ByteDance Drops Seedance 2.5 With Massive Reference Boost
ByteDance's new AI video model now pulls from up to 50 reference materials to generate 30-second clips.
ByteDance just unveiled Seedance 2.5, its latest AI video generation model, at a conference in Beijing. The big upgrade: it can now generate 30-second video clips from up to 50 reference materials.
That's a significant jump from Seedance 2.0, which topped out at just 12 reference inputs. More references means the model has substantially more source material to work with when producing video output.
The announcement positions ByteDance deeper into the increasingly competitive AI video generation space, where the ability to synthesize coherent clips from multiple inputs is becoming a key differentiator.
Details beyond the reference material capacity and clip length remain sparse. But the leap from 12 to 50 inputs signals ByteDance is pushing hard on making its video AI more capable and versatile with each iteration.