Japan’s Nightmarish Robot Mouth Can Sing


Warning: The following footage of a robotic mouth imitating human speech can not be unseen.

Professor Hideyuki Sawada of Japan’s Kagawa University is a man determined to advance the field of robotics at all costs. In this case, those costs include horrible emotional trauma for anyone exposed to his latest project.

Hoping to mimic biology, Professor Sawada and his colleagues have cobbled together a surprisingly lifelike robotic analogue to the human vocal system. From the ‘bot’s autonomously constricting nasal resonance chambers to its eight faux vocal cords, everything that allows you to scream “kill it with fire” is included in the machine’s artificial frame.

The thing even has lips.

Despite the robot’s nightmarish appearance, I have to admit a grudging respect for the software that powers the machine. Professor Sawada’s team coded a program that enables the ‘bot to listen to its own “voice” via microphone, then adjust its pitch and tone on the fly in an effort to more closely replicate actual human speech. In short: It learns from its mistakes, and is constantly trying to sound more like you.

I should point out that the disclaimer at the beginning of this piece was not referring to the demo video you see above. It was, in fact, referring to the following footage of this thing “singing.”

This disembodied bionic throat undoubtedly has some fascinating implications for the future of robotics. If we’re being wildly optimistic, I could even posit that the ‘bot holds a wealth of potential as a prosthetic for people with very specific injuries.

If nothing else, it would make automated calling centers a bit less sterile.

That said, there’s a large part of my mammal brain that sees this thing as anathema to the natural world. We’re beyond the uncanny valley here, people.

Before you take to the comments to decry this thing and/or my predilection for alarmist anti-robotic sentiment, please remember: Your “Daisy Bell” joke is neither original, nor clever. Try harder.

Source: ieee Spectrum

About the author