- ByteDance shelved the mid-March global launch of its AI video model after receiving cease-and-desist letters from major studios.
- ByteDance engineers are adding safeguards to prevent the model from generating content that could lead to further intellectual property violations.
ByteDance has suspended the global launch of its latest video-generation model, Seedance 2.0, following copyright disputes with Hollywood.
The decision comes after a series of clashes between the company and several major Hollywood studios and streaming platforms, The Information reported Saturday, citing two people familiar with the matter.
The Chinese parent company of TikTok had originally planned to roll out the new multimodal video-generation model to global customers in mid-March.
The company intended to offer API access to startups and enterprise clients through its in-house cloud platform, BytePlus.
It also planned to provide a standalone app for general consumers outside of China, but these rollout plans are currently in limbo.
With no new launch date in sight, ByteDance's legal team is working to identify and resolve potential outstanding legal issues, according to the report.
Meanwhile, the company's engineers are adding safeguards to prevent the model from generating content that could trigger further intellectual property infringements.
Disney sent a cease-and-desist letter to the Chinese firm last month, accusing it of using Disney characters without permission.
Disney alleged that ByteDance pre-packaged the model with a pirated library of characters from franchises including Star Wars and Marvel.
Other major studios quickly followed suit with their own cease-and-desist letters, including Netflix, Warner Bros., Paramount, and Sony.
The Motion Picture Association called the copyright infringement systemic, arguing it was a deliberate feature of the model rather than an accidental bug.
ByteDance officially unveiled the model on February 12 this year, designing the system for professional use in film, e-commerce, and advertising.
It supports four input modalities — text, image, audio, and video — and can process these elements simultaneously to generate 15 seconds of high-quality audio and video.
ByteDance announced the model's API pricing earlier this month, with pure video generation costing 46 yuan ($6.67) per million tokens.
This means generating a 15-second video consumes about 308,880 tokens, bringing the cost of pure video generation to about 1 yuan per second.
However, the Volcano Engine website indicates that the model currently does not support broad API access and is restricted to internal use and an initial batch of invited enterprise clients.
($1 = 6.8966 yuan)
