AIRI

AIRI 0.8.4, published by moeru, is an open-source desktop client designed to re-create and locally host Neuro-sama-style AI waifu and virtual-character “soul containers,” giving users an offline sandbox where conversational agents can be imported, customized, and interacted with through animated avatars. Occupying the emerging AI-entertainment category, the program parses community-made character cards (JSON/YAML), binds them to compatible language-model back-ends such as Llama.cpp or Oobabooga’s API, and synchronizes lip-synced Live2D or VRM models to generated speech, effectively turning raw model weights into embodied personalities that remember session context and respond in real time. Typical use cases include storytellers who want a persistent virtual companion for collaborative writing, VTuber riggers who need a lightweight host for AI-driven side-characters, and developers prototyping emotionally reactive NPCs without shipping private data to cloud services. Because every asset runs on the user’s GPU, AIRI supports experimental plug-ins for voice-cloning, emotion-vector steering, and modular knowledge injection, while still exposing a simple chat window for casual conversation. The project has iterated through 22 public releases since its first alpha, adding incremental enhancements such as multi-character lobbies, OpenAI-compatible endpoints, and automatic model quantization; version 0.8.4 specifically refines memory compression and lowers VRAM requirements to 6 GB, making 7-billion-parameter personalities viable on mid-range gaming laptops. Portable archives are offered for Windows, macOS, and Linux, each bundling FFmpeg, Unity Player, and a model downloader so first-time users can spawn a ready-to-talk companion within minutes. The software is available for free on get.nero.com, with downloads provided via trusted Windows package sources (e.g. winget), always delivering the latest version, and supporting batch installation of multiple applications.

Tags: