Alibaba Cloud leads ShengShu Technology’s funding round to accelerate the development of multimodal AI models that process video, audio, and text, signaling a broader shift toward non-language-focused AI with real-world action.

ShengShu aims to build a general-purpose system that understands and acts in physical environments by bridging digital and real-world contexts through its Vidu video platform and related technologies.

Industry observers note that multimodal world models could become the next major phase of AI, addressing the real-world grounding gaps seen in traditional LLMs and expanding into robotics and embodied AI.

ShengShu did not provide a timeline for when commercially viable world-model systems might be ready.

HappyHorse 1.0 has debuted with strong performance in video-generation rankings, and more ATH products are anticipated.

The broader context shows LLMs excel at text but struggle with real-time physical-world understanding, prompting investment in multimodal world models.

ShengShu reported significant growth in users and revenue in 2025, though exact figures were not disclosed.

Industry watchers expect world models to define AI’s next stage if they can successfully integrate multiple modalities with real-world grounding.

ShengShu’s Vidu Q3 Pro ranks competitively in text-to-video and image-to-video benchmarks, with audio-enabled versions improving rankings.

Vidu targets independent creators and animators, promising effortless content production across styles while aiming to build a general world model using multimodal data.

ShengShu’s core concept is a multimodal world model trained on vision, audio, and touch to simulate real-world physics and causal reasoning, potentially outperforming text-only LLMs.

The company emphasizes that combining vision, audio, and touch data enables perception-action understanding that connects perception to real-world interaction.