Last week, OpenAI released Sora, its new AI video generation platform. Like many in the industry, we were blown away by the quality of the videos that Sora can create.
Demonstration videos from Sora show people walking through digitally created cities, cute animals frolicking in the snow, beautiful art galleries, and much more.
OpenAI Sora tracking shot of a vehicle driving along a dramatic cliffside road
Credit: OpenAI
One of OpenAI's test videos even shows a tracking shot of a vehicle driving along a dramatic cliffside road. Superficially, this seems very similar to the kind of work a 3D artist might create using our HDRi Maps and Backplates.
Based on what we’re seeing, will Sora cause a major disruption in the 3D rendering space?
A Problem of Specificity
Again, we were impressed with the AI vehicle rendering that Sora created. The dust kicked up by the car tires and the lighting are rendered in a fairly accurate way.
There’s a big reason, though, why these kinds of videos won’t replace the creations of 3D artists. The biggest issue is one of specificity.
In the demonstration video , the background and environment look pretty good. But take a closer look at the vehicle.
Clearly, Sora is trying to create a Land Rover. The vehicle has some surface-level similarities to Land Rover's iconic styling. But, zoom in a little bit closer!
When you actually look at the vehicle, it doesn’t look anything like a real Land Rover. In fact, it doesn’t look anything like any actual vehicle, because it is the hallucination of a computer, not a 3D model of an actual car.
Hilariously, the vehicle in OpenAI's demo video even includes the misspelled word "DANDVRR" on the back, its attempt at spelling "Landrover."
Accuracy Is Everything
3D designers rendering real vehicles pride themselves on the deep accuracy of their renderings. Digital twins used in production rendering are detailed and accurate down to millimeter-level detail.
We’ve had designers write in to inform us if a model used in one of our scenes has the wrong color brake calipers for the trim level of the vehicle depicted. That’s how devoted to accuracy the 3D design community and our clients are.
The same goes for the accuracy of the environments in which vehicles are shown. Our RAY.HDRi Maps perfectly capture real light, rather than simply simulating it using a computer's best guesses.
The end results are 3D renders that not only look amazing but also perfectly capture the details and performance specifications of individual real-world vehicles, as well as the physical environments they’re shown in.
Because generative AI video platforms like Sora can’t achieve this kind of accuracy and specificity– and likely won’t be able to for many years–they’re unlikely to disrupt the core work of 3D designers.
The Value of Video Generators
That said, based on what we’re seeing with OpenAI's Sora, we do see a future for this tech in the 3D rendering space.
AI video generation will be incredibly useful for internal design processes and teams. It will also be extremely useful for pre-planning shots for virtual production.
In both 3D rendering of vehicles and virtual production, the final shot that the audience sees is the result of a long and involved process. Before a 3D artist even begins to create a production-ready render, or models and props are brought before an LED wall, internal design teams must meticulously plan the production they’re creating.
This involves creating mockups, concept art, rough draft videos, and much else. These are used to demonstrate what the final shot will look like and get everybody on the design team–as well as external stakeholders–on the same page.
AI-generated videos will be a powerful tool to streamline this process. Imagine that a designer has an incredible idea for an innovative shot of a new vehicle.
Previously, the designer would need to spend hours or days creating storyboards or other mockups of their ideas. With AI video generation, however, the designer could simply write a detailed prompt, and in a few minutes, have a concept video showing their idea.
This video could help to get other decision-makers and stakeholders on board or win over a client to the final idea.
In short, AI video generation will be a powerful tool for ideation. For these internal processes, the specifics matter less than capturing the general concept of a future shot.
As with AI image generators, we expect video tools like Sora to turbocharge the ideation process for design teams. When these platforms become generally available, they will help the best designers to quickly and powerfully unlock and demonstrate their creative visions.
In the future, it might be possible for AI to render specific vehicles and scenes accurately. In the meantime, these tools will be exactly that: tools. They won’t greatly disrupt the industry, but they will make the most innovative designers even more efficient and impactful.
