After a one-week Hollywood-style hiatus, ByteDance's Seedance 2.0 quietly resumed global operation, at the cost of sealing off real human faces.

According to 1M AI News monitoring, ByteDance's AI video generation model Seedance 2.0 and image model Seedream 5.0 Lite have been opened to the global market through the overseas AI creative platform Dreamina (dreamina.capcut.com). CapCut's mobile, desktop, and web versions have been launched simultaneously, initially covering Indonesia, the Philippines, Thailand, Vietnam, Malaysia, Brazil, and Mexico, with further expansion to more regions.

Seedance 2.0 supports text, image, audio, and video multimodal inputs, capable of generating up to a 15-second 1080p video with native audio-video synchronization capabilities. In a blind test conducted by Artificial Analysis, Seedance 2.0 ranked first in the text-to-video category with an Elo score of 1269, and its image-to-video performance surpassed Google's Veo3, OpenAI's Sora, and Kuaishou Kelin.

The launch comes with clear security restrictions. In response to a user on platform X, the Dreamina official account confirmed that currently, uploading real person images is not supported. Previously, after its initial release in the Chinese market in February, user-generated videos of stars like Tom Cruise went viral on social media, leading Disney to accuse ByteDance of "virtual looting of Disney IP" and issue a cease-and-desist letter, followed by Paramount, Netflix, and other studios.

On March 15, The Information reported that ByteDance had postponed its planned mid-March global launch, as its engineering and legal teams needed to address compliance issues first. The interval between the suspension and resumption of the launch was about a week, with the banning of real person faces being the most visible compromise.