ByteDance Unveils Seedance 2.5: 4K Video Generation with Up to 50 References
ByteDance has revealed Seedance 2.5, the latest iteration of its AI video generation model, at a technology conference in Beijing. According to reports, the new model can generate 4K-resolution videos up to 30 seconds long from a single prompt, with the number of reference materials a user can provide increased from 12 to 50, covering images, videos, and audio files.
Expanded Reference Support
A major upgrade from Seedance 2.0, Seedance 2.5 raises the reference material limit from 12 to 50. These references can be images, videos, or audio files, giving users substantially more control over the video creation process. The model generates 4K video at 30 seconds per prompt.
ByteDance has consistently released AI video generation models that rival OpenAI’s now-discontinued Sora and Google’s Veo 3 in capability, establishing itself as a serious competitor in the space.
Release Timeline and Platforms
According to AI-focused publication Testing Catalog, Seedance 2.5 could launch as early as July 9, though that date remains unconfirmed. An official landing page for Seedance 2.5 is already live on the Dreamina website, and CapCut posted about the impending release on July 4.
The model is expected to debut in China next month, with broader availability to follow.
Copyright Concerns
The new model may face copyright scrutiny similar to its predecessor. The US rollout of Seedance 2.0 was delayed earlier this year under pressure from Hollywood over concerns that copyrighted works had been used in training data. If Seedance 2.5 delivers significant quality improvements without addressing those underlying legal questions, it could provoke a similar backlash.
ByteDance did not immediately respond to requests for comment.
Full report at CNET.
