ZD 25.31: Advances in Video Generation
Stormtrooper Greg wants you to know that AI video generation is advancing at an astounding pace.
Short form video is full of wild AI-generated videos of Stormtroopers, Sasquatch, Sacagawea, George Washington and more. Clearly something changed to make this possible. How much better has video got? And what tools should you try? Or just head down to the Distilled Spirit for some insights into energy, summer market madness and leadership. Enjoy a big band, show tune musical coda for added flavor.
In 2025, Everyone Is a Director
If you are someone with lots of creative ideas and limited artistic talent like me, it is an amazing time to be alive. The tools you can use to bring your imagination to life are truly astounding. An 8 second clip of video that would have cost thousands of dollars to shoot — if you could even get into the room with the people capable of creating it — can be had through a number of tools. Visual content creation has come of age. Multiple vendors have made major leaps in video production and tooling. Here are a few ways you can get started on your AI-powered video production today, and a few places to look for inspiration.
Google: Flow and NotebookLM
Google has created some amazing video tools. It makes great stormtrooper vlogs, very lifelike yetis and Sasquatch and other entertaining characters. The website itself is a work of art, run your mouse over the world Flow to see that in action. Beyond the front door, the tooling is really cool. Google’s intro video sets up the suite nicely. Where flow shines is the scene creation. It is pretty easy to get started, and it makes it easy to stretch your AI-generated talking movies out to reasonable lengths, albeit in 8 second chunks.
The Google videos have audio — they will say what you want. There are pretty decent background sound effects.
Google uses this superpower to combine video and audio in other ways. NotebookLM just added a video generation feature that is pretty good. It did a great job explaining this publication:
Making Sense of Chaos
NotebookLM is an amazing tool. It got even better this week with the release of a new video overview feature. You read that right — it now creates video summaries! The scripting and voice acting are first rate, and the images are simple and supportive. Really great out of the box, I can’t wait to see what they can do down the road.
If your goal is to hop in and make cool stuff immediately Google’s stack is hard to beat. Flow is simple yet powerful, Gemini understands you pretty well and the ability to generate realistic audio is astounding.
OpenAI: Sora
Flow was a response in part to OpenAI’s own image studio, Sora. You can also get to Sora from ChatGPT — it is an icon that is on the upper left at the time of this writing. It comes with your paid ChatGPT account.
The most fun part of Sora is the home page where you see everyone else’s artwork. By default yours will appear here too — make sure to turn that off in the settings if you value privacy. Beyond the leaks the package is fun and easy to use. OpenAI image generation is painfully slow. Sora enables up to 4 at once. Moreover, the UI enables easily remixing in source images to make much better outputs. It is what I generally use for this publication’s artwork.
For the video side, the output is great visually but very silent. That puts Sora quite a bit behind Google’s Veo3 at the moment. It just does not feel real without the audio, no matter how good it looks. For final output it needs to be combined with other tools to make it effective.
Runway
Runway is focused solely on media creation, for media creators. As such it is a little pricier and a bit more focused than the other tools. They released two new features in the past few weeks that are downright amazing and worth mentioning. Act Two motion capture lets you transplant the motion of one video into the visuals of a character from another image. It tracks the face, head and body and the results are astounding.
Aleph is completely new — it is chat based video changes. You can drop in a file and say “add rain” and it will do so. It is still limited — it only handles 5 seconds of video, there is no sound. But if they can enhance this tool with those features this will be the best of the bunch.
Descript
Descript is not a video generation tool like the others, rather it is an AI powered editing suite. If you need to do editing beyond the limited scene management tools this is your ticket. It is a powerful video editor that lets you change video with scripts. It has its own AI tool — Underlord — that now automates a lot of work. It can also give you feedback on your scripts and help you make a better video. If you have some old footage, the pull media clips feature is outstanding and worth checking out. Descript is a handy tool if you need to take the AI generated outputs the extra few yards and make the videos publishable.
Other Tools
There are a ton of other video tools on the market I have not had a chance to try.
