OpenAmika v1.0.0

The first open-source release of Amika, MekaHime's AI companion: a Live2D anime girl who talks with you by
voice and text in real time, remembers you, and lives with you in your city. It runs on your own Windows PC with
OpenRouter or fully local with Ollama. A hosted version runs at https://www.mekahime.com/amika.

What's in it

• Brain: three prompt tiers (surface reflex, deep reply with inner thoughts, background reflection), an
  evidence-based long-term memory (BM25 + local bge-m3 embeddings), per-call cost metering, MCP tools.
• Ears: local SenseVoiceSmall speech recognition with voice activity detection and barge-in.
• Voice: the OpenAmika voice server (vendor/tts-engine, a MekaHime build of GPT-SoVITS) with 69 acted
  emotions; one emotion tag per sentence drives her voice, her face and her motion together.
• Body: Live2D stage; her own model downloads with her voice.
• Pages: Setup, Tools, Use (with a first-run wizard) and About, plus the architecture map at /flow.
• Optional: camera eyes (local VLM), smart home (Zigbee via MCP), Discord/Telegram "Text her".

Install

Follow Install (Windows) in the README. Her voice and her
Live2D model come from the Hugging Face repository
MekaHime/MekaHime_Amika: envs\brain\python.exe scripts\setup_tts.py shows their license, asks you to type yes, and installs the voice server, upstream
GPT-SoVITS with its models, her voice and her model (about 5.5 GB of downloads; every file pinned by SHA-256).

Licenses

OpenAmika's code is MIT. Amika's voice model, reference audio and Live2D model are © MekaHime Inc. under a
personal-use license: personal, non-commercial use only, 18+, no redistribution, no training other models on them,
visible credit to MekaHime when you post her. Live2D Cubism Core keeps Live2D's own license.

Report bugs at https://github.com/zeekk0/OpenAmika/issues, by email hi@mekahime.com, or on Discord
https://discord.gg/mekahime · Support the project: https://www.mekahime.com/donate