CrasH Utils Custom Nodes

A collection of utility nodes and games for ComfyUI — visual LoRA loaders, image tools, SDXL helpers, LLM integration, and browser games for long generation queues.

---

Table of contents

• Node index
• Installation
• Loaders
  • Grid LoRA Loader
  • List LoRA Loader
  • Checkpoint Names
• Image
  • Resolution Picker
  • SDXL Resolution
  • SDXL Resolution To Dimensions
• Effects
  • Image Glitcher
  • Color Stylizer
• LLM
  • Query Local LLM
  • Extract Character Info
• Games
• Issues & contributions
• Credits
• License

---

Node index

All nodes appear under the CrasH Utils menu in ComfyUI, grouped by subcategory:

| Node | Category | What it does | |------|----------|--------------| | Grid LoRA Loader | Loaders | Multi-LoRA picker with thumbnail grid, trigger tags panel, and strength controls | | List LoRA Loader | Loaders | Same LoRA features in a compact vertical list layout | | Checkpoint Names | Loaders | Outputs the selected checkpoint name as a string | | Resolution Picker | Image | Visual picker for five common resolutions → width & height | | SDXL Resolution | Image | Dropdown of every SDXL-trained resolution | | SDXL Resolution To Dimensions | Image | Splits SDXL Resolution into separate width & height outputs | | Image Glitcher | Effects | Glitch art with chromatic aberration and scanlines | | Color Stylizer | Effects | Keep one color, desaturate everything else | | Query Local LLM | LLM | Send prompts to OpenAI-compatible APIs (local or cloud) | | Extract Character Info | LLM | Read embedded character metadata from a PNG file path | | Snake Game 🐍 | Games | Classic Nokia-style snake | | Tetris 🟦 | Games | Block-stacking puzzle | | Dino Game 🦖 | Games | Chrome-style endless runner | | Space Invaders 👾 | Games | Arcade shooter | | DOOM 👹 | Games | Full DOOM Shareware in the browser |

---

Installation

Recommended: install via ComfyUI Manager — search for CrasH Utils.

Manual:

cd ComfyUI/custom_nodes/
git clone https://github.com/chrish-slingshot/CrasHUtils.git

Restart ComfyUI. No extra dependencies — everything runs out of the box.

---

Loaders

Grid LoRA Loader

Visual multi-LoRA loader with a thumbnail grid browser, selected-LoRA panel, and per-tag trigger output. Inspired by rgthree's LoRA tooling.

Outputs: MODEL, CLIP, triggers

| Panel | Actions | |-------|---------| | Browse | Navigate folders, search, SFW toggle, double-click to add/remove LoRAs, right-click for info | | Selected | Single-click enable/disable, double-click remove, ⚙ for strength | | Trigger Tags | Toggle which tags are sent to the triggers output |

SFW toggle: When enabled, any folder named NSFW (case-insensitive) is hidden from browse and search — at any depth in your LoRA library. Turn it off to show those folders again.

Thumbnails: place an image next to each .safetensors file using the same base name — .png, .webp, .jpg, or .jpeg (.png is preferred when multiple exist). Right-click a Preview Image → Save as LoRA thumbnail to save from a generated image (saved as .png, overwrites existing).

Trigger words: companion .txt file beside the LoRA (same base name). One per line or comma-separated.

Panels: drag dividers between Browse / Selected / Tags to resize. Settings persist per node.

Typical workflow:

1. Connect model and clip
2. Browse and double-click LoRAs to add them
3. Tune strength with ⚙; disable any you want to keep but not apply
4. Toggle trigger tags for your prompt
5. Connect MODEL, CLIP, and triggers downstream

---

List LoRA Loader

Same backend and outputs as the grid loader, in a vertical list layout. Better when you want names and trigger tags readable at a glance rather than a thumbnail wall.

Browse: folder navigation, search, SFW toggle, double-click add/remove, right-click info. Trigger tags appear inline on each row.

Selected: same enable/disable, strength, and remove controls as the grid loader.

