Tripo: Rig model

A static mesh is fine for a render or a 3D print. The moment you want the character to do something - walk, run, gesture - you need a rig: a skeleton with weights that tells the surface how to move when the bones do. That's what TripoAnimateRigNode is for. Give it a model and Tripo's cloud builds a skeleton and skinning for it, so the mesh is ready for animation.

This is the node that unlocks the rest of the animation half of this pack. Without a rig, you can't retarget. With one, you can feed the result into TripoAnimateRetargetNode and drop a walk cycle or a slash animation onto the character straight from ComfyUI.

What it takes

Like every Tripo node, it's a cloud call off the model_info handle - wire in the model you want rigged and the pack does the rest. The inputs that matter:

• model_version - v1.0 or the default v2.0-20250506 rigging model. Newer, obviously.
• out_format - glb or fbx. FBX is what game engines and DCCs like Maya/3ds Max expect; GLB is fine for quick inspection and web use.
• spec - the rig convention: tripo (Tripo's own skeleton) or mixamo (Adobe Mixamo's bone naming, which matters if you plan to move the character into Mixamo's animation library or a pipeline built around it).

Under the hood there's a bit of smart behavior worth knowing: before rigging, the node calls Tripo's riggability check. If the model can't be rigged - a non-anthropomorphic blob, a chair, a rock - it raises an error instead of burning a task on something doomed. It also auto-picks the rig type from that check, so you don't have to tell it whether you're rigging a humanoid or a quadruped.

The outputs are the standard pair: model_file (path to the rigged GLB/FBX in ComfyUI/output/) and model_info for the next step - and the next step for a rigged model is almost always retargeting.

Install

Same pack, same story:

cd ComfyUI/custom_nodes
git clone https://github.com/VAST-AI-Research/ComfyUI-Tripo
cd ComfyUI-Tripo && pip install -r requirements.txt

Or grab it as "Tripo for ComfyUI" in ComfyUI Manager. Key from platform.tripo3d.ai, via env var, config.json, or the apikey field on the generation node.

The honest part

This is where the "cloud AI rigging" tradeoffs get concrete. A rig from a generative service is impressive - it handles humanoids and animals without you touching a single bone - but it's not a hand-built character rig. For a game-ready character you'll likely re-target or clean the weights in Blender or your engine. And rigging a character costs more credits than generating a draft, so don't rig your placeholder cubes. The real sweet spot is exactly what this pack advertises: draft → refine → rig → retarget, one graph, no DCC needed until the very end.