ByteDance recently launched its latest AI video creation app, Seedance 2.0, on February 7. The app has impressed analysts with its high-quality content, resulting in a surge in shares of Chinese media and AI companies.
COL Group Co., a publishing company, hit the daily trading limit with a 20% increase in shares, while both Shanghai Film Co. and Perfect World Co. saw their shares rise by 10%. Additionally, the onshore CSI 300 Index increased by 1.4%.
Analysts from Kaiyuan Securities pointed out that the app supports various input types, including text, images, videos, and audio, and has achieved significant advancements in its key capabilities. This development has sparked discussions within the industry about the potential of AI in the film and television sectors.
Global race and copyright debates
ByteDance’s launch of Seedance 2.0 signifies its push into frontier AI, directly challenging U.S. leadership in this field. According to one tech CEO, who compared the models to their American counterparts, they are “better in every way.” Furthermore, a Chinese internet analyst informed CNBC that while AI video generation is still in its early stages, some Chinese companies are already demonstrating a lead.
That advantage may stem from more relaxed training and output boundaries. A co-founder of Kapwing, a browser-based video editing platform, stated that Seedance and Seedream appear willing to recreate copyrighted characters such as Mickey Mouse or the Minions.
The cost is also a factor. According to TIME, Seedance’s usage on Freepik, a stock media and design marketplace, required only half the credits that Google’s Veo 3 would have needed.
What is Seedance 2.0?
Seedance 2.0 adds tools that allow creators to direct results with more precision, closer to how a film director distills a scene across takes.
Its “Universal Reference” feature supports up to 12 reference files per generation and takes multimodal inputs, including text, images, video, and audio. Seedance 2.0 can copy composition, camera movement, character actions, and style from those assets.
The model is being incorporated into external platforms, some of which offer 2K or greater-resolution output options.