.svg){kind=spacer}
Details
- Baidu launches a significant update to MuseSteamer, enabling real-time, interactive long-form video creation beyond its previous 10-second limit.
- Internal tests reveal the platform now renders 4K video twice as fast, and users are able to pause videos, branch storylines, or seamlessly add scenes.
- The “open-world” mode lets users stitch together video clips into large, explorable 3D environments for uses such as game maps, virtual tours, and space simulations.
- MuseSteamer’s web studio and API are now in public beta in both English and Mandarin, with enterprise clients gaining access to batch rendering and fine-tuning.
- The system operates on a hybrid infrastructure of Baidu Kunlun AI chips and NVIDIA GH200 clusters, reducing GPU costs per video minute by 35% since April 2025.
- This upgrade integrates MuseSteamer further into Baidu’s AI suite, alongside ERNIE for text and Comate for code, supporting cross-modal content creation pipelines.
Impact
MuseSteamer’s real-time, branchable and open-world video capabilities directly challenge AI filmmaking leaders like Runway and anticipated OpenAI Sora successors, heightening competitive pressure in the generative media space. By extending video duration and reducing hardware costs, Baidu opens new monetization paths for creators and brands, while strengthening China’s domestic alternatives to platforms like Unity Muse. The move also illustrates China’s drive for tech self-reliance, particularly amid global semiconductor trade tensions.
