AI, YouTube, and the Art of Making Guns
It was a claim that seemed perfectly designed to set the internet on fire: the CEO of Embark Studios, the developer behind the hit game The Finals, mentioned in an interview that they use AI to generate 3D gun models from random YouTube videos. The reaction was immediate. Gamers and artists alike jumped into a heated debate about AI art, job replacement, and the ethics of using online content to build commercial assets. It seemed like another front in the ongoing war over generative AI in creative fields.
But as the dust settled, a clearer picture emerged, thanks to some internet sleuthing by the community. It turns out, the reality is far less controversial than the initial buzz suggested. The technology in question isn’t the kind of ‘type a prompt, get a model’ generative AI that many imagined. Instead, it’s a much more established and, frankly, less magical process that has been part of game development for years.
Not Generative AI, but Photogrammetry
The confusion stems from the broad and often hyped-up use of the term “AI”. Further digging into old blog posts from Embark from as far back as 2020 revealed that the tool being referenced is a form of photogrammetry. In simple terms, photogrammetry is a technique that uses multiple photographs or, in this case, frames from a video, to capture and reconstruct a 3D object or environment.
Instead of an AI ‘dreaming up’ a new gun model from scratch, this tool analyzes the video of a real-world object from different angles and creates a 3D point cloud, which is essentially a digital scaffold of the object’s shape. This isn’t a finished, game-ready asset. It’s a rough starting point, a piece of digital clay that an artist must then take, clean up, retopologize (rebuild with clean geometry), texture, and animate. It automates the initial reference-gathering and basic modeling phase, but it doesn’t replace the artist.
Old Tech, New Buzzwords
If this sounds familiar, it’s because it is. The team at Embark is made up of many veterans from DICE, the studio behind the Battlefield series. They were pioneers in using photogrammetry to create the stunningly realistic environments and assets seen in games like Star Wars: Battlefront. They would travel the world, take thousands of photos of rocks, trees, and historical artifacts, and then use that data to build their game worlds. The process Embark’s CEO described is essentially the same principle, just applied to a video found online instead of photos taken in the field.
So why the AI framing? It’s likely a case of a CEO using the hottest tech buzzword to make a standard, albeit advanced, development tool sound more cutting-edge than it is. In the current climate, anything involving machine learning is often lumped under the catch-all term “AI,” leading to misunderstandings.
The Lingering Ethical Question
Even with the clarification that this isn’t generative AI, some ethical questions remain. Is it okay to use a random YouTuber’s video as the source for a commercial asset, even if it’s just for photogrammetry? It’s a gray area. The argument is that it’s akin to an artist watching a video for visual reference to model an object by hand. The process doesn’t steal the video’s content, but rather extracts dimensional data from it.
This is fundamentally different from large generative AI models that are trained on vast datasets of images scraped from the internet, often without the original creator’s consent. In Embark’s case, the video is being used as input for a specific task, not as training data to build a model that can then generate infinite content. Still, it’s a conversation worth having as these tools become more common.
Conclusion: A Tool, Not a Replacement
The story of Embark’s “AI” gun models is a perfect example of how the hype around artificial intelligence can create confusion and controversy. What was initially perceived as a major leap into AI-generated assets turned out to be the clever application of an established technology. It highlights that these tools are becoming more integrated into development pipelines, not as replacements for artists, but as powerful assistants that can speed up the initial, often tedious, stages of asset creation. It’s a tool that helps get a rough idea into the game quickly for prototyping, leaving the crucial work of artistry and polish firmly in the hands of human developers.
FAQs about AI and 3D Modeling
1. Did Embark Studios use AI to generate guns for The Finals?
They used a tool that leverages machine learning, but it’s not “generative AI” in the popular sense. It’s a form of photogrammetry that extracts 3D data from videos to create a rough base model, which artists then have to finish manually.
2. What is photogrammetry?
Photogrammetry is the science and technology of making measurements from photographs. In game development, it involves taking many pictures or a video of a real-world object and using software to reconstruct it as a 3D model.
3. Is this a new technology in game development?
No, not at all. Photogrammetry has been a key part of creating realistic assets in AAA games for over a decade, famously used in titles like Star Wars: Battlefront and The Vanishing of Ethan Carter.
4. Does this process replace 3D artists?
No. The process only creates a rough, unfinished 3D scan. A significant amount of work is still required from a skilled 3D artist to turn this raw data into a functional, optimized, and beautifully textured game-ready asset.
5. What was the controversy about using YouTube videos?
The concern was about the ethics and legality of using someone else’s content (a YouTube video) to create a commercial product. While not the same as using art for AI training data, it raises questions about intellectual property and fair use in the context of data extraction.
6. Why did the CEO call it AI?
“AI” has become a popular and broad marketing buzzword. Since photogrammetry software uses machine learning algorithms to process the images, it can technically be described as a form of AI, even if it’s not the ‘generative’ type that people commonly think of today.