Can you sing like Freddie Mercury? Try the FreddieMeter
YouTube, Google and Queen teamed up on new AI program, which is helping raise awareness for HIV/AIDS charity
Given that Bohemian Rhapsody has more than a billion views on YouTube, there's a decent chance that, at some point in your life, you have sung along to at least part of the Queen megahit.
But exactly how well does your singing stand up against the legendary vocals of Queen frontman Freddie Mercury?
Thanks to the FreddieMeter, an AI-powered program built by Google researchers, you can now find out.
The FreddieMeter allows fans to pick their favourite Queen song from four options — Bohemian Rhapsody, Somebody To Love, We Are The Champions and Don't Stop Me Now — then belt out their best.
The program uses AI technology that measures people's pitch, timbre and tone, then produces a percentage number to show how close they came to Mercury's near-perfect pitch.
The program is a collaboration between Google, YouTube Music and Queen, who are using it to raise awareness for The Mercury Phoenix Trust, an HIV/AIDS charity founded in Mercury's memory by Queen members Brian May and Roger Taylor, as well as manager Jim Beach.
According to Toronto-based software engineer Nicholas Frosst, who also fronts the indie band Good Kid, the project was initially inspired earlier this year when people at Google saw that Bohemian Rhapsody had crossed the billion view mark — making it the first pre-Internet era track to reach that milestone.
The research team then approached the Mercury Phoenix Trust to find out if there was something they could do to mark the occasion, and bring more awareness to the charity. Now they hope people will measure their voices and post their results to social media using the hashtag #FreddieChallenge.
While the app may seem simple enough, Frosst says that measuring the unique traits of the human voice is no small feat.
"It was a good problem for Google Research to work on, because establishing pitch and timber and sound from a voice is actually quite difficult," said Frosst in an interview with q. "And you'll see it's a lot of fun."
So how does it manage to measure pitch and other vocal characteristics?
"It's doing that using a strategy for machine learning, which is called neural networks, and the real kind of breakthrough here was that it found a more efficient way of training a neural net to establish pitch from voice," explains Frosst.
"And it did that using a strategy called self-supervision, which is where you give a computer a whole bunch of examples of people singing and then it figures out the relative pitch between them."
In other words, instead of programmers teaching a computer to recognize perfect pitch, the computer figures it out itself.
Although Frosst is the lead vocalist for a band, he admits he hasn't yet made it past the 70 percent mark on the FreddieMeter — but he's not disappointed, given that Mercury had what Frosst calls one of the best voices of the 20th century.
"If you're benchmarking yourself against Freddie Mercury," he says with a laugh, "I think 70 is a pretty good number."
First formed in 1970, Queen became one of the bestselling and most revered rock bands in history. Their popularity spiked again in 2018 with the release of the film Bohemian Rhapsody, which has grossed nearly a billion dollars — making it the most popular biopic of all time — and landing actor Rami Malek a best actor Oscar.
You can find the FreddieMeter here; make sure to post your results in the comments.