Object Call Asset Generate Preview

Of all the raw Blender API wrappers buried inside Avatar Graph (avatechai/avatar-graph-comfyui), ObjectCall_AssetGeneratePreview is one of the more clearly "it exists because everything got wrapped" entries. Avatar Graph's real job is rigging a still character portrait into a live, animatable 2D avatar with blinking and lipsync; this node has nothing to do with that. It's here because the pack's node generator turned essentially every method on bpy.types.Object (and the ID base class it inherits from) into a ComfyUI node, useful or not.

How it works

It maps to ID.asset_generate_preview() - Blender's own docs describe it simply as generating the preview image for an asset. That preview is the little thumbnail icon Blender shows in its Asset Browser panel, computed automatically from the datablock's current appearance. It only does something meaningful once the datablock has already been marked as an asset (see ObjectCall_AssetMark, the companion node) - without that, there's no asset metadata for a preview to attach to.

Why you'd reach for it

Realistically, you probably wouldn't, at least not for anything Avatar Graph's documented workflow does. Blender's Asset Browser is a Blender-UI feature for organizing reusable objects/materials/collections across projects - and Avatar Graph runs Blender headless, with no UI ever shown. Even if you generate a preview, the graph gives you no way to view or export that thumbnail image; it just sits on the datablock. Where this could matter is if you're repurposing Avatar Graph's Blender plumbing for a genuinely different pipeline - say, building a library of rigged avatar meshes you intend to later open and browse inside an actual Blender install, where Asset Browser thumbnails would then be useful.

Inputs and outputs

One optional BPY_OBJ in, one BPY_OBJ out, and nothing else - no way to configure preview size, render angle, or anything else Blender's asset system supports. The object passes straight through; the generated preview image isn't exposed as an output.

Installing it

Avatar Graph brings in Blender's full bpy module, which is strict about Python version - expect a real install, not a one-click drop-in. On Windows, use the author's bundled Python 3.10 environment (linked in their README), unzipped into your ComfyUI directory, launched via the included .bat, then install "Avatar Graph" from ComfyUI Manager. On macOS/Linux, pin your ComfyUI environment to Python 3.10.x exactly - conda is the simplest way (conda create --name comfyui python=3.10). Then cd ComfyUI/custom_nodes, git clone https://github.com/avatechai/avatar-graph-comfyui.git (their README's manual-install section actually references a stale org, avatechgg - the real repo is under avatechai). cd avatar-graph-comfyui && python -m pip install -r requirements.txt, then restart with python main.py --enable-cors-header (Mac: also add --force-fp16). That flag opens CORS so the hosted viewer at editor.avatech.ai can talk to your local ComfyUI; skip it if you're exporting .glb/.gltf from Avatar Main Output instead.

Common issues

Python-version mismatch is still the number one blocker for the whole pack - anything other than 3.10.x and bpy won't install. Specific to this node: calling it before ObjectCall_AssetMark does effectively nothing useful, since there's no asset metadata yet for a preview to attach to. And because this is one of the least-relevant nodes in the batch to Avatar Graph's actual avatar-rigging use case, don't spend much time trying to make it fit your workflow - it's a completeness artifact of wrapping the whole bpy.types.Object/ID API, not a documented feature of the pack.