A guided walkthrough of the main features, settings, and troubleshooting tools in Waifu Companion.
This app pairs a Live2D character 🎭 with a chat 🗨️ powered by an AI model 🤖. You can move 🖱️ and zoom 🔍 the character, chat 💬, and even listen to radio 🎶 while using TTS 🔊.
Type your message ⌨️ and press Enter ↩️. The AI processes your input and responds with text, emotion, and gestures 🎭. Your conversation history is saved locally 💾. Use Reset Chat 🗑️ to clear it and start fresh.
Configure your AI model in Settings → LLM Provider. You can use OpenRouter or Groq - both support free models. Provide your API key 🔑 and a model name. They support Streaming ⚡ for real-time replies!
Customize your AI's identity 🆔 and personality 🎭 in Persona Settings. The AI has Long-Term Memory 🧵: it remembers the last few dozen messages and automatically creates a Conversation Summary 📝 to retain important facts even after they leave immediate memory.
Switch characters in Settings → Model Settings 🧩. You can pick from built-in models or add your own via a .model3.json URL 🔗. The AI can trigger Emotions (Happy, Sad, Surprised, etc.) and Gestures (Nod, Shake Head) automatically during chat ✨.
Enable Voice 🔈 to hear the AI. You can choose a different voice 🎙️ for each response language. TikTok TTS offers high-quality voices for many languages 🌎. Adjust volume 🔉 and chunk limits in Voice Settings.
Generate custom backgrounds 🎨 from prompts or the current chat context in Settings → Background Image 🖼️. Manage your collection in the Background Library 📚, use the fullscreen viewer 🖥️, and choose how the image fits your screen (Contain, Cover, etc.).
Listen to an anime radio stream 🎶 while you chat. Controls are in Settings → Audio Settings 🎧. Volume is saved automatically 💾. Click the play button ▶️ to start the music!
Click the Microphone 🎤 icon in the chatbox to talk instead of typing. It uses your browser's speech recognition ✍️. Make sure to allow microphone permissions in your browser 🔐.
Set the AI's response language 🗣️ and optionally show Transliteration 🔤 (Romaji/Romanization) for Japanese or Korean. You can also enable UI Translation 🈯 to have the AI translate the entire app interface for you!
In Preferences ⚙️, you can choose to include the current Time ⏱️ and Battery 🔋 level in the AI's context so it can make comments about your day or device status 🧠. You can also adjust chatbox and bubble opacities 🎨.
If the AI connection fails or you're offline, the app switches to a Local Fallback Engine 🔌. It uses basic heuristics to keep the conversation going until you're back online 🌐. You can also force Offline Mode in Debug Settings.
Troubleshoot issues with the Debug Panel 🧰. See real-time logs 📋 for AI calls, model loading, and system events. You can even see the full context 📄 being sent to the AI to understand how it thinks.
The Chat window is fully flexible! Drag it by the header 🪟 and resize from the edges ↔️↕️. Your layout is saved 💾 so it stays consistent between visits.
• Click/Tap the character to interact 🖱️
• Use mouse wheel to zoom 🌀
• Drag the character to reposition 📍
• Enable 'Always show Settings' 🪄 if you tweak often
• Check the Links 🔗 section for more from Waifu AI!
Configure your AI model in Settings → LLM Provider. You can use OpenRouter or Groq - both support free models. You'll need an API key from openrouter.ai or console.groq.com. Both support Streaming ⚡ for real-time character-by-character responses.
Your AI isn't just a generic bot; you can define its Core Identity and Custom Personality in Persona Settings. Whether you want a helpful assistant or a specific character archetype, the AI will adapt its tone and behavior. To handle long conversations, the app uses a Memory System. As older messages are pushed out of immediate memory to save performance, the AI automatically generates a Conversation Summary 📝. This summary is fed back into the AI's context, ensuring it never truly "forgets" important events or your name.
The Live2D models aren't just static images. They feature Automatic Gaze 👁️ (following your interactions), breathing animations, and physics-based hair/clothing movement. When chatting, the AI can trigger specific Emotions (Happy, Sad, Surprised, Thoughtful, Excited) and Gestures (like nodding or tilting its head) based on the sentiment of its reply. You can also add your own models by providing a URL to a .model3.json file.
Text-to-Speech (TTS) is powered by TikTok's free TTS API, offering a variety of natural-sounding voices across many languages. You can select a unique voice for each language you use. For input, use the Microphone 🎤 icon to speak directly to your character. This uses your browser's native Speech-To-Text capabilities. Note: For the best experience, use a Chromium-based browser and ensure you've granted microphone permissions.
The AI can be made "aware" of your environment. By enabling Time and Battery context, the AI will know if it's late at night or if your device is running low on power, leading to more natural observations. If you lose your internet connection, the Local Fallback Engine 🔌 takes over, using smart heuristics to provide relevant (though less "intelligent") replies until the connection is restored.
If something isn't working as expected, the Debug Panel is your best friend. It logs everything from network requests to model loading errors. You can even enable Show Chat Context to see exactly what "thoughts" and instructions are being sent to the AI behind the scenes. If audio doesn't play, remember that most browsers require at least one manual click 🖱️ on the page before they allow automatic sound playback.