hour—please cast your vote!
https://award.aicu.ai/2026/gallery/index.ja?w=26G30&field=GAME_TOOL_APP document fetched 2026-11-15 00:36 · canonical https://reddit.com/r/comfyui/comments/1xw0c55/im_building_a_paint_app_where_comfyui_workflows