TikTok Launches Text-to-Image Generator AI Greenscreen
Photo by DeepMind on Unsplash

TikTok Launches Text-to-Image Generator AI Greenscreen

If there wasn’t a feeding frenzy at the text-to-image, AI-powered trough after DALL-E 2 achieved viral fame, there will be now. TikTok has added an “AI greenscreen” feature in the app, which—like DALL-E 2—lets you put in a text prompt the AI then renders in image form. This adds another tool that creators can use as a video background.


It’s not a particularly sophisticated feature yet—it renders abstract, strange images like many text-to-image applications. Still, similar models like Imagen (Google) or Midjourney can render strikingly detailed creations by comparison.

The vague abstraction of AI Greenscreen images might be intentional, given the enormous amount of computing power needed to render the images on top of TikTok’s ever-increasing popularity as a social media app in general.

A series of surreal, colorful images created with TikTok's AI Greenscreen featureA series of surreal, colorful images created with TikTok's AI Greenscreen feature

As The Verge notes, the choice to make AI Greenscreen simple and surreal is a matter of corporate safety since TikTok has over a billion users. A photorealistic AI product could lead to someone producing objectionable, offensive and legally actionable content.

However limited the tool may be compared to established AI art projects, TikTok’s adoption of AI Greenscreen marks a significant step forward for text-to-art from something that’s still a tech novelty to more mainstream usage. In addition to notable projects like DALL-E 2 or the related app Craiyon (formerly DALL-E mini), there are numerous similar projects in the works, such as Bitcoin podcaster and Tokenly founder Adam B. Levine’s Pixelmind. Still in beta, it is described as “A generative art experiment” and produces notably interesting and precisely-rendered art that easily could have come from a human hand.

There’s also Playform.io, which offers AI-generated art as a tool for human artists, and Hotpot.ai, which provides a host of tools, including an AI artmaker.

The U.S. Copyright Office has already had to address the question of whether an artificial intelligence application can copyright an image it creates, and the answer was that “human authorship is a prerequisite to copyright protection.”

Still, visual artists are growing more concerned that artificial intelligence will drive them out of work. If TikTok adoption truly kickstarts text-to-image AI art into broader usage, paying for the computing power necessary to create it will be just one of a host of new problems confronting the emerging industry.

Ex-Zillow CEO Takes Offerpad Public in $3B Deal Pitting Him Against Former Colleagues
GeekWire Photo/Kevin Lisota

Former Zillow Group CEO and dot.LA cofounder Spencer Rascoff is jumping back into the real estate game in a big way.

Rascoff, who left the CEO post at Zillow in February 2019, is part of a group of SPAC investors that's buying Offerpad, a 6-year-old real estate company that competes directly with Seattle-based company in a quest to transform how people buy and sell homes.

Read moreShow less
NASA Starts Recruiting the Next Wave of Astronauts

NASA says it'll take applications for its next class of astronauts between March 2 and 31 — the first step in what's expected to be a yearlong selection process.

To start the process, would-be spacefliers should click into USAJobs.com next month. For the first time in NASA history, applicants will be required to fill out a detailed online assessment that could take as long as two hours to complete.

Read moreShow less
Alan Boyle, GeekWire

GeekWire contributing editor Alan Boyle is an award-winning science writer and veteran space reporter. Formerly of NBCNews.com, he is the author of "The Case for Pluto: How a Little Planet Made a Big Difference." Follow him via CosmicLog.com, on Twitter @b0yle, and on Facebook and MeWe.

RELATEDTRENDING
LA TECH JOBS
interchangeLA