OpenAI appears to be scaling back its presence in the video generation space with the shutdown of its Sora app. However, ByteDance has now confirmed the rollout of its new audio and video model, Dreamina Seedance 2.0, within its editing platform CapCut.
According to ByteDance, this model enables creators to draft, edit, and synchronize video and audio content using simple prompts, images, or reference videos. The phased release will begin with CapCut users in Brazil, Indonesia, Malaysia, Mexico, the Philippines, Thailand, and Vietnam, with plans to expand to more markets later.
This launch follows a recent report that the model’s global rollout was paused while the company addressed intellectual property concerns. These issues had drawn criticism from Hollywood over allegations of copyright infringement, which likely explains the limited initial release.
In China, the model is already available to users of ByteDance’s Jianying app. The company states that Dreamina Seedance 2.0 can generate video from just a few descriptive words, without needing a reference image. It is also proficient at rendering realistic textures, movement, and lighting from various angles, which can be used to edit or enhance a creator’s own footage.
One practical use is allowing creators to test concepts or sketches before committing to full production. The model supports a wide range of content types, including cooking recipes, fitness tutorials, product overviews, and action-focused videos, areas where AI video models have traditionally struggled. At launch, it can create clips up to 15 seconds long in six different aspect ratios.
Within CapCut, the model will be integrated into editing features like AI Video and generation tools such as Video Studio. It will also be available on ByteDance’s AI platform Dreamina and its marketing platform Pippit.
Due to its ability to create realistic content, ByteDance has implemented safety restrictions. The model cannot generate videos from images or videos containing real faces, and CapCut will block the unauthorized generation of copyrighted intellectual property. The company notes that if these restrictions were fully operational, the model would already be available in the United States, suggesting further adjustments are underway.
All content produced by Dreamina Seedance 2.0 will include an invisible watermark to help identify AI-generated material when it is shared outside the platform. This measure could assist rights holders with takedown requests if copyrighted content is inadvertently generated.
ByteDance has stated it will collaborate with experts and creative communities throughout the rollout to iteratively improve the model’s capabilities.