humanaigc-engineering/openavatarchat is an avatar chat system that bundles speech, video, and language models into swappable modules. The README describes this repository as a modular conversational digital human implementation that supports interaction through text, speech, and video. ASR (speech recognition), LLM, TTS (speech synthesis), and Avatar components can be flexibly swapped, and according to GitHub releases the latest version is 0.6.0.
The modular structure is the core of this repository. Its purpose is to bring several digital human renderers and real-time conversation flows together in one place. It supports digital human technologies such as LiteAvatar, LAM, MuseTalk, and FlashHead, and uses VAD (voice activity detection), speech buffering, and frame rate control to reduce response latency. It is a Python repository released under the Apache-2.0 license, with 3810 GitHub stars.
However, the provided record contains no changelog for 0.6.0, so it cannot be confirmed what changed in this release. The record also has no benchmark showing the size of the latency improvement, and the momentum score came out low at 1.0 on a 10-point scale. The repository also started on February 20, 2025, so its history is short. It is a candidate for anyone looking for an avatar chat implementation to reference for its module replacement approach, but it is too early to judge what 0.6.0 actually improved.