I don’t normally write a Sunday edition, but Monday is a holiday and everything will be different by Tuesday. Or at least, that’s how it feels with the current pace of change in AI video generation. Open AI hasn’t even released their tool yet, and everyone is writing about how it will the thing that destroyed the internet and/or the next gigantic benefit to humanity. I’ve been studying generative AI tools for the last year and thought I had my finger on the pulse, but I never even saw this coming. I spent much of Friday and Saturday evening trying to get caught up on my reading to try and make sense of this. If you’re curious about the potential ramifications of this announcement, I have a few observations below.
Increased Efficiency for Studios
The immediate and most obvious benefit is for established studios. From a small agency shooting TV commercials all the way up to Hollywood studios; they can now reduce certain costs because segments that had previously been shot by real production crews can now be simply generated. Anything that forms an establishing shot, a transition, or a background element can now be created by a text prompt. While the talent cannot be faked (yet), any business that wants to reduce it’s bottom line would at least consider using AI generated snippets to reduce costly shots. When OpenAI released a few dozen clips on Thursday, several commentators honed in on the drone footage of the California coast. Why hire a drone specialist for 5-10 seconds of final output when you can just generate it?
Here’s the prompt:
“Drone view of waves crashing against the rugged cliffs along Big Sur’s Gamboa Point Beach. The crashing blue waters create white-tipped waves, while the golden light of the setting sun illuminates the rocky shore. A small island with a lighthouse sits in the distance, and green shrubbery covers the cliff’s edge. The steep drop from the road down to the beach is a dramatic feat, with the cliff’s edges jutting out over the sea. This is a view that captures the raw beauty of the coast and the rugged landscape of the Pacific Coast Highway.”
Here’s the video The quality is incredible
The End of Stock Video
On the subject of drone photography, many major cities have outlawed or limited the use of drones for obvious (nuisance) reasons. If I’m making a film about Paris, it’s much easier for me to purchase drone footage from a stock video site. There are even plugins for Adobe Premiere that allow you to browse stock footage from within the interface. Obviously, there are plenty of people who still make a portion of their income by producing content for sites like this. And even though Sora hasn’t even been officially released yet, it’s clear that their will be a huge drop in demand for this type of work in the same way that tools like Midjourney have effected the stock photo industry. Yes, I know that there are now AI stock image sites, but they payout is usually under a dollar per image. I’m sure we’ll quickly see the arrival of Sora-driven stock video sites, but the payouts will soon plummet.
An Amazing Opportunity for DIY Creators
However, if you’re an aspiring film maker, content creator, or someone trying to start your own agency, you have an accessible pro tool at your disposal. Just like the iPhone 13 bringing down the cost of production, you’re no longer required to have access to $10,000 of equipment to make a film. To be clear, I don’t think AI content can replace 100% of what you need, and you would still need to know how to write and shoot and edit, but if you’re resourceful, you could create a great TV commercial for tenth of the cost of those larger studios I mentioned in my first paragraph. Or, if you just wanted to promote your own brand, you could easily set up a production system to make content for your Tik-Tok, IG Reels, or YouTube shorts. You could make a month’s worth of content on a Sunday afternoon and automate it’s delivery.
Bonus: Check out the RunwayML AI Film Festival: Enter here | Last year’s Winners
This represents the state of the art of AI filmmaking in early 2024, the end of the pre Sora era.
This is Going to Destroy The Competition
I don’t know if we can even call the “AI Film Industry” an industry yet. It probably feels like a thing to me because I read about it every day. The people over at Curious Refuge would agree because they’re running a school and and a YouTube channel, but maybe we live in the same bubble. It was as recently as December that I was waxing poetic about Pika Labs and it’s new web based video creation tool. And in January, I had swung back to RunwayML because of it’s ‘Multi Motion Brush’, despite their high costs and low success rate. Before this week, these tools could generate four seconds of video with the possibility to extend it seven seconds. But those extended seconds frequently looked distorted or out of place. Sora can generate up to 60 seconds, at 1920 x 1080, and it’s coherence and quality is several cuts above the existing tools. To be blunt, all the current AI video generation tools now look like toys in comparison.
The Tidal Wave of Deepfake Content
If you have a keen eye, or work in design and video, you can still easily tell that these videos are AI generated. However, the average person, whose looking at it for only 10 seconds on social media won’t be so discerning. While OpenAI says that currently ‘red-teaming’ the product to minimize abuse, the genie is already out of the bottle. The Taylor Swift saga of two weeks ago already showed that we are past the deepfake tipping point and this will usher in a new sewer of garbage content on the internet. As if modern life wasn’t exhausting enough already, you will now need to always be on your guard, and scrutinize every piece of ‘amazing’ content that you find. While OpenAI has made an amazing technical breakthrough, I think it will have negative consequences.
Here is some additional reading I referenced while writing this post.
Lucas Ropek of Gizmodo has an excellent article on this.
PetaPixel also asks, “Why would anyone want to design a product that can do this”
Ars Technica has a more nuanced take as well as a ‘how they did it’ explanation
Michael Malewicz explains why quality content will be harder to find
That’s it for today! I hope you’re having a fantastic three day weekend and you’re learning a lot and staying creative. If you have any specific questions, you’re always welcome to message me. And if you forward this newsletter to your friends, I’ll be really grateful.