Meta’s latest generative AI trick is to remake your videos in all kinds of fun and outright goofy ways. The Meta AI assistant will revamp up to ten seconds of video in dozens of ways on its mobile app, on the Meta.AI website, and inside the new Edits app from Instagram.
The preset video transformations are free in the dozen or so countries where they are available.
I wanted to try it out, but I didn’t want to wait too long for the edits, so I opted for a simple old video: a Boomerang (remember when that was the trendy video tool?) of myself enjoying a fancy cocktail in a bar. As you can see, it’s a pretty basic setup.
The edits are powered by Meta’s Movie Gen model. And while it’s limited to presets for now, Meta has promised custom text-prompt editing later this year.
Somewhat at random, I picked a few amusing presets, which are sorted into sections for fashion, locations, lighting, styles, and “weird.”
First up, the video game filter. The AI decided I belonged in some kind of neon-lit boss level from 1997.
The lighting changed to a garish blue-purple glow, my shirt morphed into something that looked vaguely armored, and the bar looked a lot like a techno party from an early PlayStation game:
Then came superhero mode. My shirt turned into a tight, digital-looking supersuit, and a cape flapped dramatically in the nonexistent wind. I looked like Thor’s third cousin from a direct-to-video Marvel film.
There was nothing particularly heroic about how I sipped the drink, but it did spruce up some of the folks in the background:
The background characters and the bar itself joined me in transforming for the Classic filter. The purple embroidered coat is a nice touch, but I don’t know that I would drink red wine when wearing such a fancy white shirt.
Porcelain mode was intriguing and took me straight to the uncanny valley. I became a doll. A glossy, glassy-skinned, dead-eyed version of myself, with weirdly rouged cheeks and a sculpted outfit.
I looked like the haunted Victorian toy kids are afraid of at night. Porcelain doesn’t drink, which might be why I don’t seem to actually have a glass in my cold, white hand.
Maybe they don’t allow drinks in the museum where all those people disappeared years ago, that this doll clearly belongs in:
For some reason, the Teddy option didn’t come off as creepy despite also transforming me into a toy. Specifically, it rendered me as an anthropomorphic bear with big button eyes and fuzzy arms gripping a possibly empty glass.
It was like watching a product of Build-A-Bear after a long week. This time, though, the transformation of the woman in the background was less of a thematic addition and more of a horrifying mish-mash of plush toys.
You can try out Meta’s AI video editing yourself now (provided it is supported in your region – the UK does not appear to be supported). It’s very simple to do. You upload the clip, tap a preset, and let Meta’s AI do its magic.
It took less than a minute per transformation. Once done, I could share the videos on social media platforms owned by Meta, including Meta AI’s Discover feed.
You might also like
https://cdn.mos.cms.futurecdn.net/6aV7phCn4MeFYjXDJM89Xc.png
Source link
erichs211@gmail.com (Eric Hal Schwartz)
