TikTok now offers a very basic AI text to image generator right in the app

Text-to-image AI systems are booming in terms of their capability and popularity at the moment, and what better proof than their appearance in the world’s most popular app: TikTok.

The video platform recently added a new effect called “AI Green Screen” that allows users to type a text message that the software creates as an image. This image can then be used as a background for a video – a very useful tool for content creators.

TikTok’s output is very basic compared to the latest text-to-image models like Google’s Imagen, OpenAI’s DALL-E 2, or the so-called Midjourney program. They only create somewhat abstract and swirl images; A power that is reflected in the dreamy nature of proposed TikTok triggers such as “An Astronaut in the Ocean” and “Venus Galaxy”. Other models, by comparison, can produce photorealistic images and complex, coherent illustrations that appear to have been drawn or drawn by humans.

The TikTok model only produces swirling, smeared, and smeared images. Not the latest style – but maybe for the best.
Photo: edge

TikTok form limitations may be intentional. First, more advanced models require more computing power, which will be expensive and consume many resources for the company to implement. Secondly, TikTok has over a billion users, and giving all of these individuals the ability to create photo-realistic images of anything they can imagine will surely produce some troubling results.

For example, we tested the models’ ability to generate nudity and bloodlines – two types of output that text-to-image generators try to determine. Images based on violent claims such as the “Boris Johnson Assassination” and “The Assassination of Joe Biden” produce mostly abstract swirls, with the almost recognizable face of the UK Prime Minister (although the familiar blond man’s mop makes the caricature particularly easy) .

The abstract nature of the model’s output means that stimuli using provocative language produce only vortices.
Photo: edge

Likewise, an order that includes nudity – a “naked model on the beach” – yields theme-appropriate colors, including flesh tones, sandy orange, and ocean blue, but nothing that would make a priest blush.

Trying to get the model to generate nude photos doesn’t get you anywhere.
Photo: edge

What’s remarkable about the emergence of TikTok’s “green screen AI,” then, is that it shows how quickly this technology has spread. Arguably, the latest text-to-image AI development cycle started in 2021 with the original release of DALL-E by OpenAI. Less than two years later, the technology is already in the hands of millions via an app like TikTok.

Given the potential of these systems for both harm and good, things will get even weirder from now on.

Related posts

Leave a Comment