Who knew the classic cat-and-mouse chase would get a high-tech makeover? A team of brainiacs from NVIDIA, Stanford, UCSD, UC Berkeley, and UT Austin have deployed an AI system that’s pumping out one-minute Tom and Jerry episodes. Just like that. From text prompts.

This isn’t your average AI experiment. The system uses something called Test-Time Training layers integrated with pre-trained Transformer models. Sounds fancy, right? It is. Traditional AI methods hit a wall with longer videos because self-attention mechanisms get computationally intensive. TTT-MMLP fixes that. The AI remembers context better, keeping Tom looking like Tom throughout the video.

Traditional AI chokes on long videos. Enter TTT-MMLP—keeping Tom looking like Tom, no matter how chaotic the chase.

The researchers fed the system about 7 hours of original Tom and Jerry content. Enough to learn the essentials. Tom chases. Jerry escapes. Chaos ensues. One test scenario even had Tom working at the World Trade Center. Because why not?

The internet’s reaction? Mixed, to put it mildly. The videos have gone viral, with viewers split on whether this is technological brilliance or animated blasphemy. Some people are genuinely impressed by the AI’s ability to maintain character consistency across different scenes. Others think the videos lack the charm that made the originals special. There’s something missing. Soul, maybe? Many commenters have expressed concerns about how this technology could devalue artistic styles and threaten the livelihoods of traditional animators.

You can generate entirely new scenarios with text prompts like “Tom chases Jerry through NYC office chaos.” The AI handles different camera angles and settings, creating somewhat dynamic visuals. Somewhat.

The technical achievement is undeniable. One-minute animations with consistent characters and semi-coherent storytelling generated purely from text. That’s a big deal in the AI world. This represents a significant leap beyond traditional video AI that could only produce short 10-20 second clips.

Will this replace human animators? Not yet. The stories aren’t exactly winning Oscars for plot development. But it’s a step. A weird, slightly off-putting step into a future where AI might eventually create longer, more complex animations.

Tom and Jerry, meet your digital doppelgängers. They’re not perfect, but they’re here.