AI creates 1-minute cartoons

PLUS: Murati’s Thinking Machines adds key ex-OpenAI talent

Author

Rowan Cheung
April 09, 2025

Good morning, AI enthusiasts. AI-generated video has always faced major limitations in length and consistency, but new research may have just unlocked a major leap in storytelling capabilities.

With researchers using a new method and a dataset of Tom and Jerry cartoons to create minute-long, coherent generations, the days of short, disconnected AI video clips may finally be numbered.

In today’s AI rundown:

  • NVIDIA and Stanford’s one-minute AI cartoons

  • Amazon’s new voice model, video upgrade

  • Create eye-catching thumbnails with GPT-4o

  • Murati’s Thinking Machines adds ex-OpenAI talent

  • 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

AI RESEARCH

Image source: NVIDIA and Stanford University

The Rundown: NVIDIA and Stanford researchers just unveiled "Test-Time Training," an AI technique that enables longer video generation than previously possible—with demos producing minute-long cartoon clips with improved consistency and storytelling.

The details:

  • The system generates full minute-long animations with consistency across scenes, significantly outperforming existing methods in human evaluations.

  • TTT layers work by using neural networks as memory, allowing the model to remember and maintain consistency across much longer sequences.

  • The team demoed the tech using Tom and Jerry cartoons, showing multi-scene stories with dynamic motion and character interactions.

  • This approach modifies existing video models, adding TTT layers and enabling them to handle videos significantly longer than their original capability.

Why it matters: AI video has seen some mindblowing upgrades over the last year, but one of the biggest constraints is still the length of clips and maintaining consistency across shots. This new approach could eventually unlock the ability to tell longer, more coherent stories—without having to stitch together hundreds of generations.

TOGETHER WITH TELY AI

The Rundown: With Tely AI, you don’t need an SEO expert or content team — AI does it all. It researches your industry, competitors, and website to find the best keywords for ranking on Google, delivering expert-level, well-researched articles that promote your company to customers.

With Tely AI, you can:

  • Publish up to 60 articles a month

  • See Google index results in just 2 weeks

  • Enjoy full automation for writing, SEO, and publishing

  • Work in English, German, Spanish, Portuguese, and other languages

AMAZON

Image source: AWS

The Rundown: Amazon just launched Nova Sonic, a new voice model for human-like voice interactions — alongside an upgraded Nova Reels 1.1 video model with upgraded quality and generation length.

The details:

  • Nova Sonic processes voice input and generates natural speech with a latency of 1.09 seconds, outperforming OpenAI's voice models by significant margins.

  • Sonic achieved a 4.2% word error rate across multiple languages and showed 46.7% better accuracy than GPT-4o for noisy, multi-speaker environments.

  • Reel 1.1 extends video generations to 2 minutes through both automated and manual modes, letting users craft content shot-by-shot or with single prompts.

  • Both models are available through Amazon Bedrock, with Nova Sonic costing approximately 80% less than comparable OpenAI options.

Why it matters: Amazon's one-two punch in voice and video shows the retail giant getting serious about the genAI race across the board. With its Act agentic browser tool, Alexa+’s AI infusion, and other efforts, Amazon is making a stronger case than ever for developers to give their AI stack a look over more popular competitors.

AI TRAINING

The Rundown: In this tutorial, you will learn how to use ChatGPT's native image generation to create custom YouTube thumbnails instantly — using just simple text prompts, reference images, or even rough sketches.

Step-by-step:

  1. Upload a reference image of yourself or the main subject in ChatGPT, then write a detailed prompt describing exactly what you want in your thumbnail.

  2. For style consistency, upload both a reference thumbnail you like and your subject image, then ask the AI to maintain the style while swapping elements.

  3. Refine results with follow-up prompts or use the edit feature to highlight areas needing changes.

  4. For maximum creative control, upload a rough sketch showing your layout, along with any reference images you want to include.

Pro tip: You can also use an image expander tool like Canva or Adobe’s Generative Fill to adjust your thumbnail for perfect YouTube dimensions (16:9).

PRESENTED BY SANA

The Rundown: Sana’s AI agent platform gives your team one powerful interface to build and deploy AI agents grounded in your company’s knowledge — helping tackle complex challenges with intelligence that never stops working.

Sana empowers your team to:

  • Build custom AI agents for any department—no coding required

  • Unify all your AI tools through one consistent, intuitive platform

  • Integrate easily into your current systems with enterprise-grade security

THINKING MACHINE LABS

Image source: Thinking Machines

The Rundown: Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, just added ex-OpenAI CRO Bob McGrew and GPT architect Alec Radford to its list of advisors—bringing the number of OpenAI alumni on its roster to nearly half.

The details:

  • 19 of 38 listed ‘Founding Team’ members have previously worked at OpenAI, including OpenAI’s co-founder John Schulman, leading as chief scientist.

  • McGrew departed OpenAI in September after eight years, joining Murati just months after announcing a break from the industry.

  • Radford was pivotal in creating OpenAI’s GPT technology and left the company last year to pursue independent research.

  • The startup was recently reportedly seeking to raise as much as $1B at a $9B valuation, though little has been revealed on its potential products or roadmaps.

Why it matters: Murati continues to gather her former colleagues, with a roster including many of the minds behind ChatGPT, DALL-E, and other AI breakthroughs. With all of the competition we already have in the AI space, there are still labs from multiple OpenAI leaders in Murati and Ilya Sutskever (SSI) quietly waiting in the wings.

QUICK HITS

  • 👁️ Gemini Live - Real-time visual AI on Android devices and via screen sharing

  • 🎥 Runway Gen-4 Turbo - Produce 10-second videos in just 30 seconds

  • 🔎 AI Mode - AI search with expanded visual and multilingual capabilities

  • 🤝 ElevenLabs MCP - Create automated voice agents on other platforms

NVIDIA released Nemotron-Ultra, a 253B parameter open-source reasoning model that surpasses DeepSeek R1 and Llama 4 Behemoth across key benchmarks.

OpenAI published its EU Economic Blueprint, proposing a €1B AI accelerator fund and aiming to train 100M Europeans in AI skills by 2030.

Deep Cogito emerged from stealth with Cogito v1 Preview, a family of open-source models that it claims beats the best available open models of the same size.

Google rolled out its Deep Research feature on Gemini 2.5 Pro, claiming superior research report generation over rivals and adding new audio overview capabilities.

Chinese scientists used the Origin Wukong quantum computer to finetune 1B-parameter models, seeing 15% training improvements and 76% reduction in model size.

AI2 and Google Cloud announced a $20M joint investment to power and accelerate AI-driven cancer breakthroughs with the Cancer AI Alliance’s research platform.

Snapchat debuted Sponsored AI Lenses for brands, using AI-powered advertising to transform users into personal brand moments.

COMMUNITY

Join our next workshop this Friday at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll have your own AI assistant that speeds up your workflow and helps you get more done, with less manual effort.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.

That's it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team

Reply

or to participate.