AIO Inpaint Info
See the working crop before it's stitched
- inpaint_info
- source
- sample
- mask
AIOInpaintInfo is a debugging unpacker, and it's the thinnest node in the pack for a reason. AIO Image Generate bundles its inpaint internals into one opaque inpaint_info output rather than spraying wires across the graph - and when something goes wrong with a crop/stitch inpaint, "opaque" is the last thing you want. This node unpacks that bundle into three inspectable values so you can see what the sampler actually worked on before the final composite.
What it does
One required input, inpaint_info, wired from the inpaint_info output of AIO Image Generate. Three outputs:
source(IMAGE) - the prepared inpaint working image that was fed into the inpaint path.sample(IMAGE) - the decoded sample before stitching or blending, i.e. what the model produced for the masked region before it was composited back onto the canvas.mask(MASK) - the working mask as the inpaint pipeline saw it.
If that sounds like exactly what you'd want to look at when a seam shows up, that's the point: the README frames these as "debug values ... so crop/stitch behavior can be inspected before final compositing." Wire source, sample and mask into preview nodes, run, and eyeball the working crop, the raw model output, and the mask that's going to blend them.
When you'd actually reach for it
You won't use this node for normal generation. You'll use it when:
- A crop/stitch inpaint comes back with a visible seam and you need to know whether the problem is the mask (feather too thin, grow too small), the sampled crop, or the blend.
- The full-frame fallback path shrinks your image and you want to confirm the working mask and source at the sizes the pipeline actually used.
- You're tuning
mask_grow_percent,mask_feather, orcontext_from_mask_extend_factoronAIO Inpaintand want to iterate on evidence instead of guesswork.
The heavy lifting still happens in AIO Inpaint - this node only reads a bundle that's already there. It's plumbing in the best sense: invisible until you need it, then exactly what you needed.
Install and wiring
It ships with the AIO pack:
cd ComfyUI/custom_nodes
git clone https://github.com/helto4real/comfyui-all-on-one-image-generation-node
cd comfyui-all-on-one-image-generation-node
python -m pip install -r requirements.txt
Or via ComfyUI Manager (search "AIO Image Generate"), restart, and it appears under AIO/Image. Connect AIO Image Generate.inpaint_info → AIOInpaintInfo.inpaint_info, and put previews on the three outputs. That's the whole workflow.
Gotcha
One real one: the older AIO Image Generate exposed direct inpaint debug outputs that some workflows used. This pack's migration moved those behind inpaint_info, so if you loaded an old workflow wired to the old sockets, you'll need to reconnect through this node - the old direct wires no longer resolve. And it only shows values when the inpaint_info bundle actually carries them, so connect it to a run that had an active AIO Inpaint config, or the outputs come back empty.
Nothing glamorous here - and that's fine. When the seam shows up, this is the node that tells you which of the three stages to blame.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| inpaint_info | AIO_INPAINT_INFO | Inpaint debug previews from AIO Image Generate. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| source | IMAGE | — |
| sample | IMAGE | — |
| mask | MASK | — |