ByteDance's Seedance 2.0 Resumes Global Launch, Blocking Real Human Faces as a Compromise

ByteDance's AI video generation model Seedance 2.0 and image model Seedream 5.0 Lite have been globally launched via the overseas AI creative platform Dreamina. CapCut's mobile, desktop and web versions are available simultaneously, covering 7 countries including Indonesia and the Philippines initially, with more regions to follow.

Seedance 2.0 supports multi-modal inputs like text and images, generating 15-second 1080p videos. In an independent blind test, it ranked first in text-to-video category, and its image-to-video performance exceeded Google's Veo3, OpenAI's Sora and Kuaishou's Keling.

The launch has clear safety restrictions: uploading real human images is not allowed. Previously, user-generated star videos led to Disney's infringement accusations, so ByteDance paused the original plan. After a week, it resumed with the compromise of blocking real human faces.