SDXL Base + Refiner text-to-image

Text-to-image with Stable Diffusion XL base model, optionally followed by the SDXL refiner for detail.

Estimated VRAM

~6.94 GB — ⚠️ ESTIMATE (unverified real-world usage). Estimate, not a measured figure: weight-size floor = sd_xl_base_1.0.safetensors (6.94 GB, sizes verified on Hugging Face/GitHub); base and refiner load one at a time. Activations add more; ComfyUI offloading can reduce it. Source, checked 2026-10-13.

Models required

• sd_xl_base_1.0.safetensors — 6.938 GB ✅ verified 2026-10-13 (listing) · sha256 31e35c80fc4829d14f90153f4c74cd59c90b779f6afe05a74cd6120b893f7e5b → models/checkpoints
• sd_xl_refiner_1.0.safetensors — 6.076 GB ✅ verified 2026-10-13 (listing) · sha256 7440042bbdc8a24813002c09b6b69b64dc90fded4472613437b7f55f9b7d9c5f → models/checkpoints

Tested GPUs

No verified test reports yet.

Nodes

CheckpointLoaderSimple CLIPTextEncode EmptyLatentImage KSamplerAdvanced VAEDecode SaveImage

Common errors & fixes

torch.OutOfMemoryError: CUDA out of memory

Launch ComfyUI with --lowvram, reduce batch size, or skip the refiner pass. ⚠️ unverified: Fix is general guidance; not confirmed against the cited source.

Low-VRAM tips

⚠️ General guidance, not individually verified.

• Use --lowvram launch flag
• Skip the refiner
• Generate at 1024x1024 with batch size 1

Source & credit

Workflow by comfyanonymous (ComfyUI_examples). Get it from the original: https://comfyanonymous.github.io/ComfyUI_examples/sdxl/. License: Workflow embedded in example image; models under CreativeML Open RAIL++-M.