Rodin - Gen2 Generate

Before the 2.5 lineup took over this pack's front page, Rodin - Gen2 Generate was the upgrade. It's image-to-3D on Rodin's Gen 2 model - the generation tier that followed the original Regular/Smooth/Sketch trio - and it's still there in the README under "Gen 2 Nodes," quietly doing the job. If your workflow predates the 2.5 nodes, or you just want a model generation call with fewer dropdowns to second-guess, this is the one.

What it is

Feed it images (one or more), it reconstructs a textured mesh with PBR or shaded materials, and it downloads the result into ComfyUI/output/<timestamp>/. Everything runs on Hyper3D's servers via the Rodin API; the node's job is submit, poll, download. The images get PNG-encoded and uploaded with your key as a Bearer token, the API returns a job UUID, and the node polls every five seconds until the mesh is ready.

The inputs that matter

The required set is compact: images, api_key, seed_, Material_Type, Polygon_count, and TAPose. You'll actually touch two of them:

• Polygon_count - ten presets from 4K-Quad to 1M-Triangle, default 500K-Triangle. Note this one includes both Quad and Triangle options up to a million. For game-engine work, the quad presets are your friend; for raw fidelity, go triangles.
• TAPose - the Gen-2 addition. Flip it on to ask for a character that comes back posed/rigged rather than a neutral T-pose. Pointless for objects, useful for characters, off by default.
• Material_Type - PBR, Shaded, All, or None; PBR is the default and the one you want for anything that'll be lit in a real engine.

Optional bbox and height_cm pin the physical dimensions - same as the other generation nodes, and wired the same way (the BBox Controlnet node feeds bbox if you want number boxes instead of typing).

What comes out

One output: model_path, a string path to the downloaded GLB. Plug it into the core Preview3D node to look before you commit, exactly as the pack's example workflow does.

Where Gen2 sits vs. the rest

The README groups the pack into Gen 1/1.5 (Regular, Smooth, Sketch), Gen 2 (this node), and Gen 2.5 (the Fast/Regular/ExtremeHigh family). Rodin's model line moved on, and the 2.5 nodes are newer and generally better - the Fast tiers beat Gen2 on iteration speed, and 2.5 Regular/ExtremeHigh beat it on quality ceiling. So why use Gen2? Two honest reasons: it's simpler (fewer knobs, no tier dropdown), and if you already have Gen-2-era results you're happy with, you don't need to change anything. It's the conservative choice, and "conservative" is a feature when you're paying per job.

Installing it

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

or install "ComfyUI-Rodin" from ComfyUI Manager and restart. Light dependencies - aiohttp for the API calls, and asyncio which is already in Python (the requirements line is a free no-op). No model downloads.

Troubleshooting

The usual pack family: MISS_API_KEY = key node not wired up; auth failures at generation time = wrong key or no credits; 30-minute polling timeout on big jobs = check the Rodin dashboard because the mesh may be done even though the download didn't land. And if a Gen2 job feels slower than you remember, that's Rodin's servers having a day - not your machine, which was never doing the work in the first place.