Reproduce “AVTR-1: Open Stack for Real-Time Interactive Avatars” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Artem Kravtsov, Dmitrii Ziganshin, Vsevolod Poletaev, Gleb Balitskiy, Anastasia Tikhonova, Egor Burkov, Vadim Lebedev
AVTR-1 coordinates fragmented voice-agent speech with five-frame motion generation and scheduled audio-video playback, including interruptions. Its streamer converts uneven speech fragments into model-ready windows while synchronizing rendered frames with audio. On SI-184, R-DGG indicates paired speaker speech predicts listener motion beyond listener history and speaker motion. For OpenAI Realtime API, mean streamer-added response latency was 390.4 ms across ten paired exchanges, not end-to-end latency.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.