Summary
Take full control of your JanitorAI formatting. This script is the ultimate tool for immersion. It automatically cleans up AI responses, formats narration and dialogue exactly how you want it, and removes annoying "thinking" text—all with a single click.
Now features a built-in Settings Menu (UI) so you can customize everything without touching a single line of code.
🌟 New in v9.0: Full Customization
No more one-size-fits-all. Click the Gear Icon (⚙️) to choose your preferred Roleplay style:
- Classic:
*Narration*"Dialogue" - Bold Narration:
**Narration**"Dialogue" - Bold Dialogue:
Narration"Dialogue" - Hybrid:
*Narration*"Dialogue" (Great for DeepSeek/coding models) - Plain Text: No formatting at all (Best for accessibility/TTS)
Key Features
🎨 Interface & Usability
- Draggable Buttons: Don't like the button placement? Just press, hold, and drag the ✏️ icon anywhere on the screen. It saves the position forever.
- Settings Menu: A visual menu to toggle specific features on or off instantly.
- Universal Compatibility: Works flawlessly on Mobile (Android/iOS), PC, and is now fully compatible with Microsoft Edge.
- Keyboard-Aware: On mobile, the buttons automatically hide when you type so they never block your view.
📝 Smart Formatting
- Dialogue Protection: Intelligently detects speech in quotes (
"...") or backticks (...) and preserves it while formatting the narration around it. - Smart Separators: Automatically detects scene separators (like
---). It prevents them from being italicized/bolded and fixes them if the AI formatted them wrong (*---*→---). - Noise Reduction: Automatically removes "garbage" lines that contain nothing but stray quotation marks or asterisks.
🧹 Advanced Cleaning
- Think Tag Remover: Instantly deletes the AI's internal monologue (
<think>...) and<system>tags so you stay immersed in the story. - General Tag Cleaner: Removes generic XML-like tags (e.g.,
<AI-response>,<segment>) but keeps the text inside them, ensuring you don't lose actual content. - Prompt Leak Fix: Detects and chops off leaked system prompts (e.g., when the bot starts a message with "The user said...").
How to Use
- Install the script.
- After the AI sends a message, you will see a Pencil (✏️) and a Gear (⚙️) button on your screen.
- To Format: Click the Pencil. The script will edit, format, clean, and save the message in under a second.
- To Configure: Click the Gear to open the menu. Choose your style (Italic, Bold, etc.) and toggle cleaning options.
- To Move: Press and hold the Pencil button to drag the interface to a spot that suits you.
What Problem Does This Solve?
LLMs are inconsistent. Sometimes they use bold, sometimes italics, sometimes nothing. Sometimes they reveal their hidden <think> process. This script enforces consistency. You get a clean, perfectly readable, book-like roleplay experience every single time, regardless of which AI model you are using.
Screenshots
Screenshots show the Before/After comparison for basic formatting.
