Merge Monoscopic into Stereo

You've got two separate 360° equirectangular images and you want them to behave like one 3D image - a stereo skybox, a VR180 export, an HDRI pair for Blender. That's exactly the gap this node fills: it takes two monoscopic 360s and packs them into a single stereo frame.

It's part of the pytorch360convert pack from ProGamerGov, which wraps the author's own pytorch360convert Python library into ComfyUI nodes. This one is the sibling of the pack's "Crop Stereo to Monoscopic" node - that one rips a stereo image apart, this one glues two halves back together. If you're generating 360s and want to view them in 3D, this is the node you reach for.

How it works

Under the hood it's delightfully boring: a plain tensor concatenation. Your two images get joined along the width (side-by-side) or the height (top-and-bottom), and the result comes out as one image. No depth warping, no parallax magic, no hallucinated geometry - if you want genuine stereoscopic depth, the two inputs need to be an actual left/right eye pair already, either because you generated them as one or shifted them by an IPD offset for your scene.

The one input to care about is merge_direction:

• horizontal (default) - first image goes on the left, second on the right. This is the classic side-by-side (SBS) layout that most VR players and SBS workflows expect.
• vertical - first on top, second on bottom. Useful for over/under 3D formats or when your viewer wants top-bottom.

The output is a single IMAGE named Stereo Image, which you can preview, save, or pipe straight into a 360 viewer. The images should match in height (for horizontal) or width (for vertical), or the concat will still run but the result will look lopsided.

Installing it

Same drill as the rest of this pack:

cd ComfyUI/custom_nodes
git clone https://github.com/ProGamerGov/ComfyUI_pytorch360convert
python -m pip install pytorch360convert

Restart ComfyUI and the nodes appear under the "pytorch360convert" heading. Easier still: ComfyUI Manager, search "ComfyUI_pytorch360convert", hit Install. The only Python dependency is pytorch360convert itself - a single small library, no model files to download, so this pack rarely triggers the dependency hell that custom nodes are famous for.

Things to watch

• The name says "monoscopic" but nobody's checking that your inputs are actually flat. Feed it two views from "Equirectangular to Perspective" at slightly different yaws and you've got a crude stereo pair in three nodes.
• This is packing, not rendering. If the two eyes aren't aligned, the 3D effect will be garbage - a stereo image is only as good as its left/right inputs.
• For viewing the result inside ComfyUI, the author maintains ComfyUI_preview360panorama; your default preview node will show you the SBS frame as a flat picture, which looks weird until you realize that's the point.

It's the least glamorous node in the pack, and that's fine - sometimes you just need two images stapled together in a way that a VR headset understands.