Lux3D Multi-View Generator

The Lux3D Multi-View Generator is the simplest node in the pack, and it does one quietly useful thing: you hand it a single image of an object, it returns four consistent views of it from different angles, as URLs.

Why is that useful? Because a good chunk of 3D generation is multi-view-first. The whole Zero123/InstantMesh lineage works by hallucinating a set of consistent views and then reconstructing a mesh from them - a single image is ambiguous, but four agreed-on angles give the reconstruction something solid to work with. So these four images are a natural input to feed back into Lux3D Image to 3D's eight slots, or to hand a game artist as a ready-made turnaround reference. It's also just a fast way to see "what does this thing look like from the back?" before you commit to a full 3D generation.

How it works

Standard Lux3D pack machinery, trimmed to the bone: the node takes your image (public URL, upstream STRING, or one connected IMAGE), uploads it if it's local, POSTs to /lux3d/v1/generate/image-to-four-view/task/create, and polls until four outputs land. That's the whole input side - there's no version, no face count, no format menu. Just image and base_api_path.

Inputs and outputs

Only two inputs:

• image - required. One image, any of the three source types.
• base_api_path - https://api.aholo3d.cn (default) or https://api.aholo3d.com. The matching API-key env var is used; no trailing slash.

Outputs are task_id plus image_1 through image_4, all plain strings holding URLs. The node polls expecting exactly four results, so if the service returns fewer you'll get an error rather than a half-empty node - which is honestly the right behavior.

Install

Same as the rest of the pack - it ships in one repository:

# ComfyUI Manager → Custom Nodes → Install via URL
# https://github.com/manycore-research/ComfyUI-Lux3D.git
# or:
comfy node install lux3d
# or manually:
cd path/to/ComfyUI/custom_nodes
git clone https://github.com/manycore-research/ComfyUI-Lux3D.git
pip install -r requirements.txt

Set LUX3D_API_KEY_CN or LUX3D_API_KEY_INTL in the environment that starts ComfyUI (keys from labs.aholo3d.cn / labs.aholo3d.com), restart, done. No model downloads - dependencies are requests, Pillow, and numpy.

The fiddly bit

The four outputs are URLs, not images in your workflow. To actually look at them you'll want to grab them with a download node or just open them in a browser - there's no auto-save here. The natural pattern is to chain them into Image to 3D: generate four views, wire the URLs into a few image slots, and let the reconstruction use your multi-views as better input than any single angle. That's where this node stops being a curiosity and earns its place in a workflow. It's a metered cloud call like everything else in the pack, but it's the cheap, fast one - worth running before you spend a full generation credit on a model you might not even like.