Artificial intelligence video field is making waves again. Recently, ByteDance quietly tested its new multimodal video generation model Seedance2.0 with some users. This model, praised by industry insiders as "the strongest on the surface," is rapidly changing the landscape of content creation with its deeply integrated native audio and video generation capabilities and **"director-level" control precision**.

QQ20260210-091340.png

Multimodal Integration and "Reference" Function Breakthroughs

Seedance2.0's biggest technical highlight lies in its ability to process four types of input: images, videos, audio, and text. Users can combine up to 12 files (including 9 images, 3 videos, and 3 audios) for creation.

  • Action Learning: The model can learn cinematography, motion paths, and effects from uploaded reference videos, and seamlessly expand existing clips or replace characters.

  • Native Sound Effects: Unlike previous versions, the 2.0 version automatically matches highly synchronized sound effects, music, and even lip movements while generating videos lasting 4 to 15 seconds, greatly reducing the complexity of post-production.

Test Feedback: Surprising and Controversial

Renowned tech blogger Tim from "Film Hurricane" used the word "terrifying" six times to describe its performance in his review, pointing out that the model can accurately simulate his own voice and details of his company's building not visible in the image, even without providing audio. This powerful "cloning" capability has raised deep concerns about "trust crises" and copyright issues from industry leaders such as Feng Ji, the producer of "Black Myth: Wu Kong." Currently, ByteDance has temporarily suspended some sensitive human reference functions.

Industry Rivalry: The "Double Superpower" of Chinese AI Video

At the same time, Kuaishou also recently launched Kling3.0. Unlike Seedance2.0, which focuses on complex multimodal references and camera movement controls, Kling3.0 excels in physical feedback, multilingual lip-syncing, and character consistency.

The confrontation between the two strong players has ignited enthusiasm in the secondary market. According to reports from "South China Morning Post" and A-share market data, due to the release of these "singularity-level" models, the stock prices of several domestic media and AI-related companies once surged by 20%, with the capital market generally optimistic about the commercialization prospects of AI video in short dramas, animation, and e-commerce fields.