ComfyUI_ScarlotSoft

One dropdown, 23 named aspect ratios — the ScarlotSoft Aspect Ratio node

A one-node face detailer with optional upscaling — the ScarlotSoft AutoDetailer

Checkpoint loader with CLIP skip baked in — ScarlotSoft Checkpoint + CLIP Skip

A CLIP encoder that assembles the prompt for you — ScarlotSoft Modular CLIP Encode

Two text boxes and a toggle — the ScarlotSoft Concatenate node

Two conditioning inputs, one combined output — ScarlotSoft Combine Conditioning

Feed prompts from a text file, one line per run — ScarlotSoft Easy Line From File

Masked inpainting conditioning without the plumbing — ScarlotSoft Easy Mask Conditioning

Slice frames out of a batch — ScarlotSoft Image/Mask From Batch

Point at objects with text, get masks — the ScarlotSoft Grounded-SAM node

A/B two images on the canvas — the ScarlotSoft Image Compare node

See your mask as a red overlay — the ScarlotSoft Image/Mask Preview node

One config panel for your sampler and your detailer — ScarlotSoft KSampler Config

Load an image with a friendlier upload button — the ScarlotSoft Load Image node

A LoRA loader with rows, toggles, and trigger-tag reading — ScarlotSoft Lora Loader

Chroma-key style masks from a color picker — the ScarlotSoft Mask From Color node

Read the workflow out of a PNG — the ScarlotSoft Metadata Reader node

Prompts from files, LoRAs auto-loaded from the text — ScarlotSoft Prompt Builder + AutoLoader

Every file gets its own output — ScarlotSoft Prompt Builder Pro

Pick megapixels, get dimensions and a latent — the ScarlotSoft Resolution node

A save node that names your files for you — checkpoint, seed, and date baked into the filename

Twenty inputs, one string — the prompt joiner that grows its own sockets

The string joiner that speaks GroundingDINO — dots are the whole point

A single-line string on a leash — and that's the whole job

Flip between two of anything — no rewiring, no second workflow

A multi-way switch with toggle pills instead of a number box

See what that string actually says, mid-workflow

Upscale, then re-diffuse at low denoise — a whole hi-res fix in one node

Turn any image into Danbooru tags, right inside the graph

YOLO boxes in, clean masks out — with optional SAM surgery