While OpenAI reportedly reduces its focus on video generation by discontinuing its Sora app, ByteDance announced Thursday that its latest audio and video AI model, Dreamina Seedance 2.0, is being integrated into its CapCut editing platform.
According to ByteDance, this model enables creators to generate, modify, and synchronize video and audio content using text prompts, images, or existing videos.
The initial rollout targets CapCut users in Brazil, Indonesia, Malaysia, Mexico, the Philippines, Thailand, and Vietnam, with plans to expand to additional regions gradually.
This CapCut launch comes after reports suggested the model’s global deployment was halted to resolve intellectual property concerns, which had attracted condemnation from Hollywood regarding alleged copyright violations. This context likely accounts for the restricted availability of the model within CapCut’s current markets.
In China, the model is already accessible through ByteDance’s Jianying application.
According to ByteDance’s official announcement, the video generation model can function without requiring reference images, even when a creator provides only a brief textual description of their desired scene. The company also highlights CapCut’s proficiency in depicting lifelike textures, motion, and illumination from various angles, suggesting its utility for editing, improving, or correcting users’ existing video material.
Furthermore, it offers creators the capability to experiment with potential concepts derived from initial ideas or outlines, prior to actual video production.
Moreover, Dreamina Seedance 2.0 is suitable for diverse content types such as cooking demonstrations, fitness guides, business or product presentations, and action-oriented videos – areas where AI video models have traditionally encountered difficulties, as detailed by the company.
Upon release, the model supports video clips up to 15 seconds in length, available in six different aspect ratios.
Within CapCut, the model will be progressively integrated into various functionalities, including AI Video editing features and generation tools like Video Studio. It is also slated for inclusion in ByteDance’s AI generation platform, Dreamina, and its marketing platform, Pippit.
Recognizing its capacity for realistic content creation, ByteDance has implemented safety measures, preventing the model from generating videos from images or videos featuring real human faces. CapCut will also prohibit the unauthorized creation of intellectual property. (It’s worth noting that if these restrictions were fully functional, the model would likely be accessible in the United States by now, suggesting ongoing adjustments.)
ByteDance further stated that all content generated by Dreamina Seedance 2.0 will incorporate an invisible watermark. This watermark will facilitate the identification of AI-generated content when shared externally, potentially assisting rights holders with takedown requests should copyrighted material inadvertently bypass the model’s filters.
ByteDance plans to collaborate with specialists and creative communities throughout the model’s deployment to continuously refine and enhance its features.