Gaussian Splat of a Strawberry

A hyper-detailed 3D strawberry drops, and the internet immediately roasts its bruised backside

TLDR: A creator made an ultra-detailed 3D strawberry by photographing it from tons of angles and released it for others to use. The comments instantly split between amazement at the effort and ruthless jokes about the fruit looking damaged, turning a tech demo into a full-on strawberry roast.

A beautifully obsessive strawberry scan just hit the internet, and instead of politely applauding the craft, the crowd did what the crowd does best: zoom in, nitpick, meme, and absolutely lose it. The creator photographed one strawberry from 90 angles, with 88 layered shots each, to build a remarkably lifelike 3D model using slang-splat. Translation for normal people: this was a wild amount of work for one piece of fruit. And yes, people noticed. One stunned commenter summed up the production stats with a single verdict: "Insane!"

But admiration only lasted about five seconds before the comments turned into a fruit autopsy. The biggest mini-scandal? The strawberry’s underside. Multiple people zeroed in on it like scandal reporters at a red carpet disaster, with one asking, "What happened to the bottom of that poor strawberry?" Another went straight for the kill with, "You might want to throw that one away :)" Ouch. Suddenly this wasn’t a digital art post anymore — it was a public intervention for damaged produce.

And of course, the thread had its delightful chaos. One commenter spiraled into a mini comedy bit about finally understanding what “3DGS” means, landing on the wonderfully silly image of fuzzy blobs splatted on a screen. Another demanded the next escalation: combine this with microscopy, basically begging the internet to make fruit scanning even more absurdly detailed. So yes, the strawberry is impressive — but the real spectacle is the comment section, where awe, confusion, and produce-shaming collided.

Key Points

  • The strawberry reconstruction was created from images captured from 90 perspectives, with 88 focus-stacked images for each perspective.
  • The imaging setup included a Nikon Z8 full-frame camera, f/7.1, 1/160 exposure, ISO 100, and a Laowa 180mm macro lens.
  • LED lighting and a blue screen were used during capture.
  • Training for the Gaussian splat was done using the slang-splat tool linked on GitHub.
  • The work is downloadable under a CC BY license, and the related COLMAP dataset is available for free via the creator’s Patreon.

Hottest takes

"throw that one away :)" — carlos-menezes
"What happened to the bottom of that poor strawberry?" — mgaunard
"Insane!" — galsapir
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.