Infinite cat videos.
A few months ago we were pondering a few things about the future of media here at Llama Towers. I speculated about a world where you could generate your own Star Wars trilogy based on the Timothy Zahn’s novels, fix the ending of Game of Thrones, or finish cancelled series like Firefly. I argued that copyright was increasingly ill-suited to this future, and that image and personality rights would become the real legal battleground.
It has been four months, and things appear to be moving faster towards this future than we could envisage.
SeeDance 2.0
On 10 February 2026, ByteDance launched SeeDance 2.0, and within hours the internet lost its collective mind. The model generates up to 20 seconds of temporally consistent 1080p video with synchronised audio, it handles multimodal inputs (text, image, video, and audio references simultaneously), and produces motion that ByteDance describes as “physics-aware”. What this means is that for the first time gravity works as it should, fabrics drape properly, and people move like people rather than like fever dreams; gone are the days of grainy footage of Will Smith eating spaghetti with a huge ‘Shutterstock’ watermark across the screen. We got some interesting viral experiments, Irish filmmaker Ruairi Robinson generated a clip of Tom Cruise fighting Brad Pitt on a rooftop using a two-line prompt. The Motion Picture Association issued a statement denouncing the tool for enabling copyright infringement “on a massive scale” and demanding that ByteDance cease its infringing activity immediately, and soon after SAG-AFTRA followed suit.
For context, the MPA had said almost exactly the same thing when OpenAI launched Sora 2 a few months earlier. That time, OpenAI listened and implemented safeguards. SeeDance 2.0 comes from the Chinese owner of TikTok, a company already involved in a geopolitical conflict with the US. I would not hold my breath for a similarly cooperative response.
The reaction from creative professionals was telling. Rhett Reese, writer of the Deadpool films, commented on the Cruise-Pitt video: “I hate to say it. It’s likely over for us. In next to no time, one person is going to be able to sit at a computer and create a movie indistinguishable from what Hollywood now releases.” Robinson himself seemed unsure whether to celebrate or despair, writing: “Today’s question is: should I be killed for typing 2 lines and pressing a button.”
Those with early access to SeeDance started doing what the Internet does best, generating a bunch of videos featuring whatever media you can possibly imagine, 15-second Lord of the Rings (the Eagles take the Ring to Mordor), or a 15-second fix to the ending of Game of Thrones (Daenerys doesn’t go mad). Hundreds of preposterous videos started filling my timeline.
And cat videos. So many cat videos. I’m partial to all the Catzilla and Ultraman cat content out there. I’m a llama of simple tastes.
The previous generation of AI video models, namely Sora 2, Veo 2, Runway Gen-4, Kling, already felt like they were approaching a threshold. SeeDance 2.0 has crossed it. We will soon live in a world where anyone with an internet connection will be able to generate video content featuring recognisable characters and visual styles at a quality level that would have required a professional production team and a significant budget just two years ago.
And video models are continuously improving. This is the worst the technology will ever be.
Creative industries onboard
But here is the really interesting part. While the internet was panicking about SeeDance, the entertainment industry had already been quietly making its peace with generative AI, and not in the grudging, teeth-clenched way that most people expected. It has been a common theme in this blog that a large part of the creative industry will adopt AI, and this still seems to be the direction of travel.
In December 2025, Disney announced a three-year licensing agreement with OpenAI. Under the deal, Sora users will be able to generate short-form social videos featuring more than 200 characters from Disney, Marvel, Pixar, and Star Wars. A full parade of Disney IP will be licensed for user-generated AI content. ChatGPT Images got the same access for still images. Curated selections of fan-generated Sora videos will even appear on Disney+. And to underline the seriousness of the partnership, Disney invested $1 billion in OpenAI and will become a major customer of OpenAI’s APIs for building new tools and experiences across the company.
This is the single most powerful intellectual property holder in the entertainment industry actively licensing its characters for AI-generated content. Not tolerating it. Not looking the other way. Licensing it, investing in it, and planning to distribute it on its own streaming platform.
I wrote in October that studios were “seeing what is going on and have decided that AI is where it’s at.” I thought this process would take years. It took weeks.
It’s important to point out that the Disney-OpenAI deal has several guardrails. It excludes talent likenesses and voices, a bright line that respects SAG-AFTRA’s agreements. The two companies have established a joint steering committee to monitor user creations against a brand appendix of prohibited uses. The content is limited to short-form social video, not feature-length productions. But the principle is established: Disney believes that licensing its IP for AI generation is a business opportunity, not a threat.
If the Disney deal was surprising, the music industry’s transformation was even more dramatic, largely because it happened so quickly after what had been very aggressive litigation.
In June 2024, the three major record labels banded together to sue Suno and Udio, the two leading AI music generation platforms, for mass copyright infringement, but the lawsuit lasted about a year.