Start with Why Try AI’s beginner’s guide to AI video. Check out Heather Cooper’s post on how AI Video Tools are Evolving for some other options. Finally, this video for a really good idea of how you can use these tools -- and how they can be strung together -- to produce a real, usable video clip:
What Is Video Worth Now?
When anyone who can write a prompt can call themselves a director, does video lose prestige? In practice, the opposite is happening. Distribution keeps expanding—social feeds, streaming platforms, storefront screens—so the hunger for compelling motion grows. AI destroys technical gatekeeping but not narrative craft. Crafting a through‑line, knowing when to cut, and coaxing emotion from pixels remain human skills.
Musical Coda
The Distilled Spirit
Your Money and The World
Some ways to think about investing in a pretty crazy time.
Energy for AI and Everything Else
⛽ Why EVs Won’t Reduce Overall Oil Demands ()
Petroleum production is a complicated business, and transport and auto manufacturing more so. Electric Vehicles are not a magic bullet to end fossil fuels.
☢ Standard Oil of Nuclear ()
Meet the innovators working hard on building nuclear power in America.
💂♀️ Quiet War for AGI Infrastructure ()
The paranoid case of AI infrastructure construction.
🏭 AI Energy Transition ()
Is AI pushing us towards a hard wall of energy production reality?
Summer Markets
📊 How to Read a Cashflow Statement like an Investor ()
How to look at a cashflow statement and get to the bottom line.
🏖 Slow Burning Summer Crashes (The Diff)
Wall Street goes to the beach in August. And they talk at those beach houses.
🤐 Data Denialism Won’t Fool Anyone ()
A dive into revisions on labor market numbers.
📊 It is Difficult to Hide the Truth ()
Today’s reminder that it is really, really hard to hide big economic numbers.
😈 Betting on Chaos ()
To some extent, there has always been a for-profit prediction market. Now there is a direct one. Learn more from a professional forecaster.
Self Improvement
Here are some ideas for self-improvement and different ways to think about problems.
Thinking Better
🏃♂️ ADHD is a Personality Trait ()
Kids who can’t sit still might not need psychoactive substances after all.
👨💻 How I Built My Second Brain in Three Hours ()
You can make an automated 2nd brain with Notion, Make and Gemini very easily.
😁 Teaching ChatGPT to Boost My Joy ()
ChatGPT can help you get to the root cause of your unhappiness and help build an action plan.
Leading Better
🎲 Building Lasting Experiences ()
How to build a lasting experience through the lens of game design.
⚔ Ricover Corpus ()
Admiral Rickover built the nuclear navy. He has many leadership lessons to share, and they are now in a handy searchable corpus.
🤖 Why Most AI Implementations Fail ()
Most AI implementations fail for a few predictable reasons. Learn about them, and what you can do different, here.
Artificial Intelligence Today
What to think about AI this week.
Building With AI
✈ Enough Co-Pilots We Need HUDs (Geoffrey Litt)
Having a co-pilot to ask questions is useful, having a system that warns you directly is more powerful.
👁🗨 How Cursor Serves Billions of Requests ()
Ever wondered how a complicated AI app brings all the threads together?
Agentic AI
ChatGPT agent was some hot news this week. Here are some ways to think about your new toy.
📖 Everything I’ve Learned About Agent So Far (Leon Furze)
Did you want to know what agent runs on and what it is capable of? This is your guide.
👨 Agent Does Not Think They Are a Robot (Ars Technica)
Agent clicks through CAPTCHA’s with glee. They are likely to get more annoying soon.
✉ MCP: Stop Playing Messenger to Your AI ( )
A very good guide to getting started with MCP and Claude. If you have Claude pro you should really check this out.
🥼 Robotics Levels of Autonomy (Semi Analysis)
An excellent explainer of the various levels of robotics technologies and the current challenges in the field.
Did you enjoy reading this post? Hit the ♥ button above or below because it helps more people discover great Substacks like this one and it helps train your algorithm to get you more posts you like. Please share here or in your networks to help us grow!
Wyatt, kind thanks for the mention 🙏
Thanks for the mention, Wyatt!