The TikTok application learned to generate images by textual description

The TikTok application learned to generate images by textual description

Artificial intelligence systems capable of generating images by textual description are gaining increasing popularity. Evidence of this is the appearance of such a tool in a popular short video service, TikTok.

Recently, a new effect was introduced on the platform called "AI greenscreen." The interaction with this tool suggests that the user in a special window will introduce a text description of the desired image, and then the algorithm will create a full image. In the future, TikTok users can use the images generated in this way as a background for the video, making the new effect useful for content authors.

Note that the capabilities of the iI-algorithm TikTok are rather modest compared to more advanced systems such as Imagen from Google or DALL-E 2 from OpenAI. Accessible users of the platform can only form abstract and unclear images. This means that the algorithm is not suitable for creating photorealistic images or complex illustrations.

The limitations of the AI greenscreen tool can be intentional because it requires more computing capacity to create photorealistic images. In addition, TikTok has more than a billion users, so it is difficult to imagine what could come to mind to draw each of them. It is likely that developers initially made sure that users are not able to create images that in one way or another violate the service rules.