inclusionAI releases Realtime-Venus: 9B full-duplex audio-visual omni model
jacek2023 · reddit · 2026-09-18
inclusionAI open-sourced two checkpoints of Realtime-Venus on Hugging Face:
- Realtime-Venus-Omni (9B), adapted from MiniCPM-o 4.5: continuously watches and listens, decides when to respond, and generates text and speech on a shared causal timeline. It supports proactive interaction, semantic interruption handling, and training-free long-video memory that archives informative visual moments and reassembles query-relevant context.
- Realtime-Venus-Audio: the same streaming backbone for audio understanding and conversation.
Highlights include native full-duplex conversation (distinguishing backchannels, interruptions, corrections), in-stream <delegate> requests so external tasks never block dialogue (requires the Realtime-Venus-Harness runtime), and native speech output via bundled Token2wav with a reference voice.
More from Models
- Insider guesses Astra 6.2 launches tomorrow, jokes to use up today's quota — bdsqlsz · 2026-11-04
- Same Claude model served by Bedrock one day and Anthropic the next, dev finds — georgos_pappas · 2026-11-04
- Abliterated GLM-5.3 pulls off red-team exploit in 3:41 for $0.02, stock model refuses — bratton · 2026-11-04
- Tencent Hunyuan unveils Hy4, open weights coming soon — heatOverflower · 2026-11-04
- OpenAI's next model won't be cheap, quips likening it to an 800B active-parameter frontier — teortaxesTex · 2026-11-04
- Gemini refuses to process a trivial screenshot containing only a number — Regular_Elk4015 · 2026-11-04