MateChat

A MelonLoader mod that enables AI chat and voice interactions with your Desktop Mate! ♪

MateChat now supports multiple AI language models and text-to-speech capabilities, allowing for even more immersive interactions with your desktop companion~ ✧

✨ Features

• Chat with your Desktop Mate using various AI providers
• Text-to-Speech support powered by GPT-SoVITS
• Cute and minimal interface matching the game's aesthetic
• Highly configurable settings for both chat and voice
• Quick access through mate's menu or keyboard shortcut

🎀 Installation

More detailed instructions (step by step) can be found in the Wiki.

1. Install Desktop Mate & MelonLoader

   • Download the latest Desktop.Mate.zip
   • Extract the contents to match your Steam installation directory
   • The extracted folder should contain:
     • Mods/ directory with the MateChat dll
     • UserLibs/ with required libraries

2. Configure MateChat

   • Start Desktop Mate
   • Right-click and select MateChat from the menu
   • Configure your settings in UserData/MateChat.cfg

3. Setup Text-to-Speech (Optional)

   • Clone GPT-SoVITS repository
   • Install pretrained models
   • Run pip install -r requirements.txt
   • Place a reference audio clip (<10s) in the root directory as reference.wav
   • Start the TTS server: python api_v2.py

📝 Configuration

[MateChat]
# Core Settings
CHAT_KEYBIND = "F8"
ENGINE_TYPE = "Cloudflare"
API_KEY = "your_api_key"
ACCOUNT_ID = "your_account_id"  # Optional for OpenRouter/OpenAI
MODEL_NAME = "llama-3.1-8b-instruct"
NAME = "USER"
AI_NAME = "Miku"

# System Prompt
SYSTEM_PROMPT = "You are a cheerful digital companion inspired by Hatsune Miku! Keep responses brief and energetic..."

# UI Settings
CHAT_WINDOW_WIDTH = 400
CHAT_WINDOW_HEIGHT = 500
CHAT_WINDOW_X = 20
CHAT_WINDOW_Y = 20
CHAT_WINDOW_FONT_SIZE = 24

# Advanced Features
ENABLE_AUDIO_MODEL = false
BASE_URL = ""

🎤 Text-to-Speech Configuration

Enable and configure TTS in your MateChat.cfg:

# Text-to-Speech Settings
ENABLE_TTS = true
ENABLE_AUDIO_MODEL = false      # Not yet supported, placeholder

TTS_ENGINE = "GPT-SoVITS"      # Currently only GPT-SoVITS is tested
TTS_API_URL = "http://localhost:9880"  # Default GPT-SoVITS server URL

# Voice Configuration
TTS_TEXT_LANG = "auto"         # Language detection for input text
TTS_REF_AUDIO_PATH = "reference.wav"   # Reference voice file (on server side)
TTS_PROMPT_TEXT = "[voice of reference.wav]"
TTS_PROMPT_LANG = "ja"         # Reference language (ja/zh/en/etc)

# Processing Settings
TTS_TEXT_SPLIT_METHOD = "cut5"
TTS_BATCH_SIZE = 1
TTS_MEDIA_TYPE = "wav"
TTS_STREAMING_MODE = false      # Streaming support pending GPT-SoVITS updates

Important Notes:

• The reference.wav path is relative to your GPT-SoVITS server directory
• Use a clear, high-quality voice sample under 10 seconds for best results
• Keep the GPT-SoVITS server running while using TTS features
• Streaming mode is implemented but waiting on full GPT-SoVITS support

💭 Coming Soon

• UI overhaul with improved aesthetics
• Speech bubble integration
• Enhanced character animations
• Speech-to-text capabilities
• And more exciting features! ♪

🌟 Support

Join our Discord server for help, updates, and cute chat screenshots!

The mod performs automatic configuration testing on startup and reload to ensure everything is working correctly. Check the MelonLoader console for detailed feedback~

---

Made with ♥ for Desktop Mate