Before you start
A version-aware workflow for selecting a maintained wrapper, reading its exact CUDA and VRAM requirements, importing an example graph, and validating each output stage.
Prerequisites
- A working ComfyUI install
- Git
- Enough disk space for checkpoints
- A GPU that matches the selected wrapper's requirements
Step by step
Choose one maintained wrapper
Inspect its latest release notes, supported operating systems, dependencies, and open issues. Do not combine nodes from unrelated wrappers until the baseline works.
Record versions
Save the ComfyUI commit, custom-node commit, checkpoint identifier, PyTorch build, CUDA version, and attention backend.
Import the smallest official example
Use the wrapper's own workflow and sample input before adding background removal, multi-view, smoothing, or export nodes.
Validate each stage
Check shape output before texturing, then inspect the final GLB for triangle count, material channels, and alpha behavior.
The workflow is reproducible after a clean ComfyUI restart.
Known trade-offs
- ComfyUI wrappers are community projects and can lag behind upstream changes
- A workflow file does not install missing custom nodes or checkpoints
- Quantized or offloaded workflows can change quality, memory use, and compatibility