Nodes/comfy-ovum/IMAGE_EX Context
ComfyUI Node

IMAGE_EX Context

Override one field of an image record and keep the rest

By sfinktah·Created about a year ago·Updated 10 months ago· 7
IMAGE_EX Context
  • image_ex
  • image
  • mask
  • prompt
  • workflow
  • edit_list
  • IMAGE_EX
  • IMAGE
  • MASK
  • FILEPATH
  • PROMPT
  • WORKFLOW
  • EDIT_LIST
filepath

IMAGE_EX Context is a pass-through node with a superpower: it lets you modify one part of an image record - the image, its mask, its filepath, its embedded prompt or workflow - while every other part flows through completely untouched. The pack's own description nails the pitch: use it to change parts of an IMAGE_EX structure "without needing to demux and remux it completely."

First, what's an IMAGE_EX? It's the richer image record Ovum's workflow-aware loaders produce. Core ComfyUI's Load Image gives you image + mask. The pack's "Load Image with Workflow" instead hands you a bundle - {image, mask, filepath, prompt, workflow} - so you can carry an image around with its provenance attached. That's the container this node edits.

How it works

convert() takes a base image_ex dict, applies whatever overrides you connected, copies the result safely, and hands back the patched bundle plus each field demuxed onto its own socket. Anything you don't connect comes straight through from the base. It's a pure merge - no mutation of the input, no side effects, just "base plus overrides."

Inputs and outputs

All inputs are optional:

  • image_ex (DICT) - the base IMAGE_EX record, usually from Load Image with Workflow.
  • image (IMAGE), mask (MASK) - visual overrides.
  • filepath (STRING) - path override.
  • prompt (DICT), workflow (DICT) - metadata overrides.
  • edit_list (LIST) - threaded through; the single-context node mirrors it out but doesn't merge it into the dict the way the other fields do.

Outputs: IMAGE_EX (DICT), plus IMAGE, MASK, FILEPATH, PROMPT, WORKFLOW, EDIT_LIST - the full bundle and every field on its own socket. That split output is the hidden gem: you can keep the bundle for downstream context-aware nodes while also grabbing, say, just the IMAGE to feed a sampler that doesn't care about metadata.

When you'd reach for it

  • Fix one thing, keep the rest. Swapped a mask or want to re-point a filepath? Override that one input; provenance survives.
  • Threading metadata through edits. Load a workflow-baked image, patch its prompt, pass the updated record on.
  • Deciding what "context" means to you. Need the bundle and the plain image simultaneously? It hands you both.

Installing it

It ships in sfinktah/comfy-ovum:

cd ComfyUI/custom_nodes
git clone https://github.com/sfinktah/comfy-ovum

or search comfy-ovum in ComfyUI Manager, then restart. No models.

Gotchas

IMAGE_EX is this pack's convention, not a core ComfyUI type - so the natural source is Ovum's own workflow-aware loaders, and the natural consumer is other nodes that know the shape. If everything upstream is plain core Load Image, this node has less to offer you. And if you're processing a batch rather than a single image, the list variant (IMAGE_EX Context (List)) is the one you want - this node is deliberately single-image. For one record with a one-field fix, though, it's exactly the tool.

Categoryovum/image

Inputs (7)

NameTypeDefaultDescription
image_exoptDICT
imageoptIMAGE
maskoptMASK
filepathoptSTRING
promptoptDICT
workflowoptDICT
edit_listoptLIST

Outputs (7)

NameTypeDescription
IMAGE_EXDICT
IMAGEIMAGE
MASKMASK
FILEPATHSTRING
PROMPTDICT
WORKFLOWDICT
EDIT_LISTLIST