1/ 🎬 What happens when NVIDIA teams up with Stanford? You get an AI that can generate full-blown 1-minute cartoons—yep, like an actual animated Tom & Jerry-style episode—from just a text prompt. It’s wild, fast, and might just shake up animation forever. 🧵👇

🤖 Software Dev by day and AI & Crypto
⭐ I love sharing all the new tech developments with you
📫 DM me for collaboration
🌐 AI MARKET WATCH --> t.co/WMNyhRhSAO

3/ 🍿 And get this—they trained it on 81 episodes of classic Tom & Jerry. That’s why the output feels so “cartoon logic” accurate. Think slapstick chases, over-the-top animation, and even scene continuity from one shot to the next—all generated from a single prompt.

4/ 🗯️ Example prompt? “Tom works in a NYC office and chases Jerry across desks.” The model outputs a full scene, complete with camera motion, props, and pacing. No human animator touched it. This could totally change how cartoons, games, or even TikToks are made.

5/ NVIDIA + Stanford just opened the door to fully AI-generated cartoons, and the tech’s evolving fast. Whether you’re hyped or a little uneasy, there’s no denying it’s a glimpse at where storytelling is headed.

https://test-time-training.github.io/video-dit/assets/ttt_cvpr_2025.pdf

https://www.cartoonbrew.com/tech/nvidia-mined-hours-of-classic-tom-jerry-shorts-to-generate-new-ai-horrors-246696.html

https://www.livemint.com/ai/tom-and-jerry-enter-ai-era-one-minute-episode-by-nvidia-and-stanford-sparks-debate-11744538509236.html

2/ 🧠 The secret sauce is something called Test-Time Training (TTT). Basically, it helps the AI remember what happened earlier in the animation—so characters don’t randomly disappear or start floating mid-chase. Long-form AI video just got a serious upgrade.