⭐ Star Save Panorama JPG+

The plain Star Save Panorama JPEG is the "make my 360 image viewable" node. Star Save Panorama JPG+ is that same node, plus a party trick: feed it a depth map and it will generate a stereoscopic 3D version of the panorama - side-by-side or top/bottom - alongside the normal metadata-tagged JPEG. If you're building 360 worlds and want people to actually feel the depth instead of just panning around a flat wrap, this is the one to reach for.

How it works

Two separate jobs run in one node. First, the standard panorama save: each image is written as a JPEG with the Google GPano XMP block injected into the file (same trick as the plain node, cylindrical projection). Second, if stereo_3d is on and you've connected a depth_map, it builds a second image by shifting pixels horizontally in proportion to depth - closer stuff shifts more, far stuff less - and packs the result into the layout you picked. Gaps left by the shifting are filled with OpenCV inpainting (cv2.inpaint, TELEA), which is why gap_fill defaults to "Inpaint" rather than "None": the inpaint pass is what makes the stereo view look clean instead of torn.

Because the depth-to-shift math runs through OpenCV, opencv-python is a hard dependency for the stereo part. It's in the pack's requirements.txt, so a normal install covers you.

Inputs that matter

• images - the panorama batch.
• quality - JPEG quality, default 95.
• stereo_3d - the master switch. Off by default, and without a depth_map connected it does nothing.
• stereo_layout - SBS or Top/Bottom, depending on what your viewer/headset expects.
• depth_scale - strength of the 3D effect; higher shifts pixels further. Default 5, and the shift scales with image width.
• invert_depth - flip this if your depth map uses white for far instead of near; get it wrong and the whole scene reads inside-out.
• gap_fill - Inpaint (best quality) or None.

Optional inputs: depth_map (the IMAGE that makes 3D possible) and base_path (a save folder, e.g. the path output from Star Save Image+ if you want everything in one place). There's also subfolder and filename for path control.

Outputs

It returns two IMAGE sockets: image (what you saved) and 3d_image (the generated stereo view). Wire 3d_image into a preview node before committing - you want to eyeball the depth effect once before batch-saving dozens of files.

Installing and troubleshooting

Same pack install as everything else in StarNodes:

cd ComfyUI/custom_nodes
git clone https://github.com/Starnodes2024/ComfyUI_StarNodes
pip install -r requirements.txt

ComfyUI Manager users can just search "Starnodes" and click Install.

The one gotcha people hit is ordering: if you're generating a depth map from a ControlNet/depth node, make sure it matches the panorama's dimensions. The node does resize the depth map to the image size if they differ, but a wildly mismatched or garbage depth input will produce a garbled stereo view no matter what. Check the 3d_image output before you trust it.