Crop 360 to 180 Equirectangular
Grab the central half of a full panorama and drop the rest
- image
- Cropped 180 Image
Sometimes you don't want the whole sphere. Maybe you only care about the front 180° of a panorama, maybe a LoRA or model you're using only accepts 180° inputs, or maybe you just want to halve the resolution you're working with. Crop 360 to 180 Equirectangular takes a full 360° equirectangular image and keeps only the central 180°, slicing out the middle half of the width.
It's the exact inverse of the pack's Pad 180 to 360 Equirectangular node. Pad doubles you out, this halves you back - the two are meant to be a matched pair, and they're both dead simple.
How it works
It takes the width, cuts from one quarter to three quarters, and keeps that middle slice. So of the image's full 360°, you keep the front-facing 180° - the region that maps to the front of the panorama - and drop the back half entirely. The height is untouched, so the output keeps the same vertical resolution and the standard equirectangular aspect.
There's exactly one input (the image) and one output (Cropped 180 Image). Nothing else to configure, no geometry to reason about - if the input is a clean 2:1 360, the output is a clean 180° slice of it.
When you'd actually use this
The most common reason is model compatibility: some 360-trained LoRAs and pipelines are built around 180° inputs, so you crop the pano, process, and pad back out when you need the full sphere again. It's also a lazy way to crop out a badly-artifacted back half - if your pano's rear hemisphere is garbage (seams, missing geometry), cropping to the good 180° is a legitimate salvage move, and you can pad and inpaint the rest later.
Nothing surprising happens in this node - it's a plain slice, no resampling, no quality loss, no resizing. It runs instantly.
Install
Same as the rest of the pack: ComfyUI Manager, search "ComfyUI_pytorch360convert", or git clone https://github.com/ProGamerGov/ComfyUI_pytorch360convert into ComfyUI/custom_nodes and restart. The requirements.txt in the pack is empty, so you may need to run python -m pip install pytorch360convert in your ComfyUI environment before the nodes will import. No model files, nothing to download.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| Cropped 180 Image | IMAGE | — |