Realtime-Venus

About

Realtime-Venus is a full-duplex multimodal AI system from InclusionAI that continuously processes synchronized audio and video, decides when to respond, and generates text and speech on a shared causal timeline, including handling interruptions and corrections mid-conversation. It ships as two 9B-parameter checkpoints, an audio-visual Omni model built on MiniCPM-o 4.5 and an audio-only model, and supports in-stream task delegation to external tools plus training-free long-video memory that retrieves relevant past visual moments without additional training.