The public GitHub repository open-llm-vtuber/open-llm-vtuber is a locally run AI companion that talks by voice without using your hands, allows you to interrupt it while it speaks, and puts a Live2D avatar on screen. The latest release is v1.2.1. However, the record reviewed this time contains no changelog for this version.
The setup is to talk to the avatar on screen and hear its answer out loud. The gist of the project is an AI companion that combines voice and visual input with Live2D characters. The README says it supports real-time voice conversation and visual perception. It can be used on Windows, macOS, and Linux as a web version and as a desktop client, and it can also run fully offline on your own computer. Speech recognition (ASR) is the technology that turns speech into text, and text-to-speech (TTS) is the technology that turns text back into sound. Both of these and the LLM each integrate with several solutions, and character customization is supported as well. The repository is written in Python, was created on November 24, 2023, and has 14221 stars.
Still, the limits of the available evidence are clear. The license shown by the repository is NOASSERTION, so the actual terms of use cannot be confirmed. The README summary reviewed this time gives no figures for voice latency or recognition accuracy, and no benchmarks that back up performance either. Of the score items, only momentum is filled in at 5.5 out of 10, while technology, quality, and novelty are left empty. Until the terms of use are confirmed, it is safer to hold off on any adoption decision.