A locally hosted, privacy-focused, uncensored roleplay chat application. Run powerful AI waifus directly on your NVIDIA GPU with a modern, visual-novel style interface.
- 100% Offline & Private: No data leaves your computer.
- Uncensored Roleplay: Powered by the L3-8B-Stheno-v3.2 model, designed for creative and complex storytelling without refusals.
- Immersive Visuals:
- Custom Avatars: Upload your own character images (Neutral, Happy, Angry, etc.).
- Dynamic Backgrounds: Set unique wallpapers for each character.
- Glassmorphism UI: A sleek, neon-styled interface.
- Multi-Character System:
- Create and edit unlimited characters via the built-in Editor.
- Switch between them instantly.
- Roleplay Tools:
- User Persona: Define your own name and appearance so the AI knows who you are.
- Regenerate (Reroll): Don't like a reply? Reroll it instantly.
- Edit Messages: Fix typos or steer the story by editing any message (yours or hers).
- Infinite Chat: Automatic context sliding lets you chat forever without crashing.
- Persistent Memory: Save and load your chat sessions at any time.
The app has been upgraded with advanced roleplay features to make your characters feel truly alive!
Characters now react visually to the conversation!
- Mood Awareness: The AI automatically detects if it's happy, sad, angry, or blushing.
- Sprite Switching: The avatar changes instantly to match the mood (e.g., smiling when complimented, frowning when insulted).
- Customizable: Map your own image files to specific emotions in the Character Editor.
Solve the "goldfish memory" problem with the new Diary System.
- Episodic Memory: The AI writes diary entries about your adventures from its perspective.
- Context Injection: Recent diary entries are automatically fed into new chat sessions, so the character remembers what you did yesterday.
- End of Day: Click "End Day & Write Diary" to generate a summary of your session.
Build a deep, consistent world.
- Keyword Triggering: Define keywords (e.g., "The Crystal Kingdom", "Excalibur").
- Smart Injection: When you mention a keyword, the app secretly injects the description into the AI's mind.
- No Confusion: Keeps the AI on track with complex lore without using up all its memory.
- Text-to-Speech (TTS): The waifu now speaks to you!
- High Quality: Uses natural-sounding Neural voices (like Aria, Guy, Jenny).
- Customizable: Adjust the Pitch (Hz) and Speed (%) to create unique voices for each character.
- Character Card Import:
- Easily import characters from other sites (Chub.ai, SillyTavern) by dragging and dropping V2 Character Cards (PNG) or JSON files.
The app is now a lightweight Life Sim!
- Status System:
- Affection: Being nice increases affection, insults decrease it.
- Energy: Talking drains energy. If it hits 0, they might get tired.
- Locations:
- Travel to different places (Park, Cafe, Beach, etc.).
- The AI knows where you are and changes the context of the conversation.
- Time & Routine:
- Real-time clock awareness.
- "Wait 1 Hour": Button to advance time and restore Energy.
Focusing on visuals and atmosphere to pull you deeper into the world.
- Dynamic Backgrounds:
- Map different images to different locations (e.g., a cozy room for "Home", a sunny park for "Park").
- Backgrounds switch instantly when you travel.
- True Time System:
- A 24-hour clock tracks the day (00:00 - 23:00).
- Day/Night Cycle: The app's visual atmosphere changes automatically based on the time (Morning glow, Day clear, Sunset orange, Night dark).
- Enhanced Editor:
- New tools to easily upload and assign backgrounds to specific locations.
Refining the user experience with better tools to manage your AI's data.
- Diary Editor:
- You can now Edit or Delete specific diary entries to fix incorrect memories.
- Character Export:
- Easily back up your character (including images and memories) to a ZIP file for safekeeping or sharing.
- "Brain Scan" Debugger:
- A new tool for power users to see exactly what the AI is "thinking" (the raw prompt sent to the model).
- New Interactions:
- Give Gifts: Give items like Coffee or Flowers to boost affection.
- "Let Her Speak": A button to let the AI continue talking without you interrupting.
The first step towards a fully multimodal AI.
- Visual Perception:
- "Show her something": Upload an image in the chat.
- Image Captioning: The app uses the BLIP AI model to "see" the image and convert it into a description (e.g., "User showed an image: a small cat sleeping on a laptop").
- Reaction: The character reacts to what she sees in character!
Now you can talk to her directly!
- Voice Input:
- Uses OpenAI Whisper (Local Base Model) for high-accuracy speech-to-text.
- Click the microphone icon 🎤, record your message, and it will be transcribed instantly.
- Works completely offline.
The AI now has a subconscious.
- Dream System:
- When you click "End Day (Sleep)", the AI reflects on the day's diary entry and generates a dream.
- Dreams are abstract, emotional, and sometimes surreal.
- Dream Journal: Read her past dreams in the "Memory" sidebar.
She lives her own life when you're not around.
- Offline Simulation:
- The app calculates how much time has passed since you last chatted.
- Activities: She might read a book, clean her room, sleep, or practice cooking while you are away.
- Stat Changes: Being away too long might drain her energy (if she stayed up waiting) or restore it (if she slept).
- Welcome Back: When you log in, you'll see a toast notification telling you what she did, and she will know it too!
- NVIDIA GPU: You need a card with at least 6GB VRAM (8GB recommended).
- CUDA Toolkit: Install the NVIDIA CUDA Toolkit to enable GPU acceleration.
- Python: Install Python 3.10 or newer.
- Important: Check the box "Add Python to PATH" during installation.
- Double-click
install.bat.- This will create a virtual environment, install all dependencies (configured for NVIDIA GPUs), and download the 5GB AI model.
- Once finished, double-click
run.batto start the app.
- Open the sidebar and expand "🛠️ Character Editor".
- Select "Create New".
- Fill in the Name, Description, and Scenario.
- Upload Images:
- Upload an Avatar image. Copy the filename it gives you into the JSON map (e.g.,
"happy": "my_image.png"). - Upload a Background image and paste the filename into the "Background Image" field.
- Upload an Avatar image. Copy the filename it gives you into the JSON map (e.g.,
- Add World Info: Use the Lorebook section to add important keywords and their descriptions.
- Click "Create Character".
- Type your message in the bottom box.
- Regenerate: Click the 🔄 button to redo the last AI response.
- Edit: Click the ✏️ pencil icon next to any message to change the text.
- End Day: When you're done, scroll down to the "Diary" section and click "End Day & Write Diary" to save the memory.
- Go to the "💾 Memory" section in the sidebar.
- Type a name and click "Save Current Session".
- To resume later, select a file from the list and click "Load".
"CUDA not found" / Slow Generation:
- Ensure you installed the NVIDIA CUDA Toolkit.
- Re-run
install.batto force-reinstall the GPU-optimized libraries.
"Context Limit Exceeded":
- The app automatically handles this by trimming old messages. You shouldn't see this error.
App crashes on startup:
- Make sure you have enough free RAM (System Memory) and VRAM. The model requires about 5.5GB VRAM.
app.py: Main UI logic (Streamlit).brain.py: AI inference engine (Llama-cpp).character_manager.py: Handles saving/loading characters and chats.characters/: Folder containing all your waifu data and images.models/: Where the GGUF model file lives.
MIT License. Feel free to modify and share!
