ZML_全景图预览

A preview node that does zero processing: you feed it a 360° equirectangular panorama and it opens an interactive three.js viewer where you can look around inside it - drag to pan, scroll to zoom, and inspect the scene from the inside out. The 输出 output just passes the image through unchanged. ZML_PanoViewer is pure eyeballs.

If you've ever tried to judge a spherical panorama from a flat thumbnail, you know the problem: an equirectangular image smears the top and bottom into nonsense, and "is the horizon straight / is there a seam / is the floor warped" are unanswerable from the flat projection. This node answers them by putting you in the scene.

How it works

• 全景图像 (IMAGE) - the equirectangular panorama (2:1 aspect is the usual convention for 360° images). Wire it in, run the workflow, and click the node to open the viewer.
• 预览模式 - 360°预览 (the default) or 平面预览 (flat preview). The flat mode is the later addition - the README's 2025.10 notes mention "全景图预览节点增加了平面预览的功能" - and it's there because sometimes you do want the plain equirectangular view for checking stitching artifacts at the edges.

That's the entire input set, plus the pass-through 输出 so you can chain it mid-graph instead of making it a terminal node. Since it returns the image, you can drop it in front of a SaveImage and get both the view and the save from one run.

The honest state of it

The README is refreshingly upfront about this node's maturity: it was published early, "still in testing," and the author pushed it out because someone had reported a bug in the previous version and he'd rather ship the working build than hold it. That's the tone to set expectations by. The viewer works, but it's a test-build utility, not a polished showcase - if it misbehaves on your ComfyUI version, that's the kind of thing the author's README explicitly invites as an issue report.

The one hook the README mentions: it pairs with a "kontexet" (Kontext) LoRA - i.e. FLUX Kontext-style panorama workflows that generate 360° content. If you're producing panoramas for VR, this is the fastest way to QA them without exporting to a separate viewer.

Install

One node in ComfyUI-ZML-Image:

cd ComfyUI/custom_nodes
git clone https://github.com/zml-w/ComfyUI-ZML-Image
# restart ComfyUI

or ComfyUI Manager → search "ComfyUI-ZML-Image". The 3D viewer is served from the pack's bundled three.min.js - nothing extra to download. Chinese-first UI; translation patch at https://github.com/zml-w/ZZZ_ZML_English_Patch.

Word to the wise: this is the pack's poster child for "the author built what he needed." It works, it's useful, and it's also the node most likely to have a version-specific UI hiccup - so update the pack regularly if panoramas are a real part of your workflow.