Why Does My Voice Sound Like a Robot on Phone Calls?

When I set out to investigate, I didn’t turn up evidence of an encroaching cyborg army. Instead, I learned that the robot-voice phenomenon is the result of benevolent technologies trying to prevent interrupted phone calls.

“When information is going over a digital network, it has to be converted to bits,” says Alex Hills, a professor of engineering and public policy at Carnegie Mellon University. In other words, data is translated into ones and zeros that can be understood by a computer. That’s true no matter the content: texts, emails, your voice on a phone call, you name it.

Your phone transmits those bits through the air as radio waves. “At the other end, the receiver converts the ones and zeros back to the original information,” Hills says. “That’s how digital works.”

But things can go wrong. Problems can crop up when the voice data is being converted from one format to another, but often it’s just interference during wireless transmission, like competing radio traffic or physical objects standing in the way.

Sometimes the call drops. Other times the signal is strong enough to maintain a connection, but part of the information the radio waves are transmitting gets corrupted—ones get flipped to zeros, or strings of bits are lost altogether. It’s the digital equivalent of the snow or noise you’d hear on an old AM radio.

The solution starts with what’s called error detection. “The receiver knows it has this stream of bits, and it has an error detection system that can tell there are little bursts of bits coming through that aren’t right,” Hills says.

It’s a scenario he knows well from the time he spent in the early 1990s building the world’s first big WiFi network. WiFi also transmits data using radio frequencies. He described the effort in a book called “WiFi and the Bad Boys of Radio.” “We used a lot of that technology,” Hills says.

The errors can be relatively simple to deal with when the information is something like a text, an email, or the words and photos on a webpage. The system just asks the sender to transmit another copy of the data. But the fix is a lot more complicated with real-time, continuous streams of information like phone calls.

Your cell phone and other telecommunication systems also use error detection, “but with voice calls you don’t have the luxury of retransmission,” says Sanjay Udani, a technologist at Verizon who was one of the lead architects of the company’s FIOS network. With a call, “you can’t resend the packets because there are just milliseconds of delay,” and there’s no backup copy of your voice. 

Your phone could just play silence for the parts of the call where the information isn’t coming through correctly, but that would be hard to understand. So instead, the system tries to correct the error, guessing at what those missing bits might be and generating sound to fill the gaps in the middle of your words. “Sometimes it does a great job, but sometimes it can sound robotic or distorted,” Udani says.

Essentially, the distorted voice you’re hearing is the result of a bunch of little computers involved with your phone call, trying their best to chime in and help you get your message across.

The robot voice problem can happen to anyone, from time to time. But if it comes up frequently, engineers have a few suggestions.