Warp Preview Overlay
Peek Before You Commit
- image
- source_pose
- target_pose
- preview
Before you spend a full render on a pose, you want to know one thing: does it line up? CP_WarpThenPreview is the quick look - it runs the pack's default warp path (no geometric warping, just the target skeleton blended over your sprite) and returns a single preview image. It's CP_WarpToPose with every optional knob set to "just show me."
How it works
It calls the same WarpToPose machinery with defaults - method=none, skeleton_blend at whatever you set (0.70 default) - and returns only the blended result. That means you're seeing your sprite with the target pose's OpenPose sticks overlaid at partial opacity: exactly the "guided" image the full pipeline would feed into a repair or edit model, minus the rest of the outputs.
The point isn't precision. It's that you can throw this in front of a preview node, scrub through target poses, and reject the ones where the sticks clearly don't match your character's proportions before you pay for a generation.
The inputs
- image - your sprite.
- source_pose - the character's current pose (
POSE). - target_pose - the pose you want to move it to (
POSE). - skeleton_blend - 0 to 1, how strongly the sticks show. 0.70 default.
Output: preview - the blended image.
Installing
Part of Spit8/_ComfyUI_CharacterPose:
cd ComfyUI/custom_nodes
git clone https://github.com/Spit8/_ComfyUI_CharacterPose
pip install -r _ComfyUI_CharacterPose/requirements.txt
Restart ComfyUI; it's under CharacterPose/Warp.
Gotchas
- It's the default path only. If you wanted an actual geometric warp, this isn't it -
methodisn't exposed here. For warp control, useCP_WarpToPosedirectly. - No alignment rescue: if the target pose doesn't roughly match your sprite's size, the preview shows sticks in the wrong place and you'll be tempted to blame the node. That's what
align_to_sourcein the full node is for - this convenience version keeps it simple. - It's a preview node in spirit but not in type - wire the
previewoutput into aPreviewImageyourself.
See the pose before you commit to it. That's the whole pitch, and it'll save you a surprising number of wasted renders.
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | — | |
| source_pose | POSE | — | |
| target_pose | POSE | — | |
| skeleton_blend | FLOAT | 0.700–1 | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| preview | IMAGE | — |