Hailuo AI

What it is

Hailuo AI (Chinese: 海螺AI) is a generative AI product from MiniMax, a Chinese artificial intelligence company ^[2]. MiniMax launched it in March 2024 as a platform for text and music generation ^[1]. A product of this kind does not search an archive for material that already exists; it produces new text or new audio in response to a user's instruction. That is what makes such tools useful for drafting and experimenting, and it is also what makes them legally complicated, because the output can resemble something a person or a company already owns.

From text and music to video

In September 2024, MiniMax released video-01 under Hailuo AI, a model that generates video ^[1]. Video generation is a harder engineering problem than text generation. A video model has to keep the appearance of people and objects consistent from one frame to the next, and it has to make movement look as though it follows ordinary physics; small inconsistencies that would pass unnoticed in a single image become obvious once the image moves.

On 20 January 2025 the product gained audio functions ^[1]. Later Hailuo models include Hailuo 2.3, which appears among MiniMax's listed products ^[1]. The order in which the capabilities arrived — text and music first, video next, audio added afterwards — describes a product that grew from a single generation tool into a family of models covering several kinds of media.

Logos and characters

In December 2024, Broadcast reported that Hailuo AI could reproduce the logos of Channel 4, Channel 5 and ITV ^[1]. Requests that return a broadcaster's logo sit close to the centre of how generative models are built: a model able to produce a recognisable mark has, in some form, learned it from material it was shown. Such outputs attract the attention of rights holders because a logo identifies a particular organisation and is protected separately from the programmes it appears alongside.

The lawsuit

In September 2025, Disney, Universal and Warner Bros. Discovery sued MiniMax for copyright infringement in the United States, alleging that Hailuo AI infringed their characters ^[1]. The Standard reported that the suit was filed on 16 September 2025 in a California district court and was directed at the Hailuo AI service ^[2]. The studios alleged that Hailuo had been built on their intellectual property and that it had been marketed as a "Hollywood studio in your pocket" ^[2].

The case turns on questions that courts are still working through: whether training a model on protected works is itself infringement, and how far a service is responsible for what its users ask it to generate. Arguing that a product presents itself as a substitute for studio production brings the marketing, and not only the training data, into the claim.

Where it sits

Hailuo AI is the name under which MiniMax has released successive generation models, from the first text and music platform through video-01 to Hailuo 2.3 ^[1]. For people who use tools like this, the practical questions are how good the output is and how much control the user has over it; those questions are settled by trying the models rather than by reading about them. For the wider field, the dispute over Hailuo shows that capability and legal exposure tend to arrive together: the same ability to produce convincing video of a familiar character is what makes a rights holder take notice.