Nodes/ComfyUI-DreamCube/Insert Cubemap Face
ComfyUI Node

Insert Cubemap Face

Put a processed face back where it belongs

By cedarconnor·Created 10 months ago·Updated 9 months ago· 2
Insert Cubemap Face
  • cubemap
  • image
  • CUBEMAP
face

Insert Cubemap Face is the return half of the cubemap round trip. Extract Cubemap Face gives you a single square face as a normal image; this node takes a processed image and writes it back into a specific face of a CUBEMAP. Together they're how you build a manual, face-by-face pipeline - run each of the six faces through different nodes (depth here, upscaling there, maybe a different model on the top face because ceilings are weird), then reassemble.

It's a simple copy operation under the hood: the incoming IMAGE becomes the face, and the node returns a new cubemap so it doesn't mutate the one upstream of it in the graph. That copy-on-write behavior is a pack-wide convention and worth knowing - nothing in DreamCube quietly rewrites your inputs.

Inputs and outputs

  • cubemap - the target CUBEMAP, typically from Equirect to Cubemap or Create Empty Cubemap.
  • image - the processed face, as a standard IMAGE from any node.
  • face - which of the six slots (front, back, left, right, top, bottom) to overwrite.

Output is a CUBEMAP with that face updated. Keep in mind this node only touches the RGB face data. If you're inserting a depth map, that's Apply Depth to Cubemap Face's job, not this one.

How to install it

ComfyUI Manager, search "DreamCube", install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/cedarconnor/ComfyUI-DreamCube.git
cd ComfyUI-DreamCube
pip install -r requirements.txt

Then restart. No model downloads, no API keys - this pack is all local NumPy/SciPy math.

Gotchas

The image you insert must be square and match the cubemap's face resolution, or you'll get a size doesn't match cubemap resolution error. Upscaling a face before inserting it is a real workflow - but only if you upscale the whole cubemap's resolution with it, which DreamCube doesn't do for you. The practical loop people actually run: extract all six, process each, insert each back, then hit the result with Enforce Depth Consistency before converting back to equirect. That last step is the one that hides the seams between faces.

CategoryDreamCube/Projection

Inputs (3)

NameTypeDefaultDescription
cubemapCUBEMAP
imageIMAGE
faceCOMBOWhich face to overwrite/insert

Outputs (1)

NameTypeDescription
CUBEMAPCUBEMAP