A Blues-Singing AI Frog Is Taking Over TikTok Brazil

“O Sapo não lava o pé” (the frog doesn’t wash its feet) is the kind of song every Brazilian child learns before they can read. It’s about a frog who refuses to wash his feet. That’s the whole plot.

This week, an account called IABatida turned that nursery rhyme into a 1950s-style blues track, performed by a band of AI-generated frogs in a smoky lounge. The vocals are gravelly. The guitar is warm. There’s a stand-up bass.

It is, somehow, very good.

The video is labeled as AI-generated and has racked up more than 1.5 million likes on TikTok in a few days, with a parallel YouTube Shorts version pulling its own crowd. Brazilian commenters keep arriving in shifts to say the same thing in different ways: This should not work, but it does.

IABatida (“AI Beat” in Portuguese) has built an entire schtick out of running children’s songs and pop classics through AI as if they were recorded in another era. The account sits at 328,000 followers and 6.7 million total likes. Its 50s Motown-style cover of “Baby Shark” has 1.6 million likes on its own. An indie-rock version of the same song has another 388,900.

The trick is that the music is not bad. The arrangements have actual structure—choruses, bridges, restrained solos, and voices that actually sound pretty nice. The visuals match the era. The whole thing feels studied, not slopped together.

And that’s the point. The frog video belongs to a now-familiar species of internet artifact: the AI-native meme. Not a meme made with AI, but a meme that exists because AI exists, and would not be possible without it.

Decrypt has been tracking this lineage for a while. It probably started in late 2023 with the “Make It More” trend, where users asked ChatGPT to turn a normal image into something progressively more absurd until DALL-E gave up.

Then came “Ghibligeddon” in March 2025, when GPT-4o’s image generator drove a million people to sign up for ChatGPT in a single hour to turn their selfies into Studio Ghibli stills. Sam Altman publicly begged users to please chill while OpenAI’s GPUs melted.

Then there’s Italian brainrot—the genre that gave the world Tralalero Tralala, a three-legged shark in Nike sneakers; Bombardiro Crocodilo, a crocodile fused with a WWII bomber; and Ballerina Cappuccina, a ballerina with a cappuccino cup for a head. Tralalero Tralala first surfaced in January 2025 from a since-banned TikTok account. Ballerina Cappuccina followed in March 2025. By spring of that year, Italian brainrot characters were appearing in Ryanair ads and Loewe campaigns.

Other AI-native meme trends include the “pack” trend and the “dollification” trend started by Google after the release of Nano Banana.

What IABatida is doing fits the pattern but flips one variable. Italian brainrot mostly leans on the joke that AI generates incoherent nonsense and we love it anyway. The frog blues works, arguably, in the opposite direction. The joke is that the AI is competent. Suspiciously competent.

You watch the video expecting the punchline to be how janky it looks. Instead you sit there for forty seconds wondering whether you should pour yourself a drink.

Then the song gets stuck in your head.

That’s the new shape of the AI meme economy in 2026. Generators like Suno, Udio, and Google’s Lyria 3 can now produce three-minute songs with coherent structure from a one-line prompt. Image and video models can render a frog quartet in period-correct lighting without anyone having to model a single mesh.

The barrier to making something that looks and sounds like an actual production is now roughly the time it takes to type one paragraph.

IABatida’s back catalog already covers Aladdin’s “Arabian Nights,” the Brazilian children’s classic “Pintinho Amarelinho,” and multiple Baby Shark variants in styles ranging from 50s Motown to indie rock. The frog is just the latest entry. The next cover drops whenever the algorithm cools down.