Live2D Full app ↗
🔍
Companion
🎭 Character ›
🧠 Persona ›
🌐 Language ›
AI
🤖 AI Provider ›
🧵 Memory ›
🍃 Ambient & Queue ›
Appearance
🖼️ Background ›
🖥️ Interface ›
Sound
🔊 Voice ›
📻 Radio ›
Advanced
🧩 AI Context ›
🐞 Debug ›
Help & About
❓ Help & Tutorial ›
🔗 Links ›

🌐 Language

🤖 AI Provider

✨ WaifuAI Cloud (Default)

Active automatically for all guests. Powered by WaifuAI Cloud. No API key required!
To bring your own API key instead, simply check one of the custom providers below.

OpenRouter

When enabled, chat, translation, and summarization will call OpenRouter using your API key and model name.
This is stored locally in your browser only and sent directly to openrouter.ai over HTTPS.
First fallback model. Used if the primary model fails.
Second fallback model. Used if the primary and first fallback models fail.

Groq

When enabled, chat, translation, and summarization will call Groq using your API key and model name.
Get your free API key from console.groq.com. Stored locally in your browser.
Examples: llama-3.3-70b-versatile, llama-3.1-70b-versatile, mixtral-8x7b-32768, etc.

OpenAI Compatible

Connect to any OpenAI-compatible API endpoint (local or remote) by specifying a custom base URL.
The base URL of your OpenAI-compatible API (e.g. http://localhost:1234/v1 for LM Studio, or a custom endpoint). Do not include /chat/completions.
API key for the endpoint. Use any value if your local server does not require authentication.
The model name as expected by your API endpoint.
Optional CORS proxy URL to prefix API requests. Leave empty if the API endpoint supports CORS directly.

🎭 Character

When enabled, selecting a model adds it without removing existing ones.
Shape of the pictures she draws for you.

🧠 Persona

🧵 Memory

Current: 30 messages
Trigger after: 30 messages
Memory limit: 32 KiB. Older text is kept until compression succeeds.

🧩 AI Context

Provides the AI with the current date and time for context-aware responses.
Allows the AI to know your device's battery level and charging state.
Forces the AI to respond in JSON for emotional reactions, but may cause errors on weaker LLMs.

🔊 Voice

Primary voice synthesis using TikTok's free TTS API.
Matches the TikTok voice to the language of each reply. Turn this off to always use your selected voice.
High-quality local TTS. Preloaded in background for silent handoff.
Used automatically if the primary voice provider fails (e.g. rate limit).
Max Characters: 300
Volume: 1.00
Speech-to-text engine for the chat microphone. Records your mic and transcribes via WaifuAI Cloud - works in browsers without the Web Speech API.

📻 Radio

🖥️ Interface

Current: 0.9
Current: 0.3
If enabled, the settings panel opens on every load. Otherwise, it remembers your last open/closed state.

🖼️ Background

Current: 1.0
Draws a new background when the chat moves somewhere new. It never replaces a background you picked yourself.

🍃 Ambient & Queue

Queue messages sent while the AI is still writing its reply. You never wait for her to finish speaking: sending a message interrupts the voice.
The AI will speak unprompted after periods of silence (pauses after 10 unreplied messages on Free Cloud; unlimited with custom API key).
Trigger after: 10s
Pause after: 10 messages
Pause unprompted thoughts after this many unreplied messages (1-50 or Unlimited; clamped to 10 on Free Cloud).
Pre-fetch the next ambient comment in the background for instant response.

🐞 Debug

❓ Help & Tutorial

Open the full guide in the docs:

📖 Help & Tutorial Docs

Chat History
Chat
🧰 Debug Panel