![[2511.07399] StreamDiffusionV2: A Streaming System for Dynamic and Interactive Video Generation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2511.07399] StreamDiffusionV2: A Streaming System for Dynamic and Interactive Video Generation
Summary
StreamDiffusionV2 presents a novel system for dynamic and interactive video generation, enhancing live streaming capabilities through optimized video diffusion models.
Why It Matters
As generative models evolve, they significantly impact the live-streaming industry by improving content creation and delivery. StreamDiffusionV2 addresses critical challenges in real-time streaming, such as latency and scalability, making advanced video generation accessible to both creators and enterprises.
Key Takeaways
- StreamDiffusionV2 enables interactive live streaming with enhanced temporal consistency.
- The system achieves low latency and high frame rates, crucial for real-time applications.
- It supports scalable multi-GPU environments, optimizing resource use for video generation.
Computer Science > Computer Vision and Pattern Recognition arXiv:2511.07399 (cs) [Submitted on 10 Nov 2025 (v1), last revised 22 Feb 2026 (this version, v2)] Title:StreamDiffusionV2: A Streaming System for Dynamic and Interactive Video Generation Authors:Tianrui Feng, Zhi Li, Shuo Yang, Haocheng Xi, Muyang Li, Xiuyu Li, Lvmin Zhang, Keting Yang, Kelly Peng, Song Han, Maneesh Agrawala, Kurt Keutzer, Akio Kodaira, Chenfeng Xu View a PDF of the paper titled StreamDiffusionV2: A Streaming System for Dynamic and Interactive Video Generation, by Tianrui Feng and 13 other authors View PDF HTML (experimental) Abstract:Generative models are reshaping the live-streaming industry by redefining how content is created, styled, and delivered. Previous image-based streaming diffusion models have powered efficient and creative live streaming products but have hit limits on temporal consistency due to the foundation of image-based designs. Recent advances in video diffusion have markedly improved temporal consistency and sampling efficiency for offline generation. However, offline generation systems primarily optimize throughput by batching large workloads. In contrast, live online streaming operates under strict service-level objectives (SLOs): time-to-first-frame must be minimal, and every frame must meet a per-frame deadline with low jitter. Besides, scalable multi-GPU serving for real-time streams remains largely unresolved so far. To address this, we present StreamDiffusionV2, a train...
