Why OpenAI’s Sora isn’t ready to replace Hollywood
[Variety]
Anyone doubting the extreme pace of advances in generative AI would have had a shock last week when ChatGPT developer OpenAI unveiled its latest text-to-video tool. Dubbed Sora, it is a huge improvement on anything that the industry has seen before.
For example, the gif below cycles through a few frames of a 60-second sequence generated from the following prompt: A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage. She wears a black leather jacket, a long red dress, and black boots, and carries a black purse. She wears sunglasses and red lipstick. She walks confidently and casually. The street is damp and reflective, creating a mirror effect of the colorful lights. Many pedestrians walk about.
It is either hugely impressive or extremely worrying stuff, depending on your perspective — perhaps even both — and there are many more equally high-quality examples on the Sora website.
We used an unmodified Large Language Model (LLM) to generate animated content as part of a demo at IBC2023 last year, illustrating how easy it can be for pirates to create content in the future. We noted at the time that, despite their huge potential, LLMs also have various shortcomings. And most of them seem to have been addressed with Sora.
Video quality and realism have been dramatically improved in comparison to previous models, and video length is now up to a minute long, and crucially stays coherent and keeps the prompt in mind throughout the whole period. OpenAI says it will be able to extend this even further and that it has also solved the problem of spatiotemporal (space/time) consistency. This basically means that if an object goes out of view when it reappears, it will not have changed.
So, is it time for production companies to start using it in productions and operators to be thinking about creating their own trailers and interstitials? Variety is one of many that says no, not yet. Apart from the fact that OpenAI is only currently making it available to so-called red teams to evaluate any dangers it might pose, there are three main causes of concern about deploying the technology for mainstream production.
Two are technical. First, continuity. While indeed improved, this is yet to be guaranteed and is nowhere near the level needed to maintain any longer format content. The AI still sometimes gets confused about real-world physical constraints too, leading objects to behave oddly and break the illusion. Secondly, we don’t yet have the ability to move cameras, adjust lighting, and more. When we create environments in 3D software such as Unreal Engine, we have fine control over the creative elements of a scene. These sorts of controls are not yet there for Gen-AI.
And lastly, the big one that cannot be changed purely by an accelerated speed of development, copyright. Where Gen-AI-created content stands in terms of copyright law is still open to serious debate. Most models have been trained on the open internet, and the current guidance from the US Copyright Office is that entirely AI-generated works are not copyright-protectable but that the human-authored elements of AI-assisted work are. The interesting legal challenge to come, therefore, is how to separate the two.
“As a result of these constraints, in the near term, these tools are most likely to materialize during previsualization stages of a project, such as to rapidly develop and iterate concept art, character design or animatics,” writes Variety. “But even early-stage concept work is potentially not failsafe against infringement claims or questions of protectability if, for example, a studio, creative team or artist generates an interesting character or environment that’s then used in a human-created TV, movie or video game.”
MWC Barcelona 2024 preview
One of the main things we can predict for MWC Barcelona 2024, which starts next week, is that sustainability will be heading to the top of the agenda again. Barcelona is the capital of the Catalan region, and the Catalonian government has just declared a drought emergency after over three years of below-average rainfall. As a result, fountains at the Fira exhibition center in Barcelona have been emptied, water fountains removed, caterers told to conserve water onsite, and more.
While there will be a shortage of water, there is likely to be no shortage of news from the show. Though the really big names in smartphones, the likes of Apple and Samsung, stay away nowadays and concentrate on their own events, there are plenty of smaller manufacturers looking to disrupt the market with their own Android devices. Nothing, Honor, Xiaomi, Motorola, and more will all be launching new models in Barcelona, hoping to leverage some clear blue water between themselves and their competitors using innovative design.
AI will, of course, be everywhere. It was the same at CES, but of course, that is very much a consumer-oriented show, and it will be interesting to see the different use cases that are being shown in areas such as smart cities, Industry 4.0, and more.
Smart glasses are another CES echo that will repeat itself, and progress in the field is rapid enough that we could even see a slew of new models hitting the market as companies look to fill the lower price points. $600 for AR goggles sounded expensive in 2023, but with Apple Vision Pro now setting the high-end mark at $3500, $600 in 2024 looks like a comparative bargain.
There will be the usual attention on the continued 5G rollout, which is delayed enough that it is starting to have to share room with the forthcoming 6G spec. And a preponderance of European telco CEOs speaking at keynotes on Monday suggests that there might be some saber-rattling around topics such as fair share, the mobile world’s equivalent of the net neutrality debate, and their insistence that big tech should contribute to spiralling network infrastructure costs.
We will be there next week too. Meet with us in Barcelona to learn more about our multiscreen solutions and how they give your viewers the experience they expect in the age of mobile TV.
How 2024 will be the year of super aggregation in the US
A huge amount of new data about how consumer behavior is evolving thanks to the introduction of ad-tiers into the streaming market is now becoming available. The latest , big survey is titled Subscription Wars 2024 and comes from payments provider Bango.
Here are some of the key data points:
- 35% of subscribers have started paying for a service they previously accessed for free via someone else’s account
Advertising
- 35% of TV and video streamers have paid for an upgrade to avoid watching ads
- This rises to 71% for those streaming sports content
- 73% said they want one platform to manage all of their subscriptions in one place
- 69% would like the ability to pay for multiple subscriptions via one monthly bill
- Only 29% want their cable company to manage their subscriptions
- 50% say they want their mobile phone provider to launch a content hub
Paul Larbey, CEO of survey compiler Bango, commented in Advanced TV: “Subscribers want to jump between different content and services, but they don’t want the admin headache of managing multiple accounts and paying multiple bills. With the rise of Super Bundling in 2024, we’re expecting to see that headache disappear. At the same time, these all-in-one platforms will help drive new revenue for cell phone providers and allow subscription services to share users rather than fighting over them. It’s a win-win scenario for businesses and subscribers alike.”
