Personal upscaler. With the default settings, 720p images (realistic) with a good amount of detail work best. For anime and other content, lower settings may suffice. However, you are free to change the parameters.
Runtime profile
Source description
Personal upscaler. With the default settings, 720p images (realistic) with a good amount of detail work best. For anime and other content, lower settings may suffice. However, you are free to change the parameters.
Adjust the First Stage Denoise and Second Stage Denoise nodes to control how much the model re-imagines details. Higher values add more detail but can alter the image more. A good range is 0.4-0.7.
Used resources: https://civitai.com/models/277058/epicrealism-xl
AI-generated commentary
AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.
Ultra Upscaler is a workflow for upscaling images: give it an image and it produces an upscaled image. Its defaults are described as best suited to detailed, realistic 720p images; anime and other content may suit lower settings.
Input: an image. Output: an upscaled image.
It uses SDXL 1.0 as its base model.
The named model files are bbox/face_yolov8m.pt, controlnet-tile-sdxl-1.0.safetensors, epicrealismXL_vxviiCrystalclear.safetensors, and sdxl_lightning_8step_lora.safetensors.
The named node packs are comfyui_essentials, comfyui-impact-pack, comfyui-impact-subpack, comfyui-kjnodes, and comfyui-wd14-tagger.
Adjust First Stage Denoise and Second Stage Denoise to control how much the model re-imagines details.
Higher values add more detail but can alter the image more; the described range is 0.4–0.7.
Before use, verify that bbox/face_yolov8m.pt and the other named model files and node packs are available in your workflow environment.
Suggestion · not verified
If you increase either denoise value, inspect the result because higher values can alter the image more.
Suggestion · not verified
Does this need to be edited?
Sign in to send an edit request.
Sources
1 sourceSource excerpts
1 excerptSource context: 550 downloads · Type Workflows · Base model SDXL 1.0
Estimated VRAM requirement
Estimate unavailable
9.16 GB across 3 of 4 model files. Model file total + 25% loading overhead + 2 GB execution buffer, rounded up.
Requirements
9 requirementsObject Detector · Hugging Face · bbox/face_yolov8m.pt
ControlNet · 2.33 GB · SAFETENSORS · fp16 · full · Unknown
Checkpoint · 6.46 GB · SAFETENSORS · fp16 · pruned · Unknown
LoRA · 376 MB · SAFETENSORS · Unknown
Node pack · Registry
Node pack · Registry
Node pack · Registry
Node pack · Registry
Node pack · Registry