ByteDance's New AI Video Model Can Make 30-Second Clips From a Single Prompt
Short summary
ByteDance introduced Seedance 2.5, its upgraded AI video generator, in Beijing. The model accepts up to 50 reference files and produces 30-second 4K clips, though legal and copyright questions may slow its wider rollout.
- Seedance 2.5 accepts up to 50 reference files (up from 12 in Seedance 2.0) to generate 30-second, 4K clips from a single prompt
- China rollout planned for next month, with no international launch date set, partly due to past Hollywood pushback over alleged copyrighted training data
- Lack of clear watermarking could spark legal issues around deepfakes and misinformation as the output nears photorealism
Full summary
ByteDance has unveiled Seedance 2.5, its newest AI video generator, at a Beijing conference. The standout upgrade lets you feed the model up to 50 reference inputs—images, videos or audio clips—up from just 12 in Seedance 2.0. That jump in reference material gives creators tighter control over the look, feel and pacing of their 30-second, 4K outputs generated from a single text prompt.
The company has already impressed with earlier Seedance versions, matching strides with OpenAI’s defunct Sora and Google’s Veo 3. ByteDance plans to roll out Seedance 2.5 in China next month. There’s no word yet on international availability; past U.S. launches faced hiccups when Hollywood groups pushed back over alleged use of copyrighted training footage.
Watermarking remains a key concern. As AI-generated clips inch toward photorealism, developers and regulators want built-in markers to flag deepfakes or misleading content. If ByteDance can’t bake in clear identifiers, Seedance 2.5 could trigger fresh legal headaches—especially in markets sensitive to copyright and misinformation.