Everything else — thumbnails, .txt triggers, Preview Image save, MODEL / CLIP / triggers outputs — works the same as Grid LoRA Loader.

---

Checkpoint Names

Pass-through node that outputs the selected checkpoint filename as a STRING. Useful when a workflow needs the checkpoint name as text (logging, filenames, LLM prompts, etc.).

---

Image

Resolution Picker

Click a resolution to output Width and Height integers. Each option shows a small aspect-ratio preview.

| Resolution | Orientation | |------------|-------------| | 1440 × 868 | Landscape | | 1440 × 1040 | Landscape | | 1280 × 1280 | Square | | 1040 × 1440 | Portrait | | 868 × 1440 | Portrait |

Connect outputs to Empty Latent Image or anywhere you need dimensions. For the full list of SDXL-trained sizes, use SDXL Resolution instead.

---

SDXL Resolution

Dropdown of all official SDXL-trained resolutions (1024×1024, 1152×896, 1216×832, etc.) with aspect ratio labels. SDXL gives best results at these specific sizes.

---

SDXL Resolution To Dimensions

Splits the SDXL Resolution string into separate width and height outputs — one connection instead of two downstream.

---

Effects

Image Glitcher

Authentic glitch art with chromatic aberration and CRT scanlines. Based on the HTML image glitcher by Felix Turner.

| Parameter | Range | Effect | |-----------|-------|--------| | glitchiness | 0–100 | Corruption and chromatic aberration intensity | | brightness | 0–100 | Brightens the image; useful with scanlines | | scanlines | toggle | CRT TV scanlines |

---

Color Stylizer

Pick one color to preserve; everything else is desaturated.

| Parameter | Effect | |-----------|--------| | target_color | Color to keep (hex or RGB) | | tolerance | How closely colors must match | | desaturation | How much to grey out non-matching areas |

---

LLM

Query Local LLM

Send prompts to any OpenAI-compatible API and get text back in your workflow — prompt expansion, descriptions, or general text tasks.

Works with OpenAI, LM Studio, Ollama, text-generation-webui, and other compatible endpoints. Configure URL, system message, context length, and seed.

---

Extract Character Info

Reads embedded character metadata from a PNG file path (base64-encoded text chunks) and outputs it as a string. Useful for workflows that store character data inside generated images.

---

Games

Browser games that run inside ComfyUI nodes — no installs, no external deps. Handy while waiting on long generations.

Controls (all games):

• Click the game canvas to capture keyboard input
• Click outside the node to release it

---

🐍 Snake Game

Classic Nokia-style snake. Arrow keys or WASD; Space to pause/restart.

<img width="356" height="395" alt="Snake Game" src="https://github.com/user-attachments/assets/aedee4ce-15ff-433c-a877-35ff7e61da6e" />

---

🧱 Tetris

Arrow Left/Right move, Down soft drop, Up rotate, Space hard drop. Score, levels, next-piece preview.

---

🦖 Dino Game

Chrome offline runner. Space or Up to jump. Increasing difficulty and high score.

---

👾 Space Invaders

Arrow keys move, Space shoots. Wave-based arcade action.

---

🔫 DOOM

Full DOOM Shareware v1.9 in the browser via js-dos.

Automatic setup (recommended): add the node — it downloads ~2MB on first run from several mirrors, then launches.

Manual setup (if auto-download fails):

1. Download DOOM Shareware v1.9 from Doomworld or Archive.org
2. Place DOOM.EXE and DOOM1.WAD in ComfyUI/custom_nodes/CrasHUtils/doom/
3. Restart ComfyUI

Troubleshooting: click outside the node if keyboard sticks; check the browser console if download fails.

---

Issues & contributions

Found a bug or want a feature? Open an issue on GitHub.

Pull requests welcome — especially new games or utility nodes.

---

Credits

• Image Glitcher — Felix Turner
• LoRA loaders — UI inspired by rgthree's LoRA tooling
• DOOM — js-dos emulator; Shareware © id Software

---

License

MIT License — use, modify, and distribute freely.

DOOM Shareware is distributed under id Software's original shareware license.