ByteDance’s video generation model Seedance 2.0, after a pause of about a week on its global rollout plan, reopened the overseas market on March 23 with a “face ban” compromise. Simultaneously, it was gradually released in different regions through CapCut and the creative platform Dreamina. The image model Seedream 5.0 Lite was also launched at the same time.
(Background: Copyright lawsuits ongoing! ByteDance urgently halts Seedance 2.0 global release plan)
(Additional context: Elon Musk praises Seedance 2.0 as “AI video development is too fast”! ByteDance believes the model is not yet perfect)
Table of Contents
Toggle
ByteDance’s AI video model Seedance 2.0, after the Hollywood copyright storm, returned to the global market with a compromise approach. However, the route of user-uploaded real face images has currently been temporarily “blocked.”
Dreamina’s official X account confirmed that “uploading real images is not supported at the moment,” and the initial version restricts users from generating content from images or videos containing real faces.
Finally here! Dreamina Seedance 2.0 and Seedream 5.0 Lite are Now Available on Dreamina AI
We’re excited to introduce our latest models, which are built for deeper control and broader creative possibilities.
What Makes Dreamina Seedance 2.0 Exceptional?
→ Multimodal video… pic.twitter.com/OfkZfloi9P— Dreamina AI (@dreamina_ai) March 24, 2026
Despite the compromise, the core strength of the model remains intact.
Independent evaluation agency Artificial Analysis conducted blind tests, showing Seedance 2.0 scored Elo 1269 in text-to-video (without audio) projects, ranking first globally; in image-to-video projects, it scored Elo 1350, surpassing Google Veo3 and OpenAI Sora in both categories.
Technically, Seedance 2.0 supports four modalities: text, images, audio, and video. It can generate videos up to 15 seconds long in six aspect ratios, with native audio-video synchronization capabilities. ByteDance positions it as an open-source, cutting-edge model.
Seedance 2.0 has already integrated with platforms like CapCut and Pippit. The initial regions of release focus on Southeast Asia and Latin America, currently limited to paid CapCut users.
Seedance 2.0 first launched in the Chinese market on February 12. Less than 24 hours after launch, AI-generated videos of Hollywood stars like Tom Cruise and Brad Pitt circulated widely on social media. Homemade videos of characters from “Stranger Things” also spread.
On February 13, Disney was the first to send a cease-and-desist letter to ByteDance, accusing it of “virtual hijacking” of Disney IP, claiming Seedance preloaded “pirated material libraries” of Disney characters like Star Wars and Marvel.
Followed by cease-and-desist notices from Paramount, Netflix, Sony, Warner Bros., and the Motion Picture Association (MPA). SAG-AFTRA, the US actors’ union, publicly condemned the issue.
On February 16, ByteDance promised to strengthen safety measures, but the global rollout plan was not immediately changed.
On March 17, US bipartisan senators Marsha Blackburn (Republican) and Peter Welch (Democrat) jointly wrote to ByteDance CEO Liang Ruo Bo, demanding the “immediate shutdown” of Seedance.
It took about a week from suspension to relaunch. ByteDance’s response was to block real face uploads while keeping other features open.
Blocking real face uploads indeed reduces the most direct infringement risks, shutting down the Hollywood star face deepfake route. However, Hollywood’s copyright concerns are never limited to just real images; preloaded IP material allegations have not been officially resolved.
A bigger issue is that Seedance 2.0’s top global ranking, combined with ByteDance’s large user base worldwide, gives this model a natural potential for large-scale infringement. Face blocking is only the most superficial layer of protection. For ByteDance, this relaunch is also a way to test the waters.