In October 2025, UMG settled with Udio and struck a licensing deal. In November, Warner settled with both Udio and Suno. Under these deals, both AI platforms will retire their current models, the ones allegedly trained on unlicensed music, and launch new models in 2026 trained exclusively on licensed works. Artists and songwriters get opt-in control over whether and how their names, voices, likenesses, and compositions are used. Warner CEO Robert Kyncl described the Suno deal as “a victory for the creative community that benefits everyone.” Sony has not yet settled with either platform, but I wouldn’t be surprised if we got some agreements soon.
What is fascinating about these deals is how it repositions the relationship between creative industries and AI companies. Suno is not being grudgingly tolerated; it’s being embraced as a new distribution channel. Users will be able to create content featuring participating Warner artists’ voices and compositions.
The AI-generated future
I think there is a deeper logic at work here that a few commentators have missed. While there are dozens of ongoing copyright cases, AI presents the creative industry with ways to cut costs and find future revenue streams based on their existing portfolios. Copyright owners are beginning to understand that generative AI makes their existing intellectual property exponentially more valuable, if they control the terms of its use that is. But that means learning from the lessons of history, the opportunities missed in the early digital music revolution.
Consider what Disney is sitting on. Decades of animated characters, storylines, visual styles, environments, props, and cultural associations. In the traditional model, you monetise that IP by producing new films and series (expensive, risky), selling merchandise (modest margins), and operating theme parks (billions in capital expenditure). Each revenue stream requires massive investment and carries significant downside risk. Most sequels underperform. Most merchandise lines are modest earners.
But in a hyperpersonalised AI future, the same IP becomes an almost infinitely exploitable resource with minimal marginal cost. Disney does not need to spend $200 million USD producing a new Star Wars film, they can access their existing IP, say the ‘Heir to the Empire’ novels, and they can license the characters, environments, and visual language to an AI platform and let millions of fans generate their own versions. The fans get the bespoke experience they have been fantasising about since 1991. Disney gets a licensing fee for every generation, with none of the production risk. And because the content is user-generated and short-form, it sits in an entirely different market from Disney’s own theatrical and streaming productions. There is no cannibalisation, there is expansion.
The same logic applies to the music settlements. Warner isn’t licensing its catalogue to Suno out of generosity or resignation. It’s licensing it because Suno’s millions of users represent a vast new market for monetisable interactions with their existing IP. Every time someone generates a track using a licensed model with a Warner artist’s voice or style, that is a revenue event that simply did not exist before. The artist is compensated. The label takes a cut. The user gets something that passive streaming could never offer, the experience of participating in the creation of music. Some musicians are starting to see this, just yesterday rap legend Ice-T came under fire for appearing to adopt AI.
The SeeDance problem
And this is where SeeDance 2.0 becomes so significant, because it exposes the vulnerability in the entire strategy.
The Disney-OpenAI deal works because OpenAI is a company with a commercial incentive to maintain good relations with Hollywood. It responded to the MPA’s complaints about Sora 2 and it implemented safeguards that seemed acceptable to copyright owners, which is why it struck a licensing deal. That is how the system is supposed to function. But SeeDance 2.0 comes from ByteDance, a company operating from China under enormous geopolitical pressure, with no obvious incentive to follow the same playbook. And the open-source AI video community has even less incentive to play nice.
Content owners cannot sue their way out of this. They learned that lesson with music piracy twenty years ago, and the scale of the problem with AI-generated video is orders of magnitude larger. The only viable long-term strategy is to make the licensed experience so much better, so much more integrated with the IP that people actually want, and so much more convenient, that unlicensed alternatives become the inferior option. This is essentially the insight that allowed Spotify to largely defeat music piracy.
If you can generate a mediocre Star Wars clip with SeeDance 2.0, but you can generate a stunning, character-consistent, narratively coherent Star Wars experience through a licensed Disney platform with access to the full depth of the Lucasfilm archives, most people will pay for the latter (I know I would, Disney, take my money!) The value of the licence is not merely legal permission. It is quality, consistency, depth, and access to the real thing. Nobody wants bootleg Darth Vader when you can have the real one.
I don’t know how what is the future of SeeDance, but I wouldn’t be surprised if we witness some sort of quick agreement. The problem for US-based creative industries is that the threat of crippling copyright lawsuits may be less powerful with Chinese companies. They would have to rely on political manoeuvring, but the problem is that Trump hates Woke Hollywood and won’t lift a finger to help them unless there is some serious appeasement. Enter the Inaugural Oscar Peace Prize?
Concluding
I don’t think that SeeDance will be the future, chances are in a month we will be looking at a much better model that can generate longer videos. But one thing is clear, generative AI is here to stay, and content owners aren’t going to resist this future. The AI Wars will not be endless conflict, there will eventually be some sort of future settlement.
One thing is clear. SeeDance has shown us a future in which we can potentially interact with existing media, but perhaps also build our own.
Cats. More cats. Infinite cats.
0 Comments