研究人员推出了 Wan-Streamer v0.3，这是一种新的交互模型，它将视频概念化为持久的“世界”和动态的“事件流”的组合。这种方法允许对视频数据进行通用预训练，使模型能够根据传入的输入预测世界如何随时间变化。该模型已应用于实时全双工视听交互，将多模态用户输入映射到语音和行为动作，响应延迟约为 200 毫秒。 AI
影响 引入了一种新颖的视频预训练任务，可以增强 AI 代理的多模态理解和实时交互能力。
排序理由 该集群包含详细介绍新模型和交互范式的研究论文。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
We present Wan-Streamer v0.3, which reframes our native-streaming interaction model under a single organizing view: a video is a world plus an event stream. The world is the persistent context in which a video unfolds, including the environment, scene, subjects, ambient acoustic …
arXiv:2607.15038v1 Announce Type: new Abstract: We present Wan-Streamer v0.3, which reframes our native-streaming interaction model under a single organizing view: a video is a world plus an event stream. The world is the persistent context in which a video unfolds, including the…
We present Wan-Streamer v0.3, which reframes our native-streaming interaction model under a single organizing view: a video is a world plus an event stream. The world is the persistent context in which a video unfolds, including the environment, scene, subjects, ambient acoustic …