The new ‘AI greenscreen’ filter on TikTok is the app’s attempt to jump on the text-to-image trend popularised by the likes of OpenAI’s DALL-E.
TikTok now offers a very basic text-to-image AI generator built into its app.
The new filter, which TikTok is calling ‘AI greenscreen’, allows users to type in a text prompt that it will use to generate an image. The image can be used as a background for videos on the platform, which has more than 1bn monthly active users.
AI greenscreen is being introduced on the tail of other text-to-image tools that have recently popped up, such as DALL-E by OpenAI.
DALL-E was initially created last year and OpenAI introduced the updated version, DALL-E 2, in April. The AI system can create realistic images and artworks from simple word descriptions.
While OpenAI only recently expanded early access to the tool, it inspired the creation of a similar image generator called the DALL-E mini, which was developed by machine learning engineer Boris Dayma. This open access tool spawned millions of memes and catapulted text-to-image tools to viral status.
The new filter on TikTok can be accessed by pressing the ‘Effects’ button beside ‘Record’ on the app’s home screen. Type ‘AI greenscreen’ into the Effects search bar and select the logo that says ‘AI’.
Next, a prompt will appear to type a few words. You can type in anything you want, whether it’s a cultural reference or a version of a popular meme.
Click ‘Create’ and an image will be generated based on the text. TikTok also makes suggestions for those unsure what to type.
The Verge’s James Vincent noted that these images are “rather abstract and swirling”, differing from the photorealistic images and detailed illustrations that can be provided by other text-to-image models.
TikTok has a new text-to-image AI integrated directly into the app as a greenscreen filter. The output is all so abstract, though, that it’s unlikely to cause much harm. Here are some test prompts for various political figures being assassinated… https://t.co/VfwfeyPeMB pic.twitter.com/AqKtQIRRkX
— James Vincent (@jjvincent) August 15, 2022
TikTok is not the only company to jump on the AI text-to-image trend. Google Research unveiled its AI model Imagen earlier this year, saying it has an “unprecedented degree of photorealism” and a deep level of language understanding.
Last month, Meta revealed it was developing an AI tool called Make-A-Scene. The tool takes the whole trend in a new direction by allowing users to include rough sketches to inform the AI’s results.
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.
Source by www.siliconrepublic.com
