A step-by-step guide to running three production-ready generative AI workflows in ComfyUI using NVIDIA RTX GPUs. The workflows cover: (1) decomposing a photo into separate depth-based layers with alpha masks, (2) removing and inpainting unwanted objects from photos, and (3) converting a single photo into a textured 3D GLB model (Windows only). All workflows come from the NVIDIA GenAI Creator Toolkit, install via a single script, and run entirely locally without cloud dependencies. Hardware requirements are significant — 24–32 GB VRAM minimum — and initial model downloads exceed 50 GB.
Table of contents
What you’ll accomplishWhat to know before startingPrerequisitesTime and riskDetailed instructionsWhat’s NextTroubleshootingSort: