SSpack_ComfyUI

SSpack for ComfyUI 🦅SS

Compact, UI-friendly nodes for ComfyUI. Everything ships with the 🦅SS prefix so you can spot the pack at a glance, and the selectors come with thumbnail helpers for faster browsing.

Highlights

• LoRA and checkpoint selectors with preview thumbnails (tree/grid/list) plus stack/apply helpers for random or indexed picks.
• Text utilities: prompt filtering/clean-up and weighted text formatting; JSON picker to pull a prompt from a list.
• Image helpers: SDXL/SD1.5/FLUX auto-resize, last-saved-image loader, and an overlay annotator for quick labels.
• Lightweight cache cleaner script for ComfyUI (optional).
• Pure Python/JS—drop it in and restart ComfyUI.

Installation (quick)

1. Drop this folder into ComfyUI/custom_nodes/SSpack_ComfyUI.
2. Make sure Python deps are available (most are bundled with ComfyUI):

   pip install opencv-python pillow requests
   

3. Restart ComfyUI. The JS helpers (betterLoras.js, betterCheckpoints.js) auto-load.

Quick start workflow

1. Add 🦅SS Checkpoint Selector (3/6/12) and pick models; thumbnails appear on right-click.
2. Add 🦅SS Checkpoint Loader and feed it the selector’s stack to load a model.
3. For LoRAs: use 🦅SS LoRA Selector → 🦅SS Random LoRA Applier → 🦅SS LoRA Stack Applier into your model/clip.
4. Clean prompts with 🦅SS Filter and weight text with 🦅SS Text Weight.
5. For images: run 🦅SS SDXL Auto-Resize before your pipeline, and use 🦅SS Last Saved Image to pull the newest output; annotate with 🦅SS Image Annotator if needed.

Nodes

• 🦅SS Checkpoint Selector (3/6/12) & Loader (ss_checkpoint_loader.py)
  Build a checkpoint stack (with thumbnails) and load by index or randomly.
• 🦅SS LoRA Selector / Random Applier / Stack Applier (ss_lora_stack.py)
  Preview LoRAs, gather trigger words from metadata/CivitAI, stack them, and apply with random or fixed strengths.
• 🦅SS SDXL Auto-Resize (ss_sdxl_auto_resize.py)
  Snap images to SD1.5/SDXL/FLUX-friendly sizes with crop/pad/letterbox options.
• 🦅SS Image Annotator (ss_image_node_annotator.py)
  Overlay text labels on images, optionally compositing another layer.
• 🦅SS Last Saved Image (ss_last_saved_image.py)
  Load the newest (or Nth newest) file from an output folder (supports subfolders and filters).
• 🦅SS Filter (ss_filter.py)
  Clean and slice prompts: tokenization, NG word removal/erasure, case conversion, de-duplication.
• 🦅SS Text Weight (ss_text_weight.py)
  Wrap text with weighted brackets; optional prefix/suffix; skip when weight is 1.0.
• 🦅SS JSON Output (ss_json_output.py)
  Pick one prompt from a JSON array (index, sequential, or seeded random).

Web UI extras

• betterLoras.js and betterCheckpoints.js add previews to the right-click menus (list/tree/grid view).
  Previews are pulled from files that match the model name (.preview.png/.jpg/.jpeg or the base image).
• loraInfo.js adds a “View info…” menu item for LoRAs (uses ComfyUI-Custom-Scripts dialog if present).

Tips & notes

• LoRA tag fetching uses CivitAI’s hash API (utils.py); allow network access for auto trigger words.
• opencv-python is required for the auto-resize node; install it if ComfyUI was built without OpenCV.
• Previews load from files next to your models (name.preview.png/jpg/jpeg or the base image).
• The cache cleaner (clear_comfyui_cache.py) is optional; run it from inside your ComfyUI folder if UI widgets act up.
• All nodes live under 🦅SS/* categories in the node browser.

Quick validation

To sanity-check the Python modules:

python -m py_compile ss_*.py

If you hit issues or want another node prefixed with 🦅SS, open an issue/PR.