Text-to-image artificial intelligence systems are currently booming in both capability and popularity, and what better proof of that than their presence in the world’s hottest app? That would be TikTok, a video platform that recently added a new effect it calls the “AI Green Screen,” which allows users to enter a text prompt and then the software generates an image. This image can then be used as a background for the video – a potentially very useful tool for creators.
Compared to state-of-the-art text-to-image models such as Google’s Imagen, OpenAI’s DALL-E 2 or Midjourney’s software of the same name, the output of the TikTok system is fairly basic. It only creates fairly abstract images; the case of TikTok’s suggested prompts such as “astronauts in the ocean” and “galaxies of flowers” reflects this approach. In contrast, other models produce both realistic images and complex and coherent illustrations that look like they were drawn or painted by humans.
However, the limitations of the TikTok model are likely intentional. First, more advanced models require more computing power, which would be expensive and resource-intensive for the company to implement. Second, TikTok has more than a billion users, and giving all those people the ability to create realistic images of anything they can imagine will almost certainly give some people a few unsettling thoughts again.
As the media tested its model’s ability to create nudity and gore, the text-to-image generator tended to impose limits on both of these intents from users. The abstract nature of the model’s output means that prompts with provocative language can only produce a few meaningless swirls.
Similarly, requests involving nudity only produced some appropriate colors, but nothing that would make anyone blush.
The emergence of TikTok’s “AI green screen” is notable for showing the increasing speed at which the technology is entering the mainstream. The latest cycle of text-to-image AI arguably began in 2021, when OpenAI initially released DALL-E. Less than two years later, the technology is already in the hands of millions of people through apps like TikTok.
