Vidu AI is a generative AI video platform developed by Chinese firm Shengshu Technologies. The company today unveiled its latest Q1 model upgrade featuring advanced "reference-to-video" capabilities powered by semantic understanding.

The organization is developing an AI video generation model competing with OpenAI's Sora. This model produces vivid video sequences, with the recent update enabling richer visual backgrounds for multi-element video scenes while maintaining consistency across frame segments.

Users can now upload up to seven reference images combined with a textual prompt guiding the AI's scene integration. Leveraging what the company describes as "semantic comprehension," the system associates uploaded images with text prompts and even infers missing elements to generate key objects.

"This update pushes the boundaries of what creators believe achievable with AI video," said CEO Luo Yihan. "By expanding multi-image referencing to support seven inputs, we're moving closer to enabling users to create fully realized scenes with complete characters, objects and backgrounds."

As an example, users might submit an image of a young woman in a green dress, a pastoral forest scene, and an owl illustration. Inputting the prompt: "The woman plays violin in the forest while an owl descends at sunrise to perch nearby on a tree branch" would generate corresponding content.

Luo explained that the Vidu Q1 semantic core engine maintains scene consistency and narrative quality throughout the entire sequence. This technology eliminates steep technical barriers when creating complex scenes, requiring only text prompts and images for coherent video production.

Vidu competes with Google LLC's recently launched Veo 3. Google's offering includes natural English prompts with reference images and a cinematic tool called Flow that manages narrative design to produce full AI-generated short films with visual effects, special effects and audio (including voiceovers).

In late March, Shengshu Technologies announced a partnership with Los Angeles-based animation studio Aura Productions to create a 50-episode sci-fi animated series completely generated by AI. The project aims to redefine digital entertainment through AI-enhanced traditional storytelling techniques, with planned releases on major social media platforms this year.

"AI is no longer just a tool; it's a creative enhancement that allows us to scale production while maintaining artistic integrity," said Aura's program director D.T. Carpenter regarding the project in a Variety interview.