Today, we're excited to start rolling out the innovative new video and audio model Dreamina Seedance 2.0 in our editing platform CapCut, providing imaginative new ways for our communities to bring their creativity to life. Our phased rollout for paid CapCut users will bring Dreamina Seedance 2.0 to certain users in Indonesia, Philippines, Thailand, Vietnam, Malaysia, Brazil and Mexico, with expansion over time.
To support Dreamina Seedance 2.0's advanced capabilities, we're launching the model with firm safeguards against violations of our safety policies and the unauthorized use of peoples' likeness or intellectual property. As it rolls out, we'll continue partnering with experts and creative communities to learn, iterate, and continuously improve on the ways this model can empower human creativity.
Unlocking the next level of imaginative storytelling
Dreamina Seedance 2.0 will unlock new possibilities in visual storytelling by empowering creators to seamlessly craft, edit and sync high-quality video and audio content with a few intuitive steps. The model excels at turning both simple and complex prompts into cohesive content of exceptional quality, making it possible to turn great ideas into original, authentic content simply by having the words, images, or reference videos to describe it.
This makes AI-empowered creativity more accessible to creators of all kinds. It also gives professional creators and small businesses the creative controls to efficiently transform ideas into professional and polished content, reducing editing times so they can focus their energy on growing their businesses.
Inspiring creative communities
As we launch Dreamina Seedance 2.0 in CapCut for the first time, our approach is inspired by insights from early beta tests and collaboration with trusted community members, whom we'll keep working with closely to test new or updated experiences as we continue to iterate.
Their early feedback showed how this technology can inspire and empower human creativity through every step of the creative process. For example:
• Dreamina Seedance 2.0's exceptional text prompt adherence empowers people to create rich, detailed content even without reference images—whether the prompt is just a few words, or a multi-paragraph description with every detail of their creative vision.
• Especially when combined with seamless audio syncing, this makes it particularly helpful for creators in the early stages of ideation, who want to test out potential ideas based on early concepts or sketches.
• Dreamina Seedance 2.0 is exceptionally good at rendering realistic textures, movement and lighting across a wide range of visual perspectives and angles, so people can use it to precisely edit, enhance, or correct their own uploaded footage.
• Its sophisticated understanding of physics and motion also makes the model particularly useful for science, motion or action-focused content types that have historically been challenging for video models.
• The model's strong text and graphic rendering abilities make it easy to create explanatory videos of anything from cooking recipes, to fitness tutorials, to business or product overviews. This embeds it as a potentially game-changing tool for small businesses, freelancers, or other professionals.
Over the coming weeks, we'll be introducing the model into a range of CapCut features—from editing features such as AI Video, to tools like Video Studio that focus more on pure generation. We're also excited to bring it to our AI generation platform Dreamina, and our professional marketing platform Pippit. Across these products, the model will support clips of up to 15 seconds long in 6 aspect ratios, making it easy for creators to translate and share their ideas across platforms and audiences.
Innovating our safeguards for advanced model capabilities
As an all-in-one editing platform, CapCut has become a go-to resource for creators seeking intuitive tools to bring their ideas to life. As Dreamina Seedance 2.0 brings new video and audio capabilities to beloved editing features, the model has undergone extensive testing by global safety and intellectual property teams who collaborated with a third-party redteaming partner. We're also incorporating strong safeguards against the unauthorized use of peoples' likeness and intellectual property. These include:
• Restricting certain capabilities for user accounts in this initial rollout, including the ability to make videos from images or videos that contain real faces.
• Implementing advanced technologies designed to block the unauthorized generation of intellectual property, which we will continuously bolster and expand on through proactive monitoring, in-app reporting, and direct collaboration with creatives and rightsholders.
• Introducing an advanced "invisible watermarking" technology that will help us identify content made with this model and take action against it when needed, even if it has been shared or altered off our platform.
CapCut's continuous investment in safety by design
These new safeguards build on the measures CapCut already has that are designed to protect our community, advance AI transparency, and prevent harmful content, intellectual property infringement and privacy violations through application of our rules.
• To give viewers transparency into how content was created, we label AI-generated content with visible watermarks and embed C2PA Content Credentials, a technology that helps other platforms or individuals identify AI-generated content.
• We prohibit harmful content including nudity and explicit content, shocking and graphic content, and harmful misinformation. We apply these policies through multiple layers of moderation with both advanced technologies and expert human teams. This includes reviewing content before and after it's uploaded to our platform to find and take action on violations.
• We also prohibit unauthorized use of intellectual property and take steps to prevent unauthorized content from being uploaded or created on our platform.
• We build specific safeguards into individual features by design to prevent their misuse. For example, when people use generative AI features, we block prompts or source material that violates our policies, and use technology to review generated content before it's shown to users.
Iterating with experts and our community
AI is a rapidly evolving technology, and there is always more to learn about the ways it can be used. As we expand access gradually, we'll continue partnering closely with experts, intellectual property rightsholders, and our creative community to build responsible, innovative tools that help people bring great ideas to life.