tri3d-comfyui-nodes

Fix a mannequin's neck so the garment model's head doesn't float

Label every body part and garment on a batch

Local model or PhotoRoom API

Drop masks too small to matter, and get a flag saying so

A 'press the empty-cache button' node that passes your image straight through

ClipDrop's remove-background API as a ComfyUI node — key required, no local model

Replace a background by typing a sentence — ClipDrop's API, one node away

Split a side-by-side composite into its two halves, batch included

Crop to your mask, fit the aspect ratio, and pad the sides to keep it centered

DWPose detection with a built-in JSON dump (no separate preprocessor pack needed)

Pick individual face parts out of a facer parsing map (eyes, lips, brows, hair)

Turn ATR segmentation maps into the exact masks you need (and only those)

Cut garment parts out of a batch — the workhorse of the VTON pipeline

Keep the part, black out the rest — the non-cropping version of part extraction

Body-part crops from the coarse Pascal palette (torso, arms, legs, done)

Cut a crop around the torso (or shoulders) so you can regenerate it in place

Is that still the same person? A face-similarity gate for your workflow

Fill a mask by cloning neighboring pixels — no model, no VRAM, instant

Build a mask from 'close enough to this color'

Turn a number into an image so you can actually see it

Find where your histogram actually starts and ends, for sane rescaling

Mean and standard deviation of an array, masked to the region you care about

Let the gradient find the background cut point for you

It's called HistogramEqualization in code, but everyone uses it to fix a neck

Two images, one canvas

Give the model room below the face so it can grow a neck and body

Crop an image to exactly the box your mask defines

Paste the regenerated crop back where it came from

Edges only where the garment meets skin

Detect a bottom-half shot and branch your pipeline

The return trip for recoloring

Garment parsing, but you must bring the model file

Image matting, with the weights fetched for you

See the skeleton your DWpose file actually contains

Force one region's lighting to match another's

Background removal that hands you the alpha channel as a proper MASK

Is that mask actually covering anything?

The node that makes garment crops the tall, skinny shape try-on models want

Auto-swap flagged images for a safe alternative — Stability's own safety checker, in a node

A paid-API background cutout that silently does nothing without a key

Retarget a person's pose to a reference so your garment warper gets the same skeleton

Render a pose skeleton back into an image (without fighting the preprocessor stack)

Paste garment parts back onto a person using ATR segmentation colors

The PASCAL-VOC-flavored sibling of Position Parts Batch

Stamp a cropped pose part back onto the original image by coordinates

Force one garment's colors to match another, masked region only

The color-match node with per-channel dials

Skip the reference image, dial the target mean and sigma yourself

The blunt-force color match in the pack's recolor family

Sweep the specks and flyaways out of your masks

The stats-transfer workhorse

Split an image into L, A, and B channels you can edit independently

Matting with a trimap, for hair and edges that refuse to be cut

Turn any pack's POSE_KEYPOINT into a JSON file TRI3D's pipeline can read

Write output exactly where you want it, not into ComfyUI's folder

The node that writes your prompts and sidecars where you need them

Drop a subject onto a new background at the right scale, bottom-center, no math

The same paste, minus the fit-guarantee — and it will crop your subject

Paste a subject onto a new background and get a usable shadow layer for free

The linear level-stretch utility

The skin seam mask that fixes two-pass virtual try-on

Cut the head out of a photo in one node

Skip the head but keep a neck — the try-on way

Erase everything above the lips in one pass

Turn a lower-body mask into one smart rectangle

Smart Box's black-fill twin (yes, they shipped both)

The tiny contains-checker that unblocks your branching

Swap pixels between two images, guided by a mask

Make ComfyUI wait for an external process, then read its file