Myuri Thiruna, a freelance photographer in Toronto, used to post frequently on Instagram and discuss photography with other users. But she said she had stopped two years ago, feeling “drained” by the demands of social media and the pursuit of followers and trends.
Then in July, Ms. Thiruna discovered Can of Soup, a new invitation-only social network where people make fantastical images of themselves with artificial intelligence and share the images with others. Enthralled by those abilities, she created A.I. images that showed her sitting on a unicorn floating in an ocean and her wearing a jacket made of Froot Loops.
Ms. Thiruna, 33, also commented on other users’ posts, chatting with them and making images together. She now spends as much five hours a day interacting with others on the app, she said.
“I met so many people on this app that I didn’t know before, and it goes beyond just posting and getting the likes,” she said. “It’s this meaningful connection with people and being also inspired by what they’re doing.”
Social networking apps are beginning to integrate A.I. into their image capabilities to make their platforms more social. After Facebook, Instagram and other apps have become more corporate over the years, A.I. imagery presents a way for them to bring back the whimsy and fun so users can rediscover what was once the point of the platforms: to share and interact with one another.
Large social platforms and new apps alike are incorporating A.I. image features. Last month, Snapchat announced Dreams, an A.I. imaging feature that lets users in Britain, Australia and New Zealand create outlandish selfies. TikTok last year rolled out several in-app filters that use A.I. to transform selfies into the style of a comic or a dreamlike character. BeFake, a social app launched in August, is also experimenting with A.I. selfies and images.
On Wednesday, Facebook, Instagram, WhatsApp and Messenger jumped in as well. Meta, which owns the apps, said the services would now offer A.I. tools for instantly generating photorealistic “stickers,” which can be shared. It added that it would introduce similar tools for editing and restyling existing images. These tools could put cowboy boots on two babies in a family photo, for instance.
“You can generate imagery inside of your chats,” said Ahmad Al-Dahle, Meta’s vice president of generative A.I. While most image-generation tools need 10 to 20 seconds to create an image, he added, Meta’s new tool needs only five.
The growing number of A.I. imagery tools in various apps underlines how “using A.I. interactively is where social media will go,” said Sam Saliba, who was Instagram’s global brand marketing lead and is now a marketing and branding consultant in Silicon Valley.
The trend takes A.I. images further than the apps that allowed people to produce A.I.-generated images without conversing or easily sharing them in an online community. Those apps included Lensa AI — which let people create A.I. selfies in styles like “cosmic,” “fairy princess” and “anime” — as well as Remini, Snow and Wombo. Interest in those apps peaked in mid-December, and downloads have since declined, according to the market intelligence firm Apptopia.
Ben-Zion Benkhin, the founder of Wombo, said many people didn’t stick with A.I. apps that were merely a “creation tool” and that gave users no ability to chat with one another about what they had produced.
“All of these apps are very limited,” he said. Adding social networking, he said, “does connect you to the other people.”
That understanding has helped drive new apps like BeFake, which has melded A.I. image features with socializing and sharing. BeFake prompts users at a different time every day to take a picture with their smartphone’s front and back cameras and then has A.I. transform the image.
Users need to share their posts before viewing other people’s posts. The concept was borrowed from BeReal, a photo-sharing app that has been popular among young users.
BeFake connects people through their creativity, said Kristen Garcia Dumont, one of the app’s founders. “What that means to each person is unique and intriguing, and you get to explore that with whoever you want in the app,” she said.
BeFake’s parent company has raised $3 million, and the app has tens of thousands of users, said Ms. Dumont and her co-founder, Tracy Lane.
Hayley Fligel, 17, a high school student in Burlingame, Calif., said she began using BeFake in July after a friend invited her to join. It’s different from apps like Snapchat, TikTok and Instagram, which are stressful because “if you want to take pictures or videos of yourself, you have to get ready, you have to get dressed, and you have to be doing something or have a nice background around you,” she said.
She said she could use A.I. on BeFake to make herself look like Taylor Swift or appear that she was playing volleyball, which shows “a more personal snippet of who you are.” While she seldom interacts with others on Instagram, she said, she comments on her friends’ posts on BeFake and browses a “Discovery” feed for inspiration from other posts.
Gabriel Birnbaum, who created Can of Soup with Eric Meier in May, said the point was to encourage creation and have fun. “It’s an app where you spend time with your friends,” he said.
Since then, he said, he has seen many creative and social moments happen in the app. In particular, a feature called Stir — which lets users put themselves in scenarios that someone else created — makes up one in four posts on the platform, Mr. Birnbaum said, with people inserting themselves into an A.I. image of Einstein inside a black hole in space, for example.
Mr. Birnbaum, who declined to disclose Can of Soup’s funding and number of users, said he didn’t plan to roll out the app widely until it had the “right trust and safety” with users comfortable with the content and whom they create photos with.
“I like the creation aspect and people liking my work and interacting with them,” said Alex Rosenblatt, 35, of San Francisco, who has used Can of Soup since June. “Most of my interactions on it are with people I don’t know, actually.”
Cade Metz contributed reporting.