New open-source AVATAR gives your AI agents a face
Talking to a voice model today often feels like speaking into a well-lit void. The answers can be sharp, the TTS can sound human, and still something is missing: there is no presence on the desk. No glance when the model starts. No small gesture when a thought lands. Just a waveform, a chat log, or a black window that happens to speak. That gap is why we built AVATAR — an open-source desktop…
In a world where voice models often feel distant and devoid of personal touch, ARPA Hellenic Logical Systems has developed an open-source solution to bring a touch of life to AI agents. The AVATAR project creates a VRM character that appears on your screen, moving and lip-syncing to live audio, giving voice models a visual presence alongside their auditory output.
AVATAR functions as a desktop overlay for Windows, running transparently and always-on-top. Users can drag the glass bar to move it, pin it over their preferred applications, and let it react to sound from the system output, chosen app windows, microphones, or files. The software comes with sample characters and allows users to import their own .vrm and environment folders through the Settings → Directories option.
The tool is designed with users in mind, offering a face for local LLMs, voice pipelines, and agent stacks. It's particularly useful for streamers, educators, students, and accessibility enthusiasts who want to keep a visual cue during audio-based tasks. The project is MIT-licensed, with contributors encouraged to build their characters and share them through the VRoid Hub.
The latest version, v0.7.0, focuses on improving motion capabilities, allowing users to load their own .vrma clips and bind key triggers to specific gestures. The application remains lightweight and does not rely on cloud services, keeping all data local and under the user's control. To get started, users can download the installer from the v0.7.0 release or clone the repository from GitHub and run the desktop version.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
