How do you learn to play an instrument, or to make the question even simpler, how do you learn to play a new song you just discovered? Maybe chord charts if it’s popular enough, maybe dozens of web searches, but more likely you’re stuck trying to figure the song out on your own. Since software is continually improving to the point of being able to auto-detect chords from any song in your iTunes library, there are new options available. Billed as the main new feature, the Mac app Capo 3 automatically extracts chords, as well as tries to be a musician’s reverse-engineering toolkit by relying on computing power and plenty of code.
The flip side to using machines and offering a magical experience—if it works and a terrible one if it doesn’t—is human power to reverse engineer songs. Soundslice doesn’t currently use any algorithms to pick out chords like Capo, instead the company is building tools and interface elements to help users more easily annotate the music as well as share it. So which is the better way to dissect a song to learn its parts?
"Capo aims to be a tool that unapologetically goes out there and gives you a fighting chance of filling in the blanks," says Capo developer Chris Liscio. The program actually started off as a way for users to slow down the speed of songs without changing the pitch, a key feature to learning to play a song by ear. The second version added chord detection, but only visual cues, so the user still had tell the program what note was being played. In the latest version just released, Capo continues to pursue a brute force type method of trying to automatically detect and display the chords on top of the app’s unique spectrograph.
"It’s very easy to get lured into the idea of, you drag a song into it [Capo] and it tells you exactly how to play a song, it’s just not possible" says Liscio. He isn’t naive to the fact that once a cake is baked, once it’s been mixed and shipped, there just isn’t any program or solution that’s going to reverse the process perfectly. "There are advances in technology that are trying very hard to give users tools to try and isolate instruments in the spectrum, but at the end of the day there’s always going to be a user element to this process." The app’s popularity would indicate that users are at least happy to have a starting place when trying to figure out how to play a song.
"I'd want to add automatic music transcription features [to Soundslice] when they hit a very high degree of accuracy," says Soundslice cofounder Adrian Holovaty. "Until then, we're focusing on making it as easy as possible for a human to transcribe the music." Soundslice allows a user to search videos on YouTube and then easily add chords or notes for different stringed instruments. Transcribing songs from YouTube videos in the browser makes Soundslice a web app that makes the process of sharing a user’s work a natural extension of the transcription process.
There are plans to add auto measure detection to pro Soundslice accounts once they come become available, but the company’s still not chasing computer powered reverse engineering. "I've talked with more than a dozen PhD researchers around the world about this—people who have devoted years of their lives to the problem of automatic music transcription—and nobody is optimistic. It's basically an artificial-intelligence-level problem." says Holovaty.
So while programs like Capo or Riffstation, among many others, attempt to off put some of the hardest work to a machine, learning a new song will always be a human endeavor. Even if it was possible for a computer to take a song and completely dismantle it, Capo’s Liscio says he’s always had an internal struggle working in this type of software. "If the technology existed that I could easily extract everything about a song and lay it out in front of you, is that really right to be releasing out into the world?"