The technology to break down a song into parts has been around for a long time, but I wonder if this means that it has become even more accurate.
This is a feature of Spleeter, an AI tool released last year by the music streaming service Deezer, which uses a process called "sound source separation" to separate any song's audio file into four stems for each type or group of instruments.
It's been possible to split a song into four stems based on the frequency range of each instrument or part, but I guess they've kept the basic principle and added an AI decision to it. The sample only shows the separation of the vocals, but from listening to it, it seems to be almost perfect at extracting just the vocals.
Changes" with split vocal and music tracks using the Spleeter library
It says, "The accuracy of the separation varies depending on the song and the instrument, and the separated bass and drum stems tend to be blurred or distorted," but I'd like to hear how well it can separate instruments other than vocals.
However, using this kind of technology, it's possible to extract a certain part of a song and use it for other purposes, which brings up another copyright issue.
Composition is generally defined as "creating a melody and chord progression. I've always wondered about that. I often compose music with guitar riffs as the main part, and there are many songs out there where the guitar riff is more important than the vocal melody. If you think of melody as a series of notes, then guitar riffs can be interpreted as "melody played on an instrument," so they fall under composition, but then you end up with everything that has a scale, and there's no end to it. It's difficult to say.
0 件のコメント:
コメントを投稿