ByteDance launches Seedance 2.5 AI video model

ByteDance has unveiled a significant upgrade to its AI video generation technology, announcing the new Seedance 2.5 model at a conference in Beijing this week. The company also launched a more affordable, lightweight version of its current model, Seedance 2.0 Mini, on a separate crypto-native platform.

The flagship Seedance 2.5 model represents a substantial step forward in capability. It can generate single video clips up to 30 seconds long without requiring post-production stitching, complete with scene changes and tempo shifts. A key enhancement is its ability to process up to 50 reference inputs-such as images, videos, or audio files-a major increase from the 12 references supported by its predecessor, Seedance 2.0. This expanded capacity offers creators greater control over character consistency, style matching, and scene composition. The company stated the model is set to launch in China in early July, though a release timeline for other markets was not provided.

Concurrently, ByteDance has made its existing video generation technology more accessible. The company launched Seedance 2.0 Mini on the Venice.ai platform, a privacy-focused AI inference service backed by the VVV...