Alibaba’s cloud computing updated AI model, Animate Anyone 2, enables users to make realistic character animations by combining a single image with a video.
According to a report by SCMP, this system advances previous motion signal-based methods by adding environmental information from reference content.
The research study from Alibaba Cloud’s Tongyi Lab demonstrated how this process extracts action data and background elements from source videos to create realistic animated clips. The updated model achieves a more natural look because it places the animated character into its original environment.
The research posted on arxiv states, “We extract environmental information from [source] videos, enabling the animated character to preserve its original environment … facilitating seamless character-environment integration.”
Chinese developers are rapidly progressing in generative AI
Animate Anyone was initially released in late 2023, enabling users to create character-based videos through static images.
OpenAI launched Sora as its text-to-video model after creating ChatGPT last year. The innovation triggered numerous comparable text-to-video model releases by major Chinese technology corporations and startup companies. In December, Sora became accessible exclusively to subscribers of both ChatGPT Plus and Pro memberships.
China demonstrates its desire to match worldwide leaders in the AI field through the recent release of Animate Anyone 2. Alibaba’s announcement follows ByteDance’s recently released OmniHuman-1 as its multimodal model for transforming pictures and recorded sounds into realistic video content.
Alibaba’s Animate Anyone 2 study used 100,000-character videos
Alibaba’s study added that The Animate Anyone 2 platform creates interactions between characters by maintaining consistent movements and “coherence with the surrounding environment.”
For this study, researchers created a database consisting of 100,000 character videos taken from the Internet, with various scenes and interactions between humans and objects, to evaluate the performance of Animate Anyone 2 across different settings.
However, Anyone 2 poses the risk of generating additional deepfake videos.
Liang Haisheng, from his Beijing-based advertising agency, states that such AI tools “fall short in terms of conveying human emotions and detailed facial expressions on video.”
Liang added that the basic visual representation in these AI-generated videos still allows businesses to transform concepts into visuals in idea pitches for clients.
Alibaba announced an increase in the company’s quarterly profits pushed by AI strategies. The company looks forward to major AI investments in the coming 3 years as an opportunity that “comes around once every few decades.”
Cryptopolitan Academy: Want to grow your money in 2025? Learn how to do it with DeFi in our upcoming webclass. Save Your Spot
