Object Voxel Remesh

Meshes built or heavily modified through an automated pipeline - which is exactly what Avatar Graph's mesh-layer generation is - often end up with messy topology: uneven face density, overlapping geometry, or a surface that's technically valid but a pain to work with downstream. Object Voxel Remesh fixes that by throwing the mesh's existing topology away entirely and rebuilding it from a voxel representation, which gives you a clean, uniformly-dense surface regardless of how tangled the input was.

How it works

It's Blender's Voxel Remesh operator, wrapped as-is. The mechanism: Blender converts your mesh into a voxel grid (think of it like a 3D pixel grid filling the object's volume), then generates a brand-new mesh surface from that grid at a uniform resolution. That's the key trade-off to understand - voxel remeshing doesn't preserve your original topology or UVs at all, it produces an entirely new mesh that just happens to occupy the same shape. It's fast and reliable for messy or organic geometry, but if you need to keep existing UV layouts or vertex groups intact, you'll want Object Data Transfer afterward to carry that data onto the new mesh, or reach for Object Quadriflow Remesh instead if you specifically need clean quad topology rather than voxel density.

The inputs and outputs that matter

There's exactly one input:

• BPY_OBJ - the object to remesh. Chain this in after whatever step left you with geometry you want cleaned up - a join, a boolean, a rough auto-generated mesh from the segmentation pipeline.

The schema exposes no resolution/detail parameters beyond the object itself, so this node runs with Blender's current voxel-size setting on the object rather than exposing it as a per-call knob - if you need a specific detail level, set it on the object in Blender first, or reach for Object Quadriflow Remesh's explicit ratio/edge-length/face-count controls if you need that kind of fine control from within the graph.

Output is a single BPY_OBJ, the remeshed object, ready to chain into the next step.

How to install it

ComfyUI Manager, search "Avatar Graph" and install, or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/avatechai/avatar-graph-comfyui

then pip install -r requirements.txt and restart. Needs Blender's bpy module, pinned to Python 3.10.x - on macOS/Linux, use a dedicated conda environment for this pack rather than an existing ComfyUI setup running a newer Python. Windows users can use Avatech's prebuilt Python-3.10 ComfyUI zip instead and install the pack via Manager on top of it.

Common issues & troubleshooting

If the node won't load, that's the pack's Python version pin, not this operator - check ComfyUI's startup console for an import error and confirm requirements.txt installed cleanly in a Python 3.10.x environment.

If the remesh runs but the result looks unexpectedly coarse, dense, or the wrong scale, that comes back to the voxel-size setting mentioned above rather than anything exposed on this node - since there's no resolution input here, the result depends on whatever voxel size was last set on the object, which can be surprising if you expected a node-level parameter to control it. And remember: because voxel remeshing discards the original UVs, any texture that was correctly mapped before this node will need to be re-unwrapped or carried over with Object Data Transfer afterward, not treated as still intact.