AI News

Trending story

inclusionAI releases Duo v1 full-duplex interaction model

1 reports1 story updatesUpdated Nov 13 17:42

Understand this story

AI overview

On November 13, 2026, Ant's inclusionAI released the Duo v1 full-duplex audio-video interaction model on GitHub. The model is described as a Reciprocal World Model that can continuously perceive audio-video input and respond in real time, supporting natural conversation and interruptions. This release simultaneously provides the model weights and offline inference code for Duo-1.0-causal, Duo-1.0-distill, Duo-1.0-distill-quant, and Duo-1.2-causal, along with online streaming pipelines accelerated by vLLM-Omni and SGLang-Omni. The official announcement describes this as a preview version for researchers; it currently supports only basic audio-video conversation, and Duo v2 will support visual interaction driven directly by user audio and video.

AI-generated from public reports

Latest development · Report

Ant inclusionAI Open-Sources Duo v1, a Full-Duplex Audio-Video Interaction Model

Developments

1 confirmed developments

  1. inclusionAI releases Duo v1 full-duplex interaction model

    Source:English text is being prepared.

Reports and discussion timeline

Browse reports and discussion by time. Their evidence levels remain distinct.

Nov 13

  1. Ant inclusionAI: New GitHub RepositoryFirst-partySelected

    Ant inclusionAI Open-Sources Duo v1, a Full-Duplex Audio-Video Interaction Model

    Ant inclusionAI has released Duo v1 on GitHub, a Reciprocal World Model that treats intelligence and physical reality as equally important, can continuously perceive audio-video input and respond in real time, and supports natural conversation and interruption. This release simultaneously provides the model weights for Duo-1.0-causal, Duo-1.0-distill, Duo-1.0-distill-quant, and Duo-1.2-causal along with offline inference code, and offers online streaming pipelines accelerated by vLLM-Omni and SGLang-Omni. The official statement describes this as a preview version for researchers; it currently supports only basic audio-video conversation, and Duo v2 will support visual interaction driven directly by the user's audio and video.

Related stories

  1. Related storiesAnt Group's GuangJian team releases visual forensics model DefakerOne2
  2. Related storiesPika releases MiniMax H3 derivative model Pika H3 Flex
  3. Related storiesVAST open-sources real-time world model EdenLive
  4. Related storiesEnglish translation is being prepared.