For nearly twenty years, Chris DeWolfe and Aber Whitcomb — the minds behind social network platform Myspace and gaming company Jam City — have made it their mission to be on top of the most popular tech trends. Earlier this yr, the veteran tech entrepreneurs shifted their focus to the web3 and generative AI craze with their latest enterprise– Plai Labs, an a16z-backed social platform development startup that goals to offer AI tools for consumers to collaborate and connect.

Plai Labs announced today the launch of its text-to-video generator, PlaiDay, joining the ocean of generative AI video tools, including Google’s Imagen, Meta’s Make-A-Video, OpenAI’s DALL-E 2 and Stable Diffusion. Nonetheless, one notable feature that stands out is the flexibility to personalize videos by adding your likeness via selfies.

Should you’ve ever desired to see a video of yourself (an animated version of yourself, that’s) flying over the Grand Canyon or battling intergalactic enemies, PlaiDay makes it possible. All you do is upload a selfie and sort a number of words, generating a short-form video that you would be able to share along with your friends and followers. The videos are currently only three seconds long, however the video duration will expand in the longer term. The corporate can also be working on adding an audio capability.

One example the corporate showed us used the prompt: “English Bobby, 1800s style, streets of London, close-up, life-like.”

While we wouldn’t call this “life-like,” PlaiDay does a good job of conveying realistic facial expressions. Plus, the people within the background are walking relatively easily. (Granted, they don’t completely move as real humans would, nevertheless it isn’t jarring.)

The personalized video is just a little wonky. For the reason that user’s selfie doesn’t show them with a mustache, the video appears as if the person’s facial hair is seemingly fading into their skin.

This other example (created by @Marvel on Discord) is ten times more distorted. And flat-out creepy, to be honest. To avoid this, the corporate recommends typing “close up” or “portrait of” to generate detailed faces.

There’s likely an extended approach to go before we’ll see cinematic AI-generated videos appear on the TV screen. Nonetheless, we’re still impressed by what PlaiDay has achieved here and might see the potential.

You may try the personalized text-to-video generator on the PlaiDay Discord channel. It’ll soon launch on PlaiDay, a free app currently in alpha. PlaiDay offers various other fun features, akin to inserting yourself into TikTok videos, a text-to-image tool, the flexibility to restyle images, a text generator for captions and more. The corporate says it would announce more details concerning the app in the approaching weeks.

“That is just the very starting of what we see as the longer term of storytelling,” co-founder and chief architect Jim Benedetto told TechCrunch in an interview. “Once you possibly can start putting yourself into these [AI] videos, you actually start to inform your personal stories.” Benedetto is the previous senior vice chairman of technology at Myspace. He also co-founded content optimization startup Gravity, which AOL bought in 2014.

The text-to-video generator is built on the corporate’s AI platform (the interior code name is Orchestra), which leverages a mixture of in-house and open-source models. That is the primary of many products that Plai Labs plans to supply.

As explained by the team, Orchestra is a “flexible set of tools that might be used to reinforce almost any business. It leverages a foundation of reusable code blocks, which is able to enable tech and non-tech-savvy folks to create and deploy latest, interesting AI applications quickly and efficiently – whatever the industry.”

Plai Labs hopes its platform will help designers and product managers create AI capabilities without the assistance of engineers.

“We’re excited since it enables a product manager with a vision to create AI that normally would take five or ten years,” added Benedetto.

Prompt:

Whitcomb explained that the AI platform isn’t limited to merely creating generative art. Soon, the platform will help with marketing campaigns, security monitoring, analytics and more.

“For instance, you possibly can generate a bunch of creatives with a diffusion process. You can then take a neural net that analyzes each one among those creatives and put some copy around that… you should use speech-to-text to make it ensure that it’s indexable and an LLM module that you would be able to use to give you the option to make use of multiple different languages or recommend other copy,” Whitcomb said.

“It’s flexible enough to do anything within the AI space,” he claimed.

The yet-to-be-named AI platform will change into widely available in the approaching months.