Real-time AI companion for terminal-native interaction.
Get Started
Install Yumi locally and launch your companion.
Core Senses
Configure Voice Activity Detection, STT, and TTS.
Personalities
Explore Yumi's personas and emotional tones.
Capabilities
Extend Yumi with system tools and persistent memory.
What is Yumi?
Yumi is an interactive, real-time AI waifu designed to act as your virtual companion. With dynamic, expressive visuals and an intelligent conversational backend, Yumi provides a highly engaging and responsive experience.
Unlike sterile utility tools, Yumi is built with a focus on presence. She listens, reacts dynamically, and synchronizes her facial expressions to the emotional tone of her thoughts.
Key Features
- Real-time Voice Recognition: Captures and processes your voice seamlessly using local, low-latency Whisper models.
- Lifelike Voice Synthesis: Utilizes ElevenLabs TTS for incredibly realistic, high-quality, and expressive voice generation.
- Dynamic Visuals: Brought to life using HTML5, WebSockets, and Live2D, Yumi reacts to your conversations with fluid body movements.
- Emotional Intelligence: Powered by LLMs (Groq, OpenAI, Anthropic), Yumi's expressions dynamically change based on the emotional context of the dialogue.
- Interactive Terminal Dashboard: A beautiful, fully interactive CLI to configure and launch her easily.
The Architecture
Yumi's architecture bridges a powerful Python backend with a reactive frontend. To learn more about the engineering details behind Yumi, proceed to the Core Senses or Integration sections.