OpenAI seems to be reducing its focus on the video generation market with the closure of its Sora app, while ByteDance announced on Thursday that its new audio and video model, Dreamina Seedance 2.0, is being introduced in its editing platform, CapCut.
ByteDance claims the model enables creators to draft, edit, and synchronize video and audio content using prompts, images, or reference videos.
The gradual rollout will start with CapCut users in Brazil, Indonesia, Malaysia, Mexico, the Philippines, Thailand, and Vietnam, expanding to more markets over time.
The launch in CapCut comes after a report stated the model’s global rollout was paused to address intellectual property issues that drew criticism from Hollywood over alleged copyright infringement. This likely explains the limited availability of the model in CapCut’s current markets.
In China, the model is available for users of ByteDance’s Jianying app.
ByteDance states the video generation model can operate without reference images, even using just a few descriptive words. CapCut excels at rendering realistic textures, movement, and lighting from various perspectives and angles, which can enhance or correct creators’ own footage.
Another use case is allowing creators to explore potential ideas from early concepts or sketches before shooting the actual video.
Additionally, Dreamina Seedance 2.0 can be applied to various content types, such as cooking recipes, fitness tutorials, business overviews, or videos with movement, where AI video models have traditionally faced challenges, according to the company.
At launch, the model supports clips up to 15 seconds in length across six aspect ratios.
In CapCut, the model will be integrated into different areas, including editing features like AI Video and generation tools such as Video Studio. It will also extend to ByteDance’s AI generation platform, Dreamina, and its marketing platform, Pippit.
Due to its capacity to create realistic content, ByteDance has added safety measures that prevent the model from generating videos from images or videos containing real faces. CapCut will also restrict unauthorized generation of intellectual property. The model would already be available in the United States if the restrictions were functioning correctly, suggesting further adjustments are ongoing.
Dreamina Seedance 2.0’s content will feature an invisible watermark to identify material created with the model when shared off-platform, assisting in addressing takedown requests from rights holders if copyright content is inadvertently allowed.
ByteDance plans to collaborate with experts and creative communities as the model is rolled out to refine and enhance its capabilities.