ComfyUI-KYNode

Pull one item out of a list or batch

Stop wiring three separate branches into three separate nodes

The adapter between boxes and segmentation

Turn coordinates into a usable box

The un-packager for a box object

Draw a synthetic mask instead of hand-painting one

Turn a plain file path into a real VIDEO socket

Crop by exact numbers, with divisible-size safety

Get a preview without saving anything

A real Python interpreter living inside a node

Mux a soundtrack onto a silent clip

The escape hatch for everything the other nodes don't cover

Image Sequence to Video — full encoder control for frames already on disk

Trim Video — cut a clip by start time and duration

Video to Image Sequence — extract frames straight to disk

Split a path into every piece you'd actually reuse

Reverse-engineer a numbering template from one filename

A fallback chain, and a real gotcha with the number zero

Read whatever's sitting at a path, mid-graph

The drag-slider comparison, for stills instead of video

The detector-driven half of the crop pair

A gate, not just a check

Stitch mixed-type values into one string without cast nodes

Turn a vision-LLM's answer into a real box

One loader instead of three

Multithreaded bulk loading, with the filtering to actually control it

A from-scratch Load Video, tuned per model

Write the formula instead of chaining ten nodes

Combine two JSON strings without writing the merge logic yourself

Caption one image with whatever VLM you've got running

Caption a whole dataset in one pass

Talk to any local LLM that speaks the OpenAI protocol

The one-file loader, when you know exactly which file

Pull one piece out of a string with a pattern

Chained find-and-replace for prompts and paths

Undo the grid math a vision LLM did before it answered

Batch-save frames with alpha and a frame range

The fast, no-metadata batch saver

The one place this pack's README actually explains itself

Build a bounding box from raw X/Y/Width/Height numbers

Pull an HTTP URL out of a video object

POST your generated media straight out of the graph

A drag-to-compare slider for two videos, right on the canvas