ByteDance Launches Seedance 2.5 Video Model on Runway

ByteDance has integrated its latest video generation model, Seedance 2.5, into Runway's platform, making it available to all users on paid plans. The updated model represents a shift from generating isolated clips to facilitating cohesive creative projects. It is accessible through Runway's Custom mode, Workflows, and Agent interfaces, offering creators a more robust toolset for long-form narrative generation.

The most notable upgrade in Seedance 2.5 is its ability to generate 30-second video clips in a single pass, which doubles the 15-second limit of the previous Seedance 2.0 model. It also supports multi-round extensions for even longer sequences. Furthermore, the model dramatically expands its input capacity, allowing up to 50 multimodal references per generation. Users can now feed the model 30 images, 10 video clips, and 10 audio clips, a massive increase from the nine images and three clips supported by Seedance 2.0.

However, these expanded capabilities come with specific trade-offs in resolution and cost. Seedance 2.5 on Runway is limited to outputs of 480p and 720p resolution. For projects requiring 1080p output, practitioners must still rely on Seedance 2.0. In terms of pricing, running Seedance 2.5 costs 30 credits per second at 720p resolution. To manage budgets during the early stages of production, creators are encouraged to use the cheaper Seedance 2.0 Fast or Mini variants for drafting and iteration.

For video editors and AI practitioners, these updates change how complex scenes are structured. By utilizing the unified multimodal audio-video architecture, creators can maintain better narrative and stylistic consistency across longer sequences. The ability to inject dozens of reference assets directly into the prompt pipeline reduces the need for tedious post-production stitching, even if users must navigate the current resolution limits by co-existing both model versions in their workflows.