WorkflowVérifiéExterne

Extremely Simplified SDXL Workflow APP — v2.0

Changed default highres fixes.

TâcheMise à l'échelleFamilleIllustriousVersionRelease 2923387 — Extremely simplified SDXL (APP).jsonEstimé VRAMNot yet characterizedUnratedClasse

Profil d'exécution

Remaining hardware or download-size estimates are not available yet.

Exigences

BSS_GAP_SLIDER_V50.safetensors
Non résolu
LoRA · Unknown
covering privates_XL_V1.0.safetensors
Non résolu
LoRA · Unknown
Non résolu
Object Detector · Unknown
hand_yolov8s.pt
Non résolu
Object Detector · Unknown
hires_fix_v10_000000100.safetensors
Non résolu
LoRA · Unknown
novaAnimeXL_ilV180.safetensors
Trouvé
Checkpoint · 6.46 GB · SAFETENSOR · FP16 (unverified) · Unknown
sam_vit_b_01ec64.pth
Dépôt identifié
Segmentation · Unknown

Commentaire généré par l’IA

Explication générée par l’IA à partir des détails de la source et de la configuration. Les suggestions sont clairement signalées.

Use Extremely Simplified SDXL Workflow APP — v2.0 in ComfyUI to generate an image from a text prompt. It takes prompt, model, sampler, resolution, and LoRA settings and produces a final image through two-pass generation with an upscale and refinement pass.

After the first generation pass, a second upscale and refinement pass uses SAM and a bounding-box detector for subject detection and detailing.
GPU memory cleanup nodes sit between passes, and the first-pass preview is hidden so only the final result is shown.

APP Mode exposes prompts, model choice, sampler settings, resolution, and LoRA controls. The Resolution Selector includes SDXL aspect ratios and megapixel scaling, and the LoRA Loader Stack has four slots.

Before loading, install rgthree-comfy, ComfyUI Easy Use, and ComfyUI Impact Pack through ComfyUI Manager.
The description names Eyeful_v2-Paired.pt as the default bounding-box detector and says it is not included with Impact Pack. Place it in ComfyUI/models/ultralytics/bbox/ and restart ComfyUI; create the folder if needed.

Required node packs and their listed revisions: comfyui-easy-use (ec4ca6717f539a8b8c48fa88645045846ca47669), comfyui-impact-pack (cd34cfdd638a996fd011be1b490c70a7e0fb505f), comfyui-impact-subpack (1.3.2), comfyui-kjnodes (01d9fa9c983273532cacdf9532c74a93c7dc86d2), and rgthree-com
The workflow lists these model files: BSS_GAP_SLIDER_V50.safetensors, covering privates_XL_V1.0.safetensors, face_yolov8m.pt, hand_yolov8s.pt, hires_fix_v10_000000100.safetensors, novaAnimeXL_ilV180.safetensors, and sam_vit_b_01ec64.pth.

Defaults are 30 first-pass steps, 15 hires-fix steps, CFG 4.5, Euler sampler, Simple scheduler, and 1344×768 resolution at 1 megapixel; CLIPSetLastLayer is set to -2.
The rgthree KSampler Config applies one set of steps, CFG, sampler, and scheduler to both passes. Linear Mode lets you change settings and resolution without editing the node layout.
As planning guidance, the description labels 8 GB VRAM suitable, 12 GB its sweet spot, and 16 GB or more smooth.
Suggestion · non vérifié