The wonders of artificial intelligence never cease to amaze, especially when coming from the virtual mouth of SpongeBob SquarePants.

A website by the name of 15.ai lets you pick from a range of popular fictional characters, type some text in and get an eerily accurate voicing of your input.

15.ai is run by a sole person, who describes the deep-learning text-to-speech tool as an example of how it's possible to get highly accurate voice replication from very small data sets.

Originally beginning as a part of the Massachusetts Institute of Technology's Undergraduate Research Opportunities Program, 15.ai has evolved into a project demonstrating that voices can be accurately cloned with just 15 seconds of data.

Freaky.

The whole thing even interprets tone and emotion using a technology known as "DeepMoji", a sentiment analysis model based on 1.2 billion tweets of data.

Among the characters currently included for your text-to-speech enjoyment are SpongeBob, Portal's GLaDOS, the cast of Team Fortress 2, Rise from Persona 4, and — of course — the My Little Ponies.

On a completely unrelated note, we overheard this exclusive conversation between some of our faves talking about Byteside — check it out.

audio-thumbnail
Listen
0:00
/0:35

AI is only getting more advanced in its capabilities, powering predictions for protein structures, making Photoshop easier, and occasionally mistaking a soccer linesman's bald head as a ball.

Impressive tech indeed.

Share this post