The era of sound source separation has arrived in earnest!

【What is “Sound Source Separation by AI,” Used by the Beatles? Exploring the Potential of Technology to Support Musical Creativity】
https://realsound.jp/tech/2023/07/post-1367452.html

・Sound Source Separation technology allows artists and music creators to extract only the vocals from past recordings and convert them to clean sound quality to create new remastered recordings and remixes, as in the Beatles’ case.

・iZotope “RX”, which also includes an AI-based repair assistant function, can remove noise from old sound sources, or synthesize and restore frequencies from past sound sources that lacked bass or treble.

・RX’s sound source separation function, “Music Rebalance,” uses AI-based machine learning algorithms to analyze existing sound sources into four parts (vocal, bass, percussion, and other) and separate them as vocal or bass-only parts by adjusting the parameters of each. By adjusting the parameters of each part, it is possible to separate the vocal and bass parts only.

・In today’s music scene, the inclusion of sound source separation functions is becoming an industry standard in DJ software. In this field, “djay Pro AI” and “VirtualDJ” are well known for their AI-based sound source separation functions.

・LINE MUSIC has a feature that allows users to enjoy a karaoke-like experience by turning off only the vocals of the song and mixing the user’s singing voice into the sound source and playing it back.

・In the film industry, there are attempts to extract and separate individual sounds from old movies that have only “mixed” dialogue and sound effects, and rearrange them in a space using the Dolby Atmos method.

・Moises” is being used by professional and amateur music creators around the world. The application allows users to separate sound sources into up to five parts by simply uploading them to the server, where they are placed in a DAW-style browser. The user can then change the volume of each part and the BPM and key of the song automatically detected by the AI to whatever he or she wants, and then download it, making it quite easy to use. By turning off any part, the sound source can be used as an orchestra for practicing instruments or singing, and its use is not limited to remixing or mashup material.

These are the quotes from the article

The era of sound source separation arrives earlier than expected

【Separating Music by Parts; AI’s Sound Source Separation Technology is Amazing. Its advantages, problems, and considerations】
https://amimako.com/music-sound-separate-sleeter-copyright-ai/

【STEM PLAYER, a special music player. The way to enjoy music is also shifting from passive to active.】
https://amimako.com/music-stem-player-kanye-west/

We have discussed sound source separation several times on this blog and have paid attention to it. I had also predicted that the era of sound source separation would eventually come, but I did not expect that the era of full-fledged sound source separation would arrive so soon.

I used the above “Moises” trial, and was astonished at how easily and beautifully it separated existing music, and how amazing it was.

In my blog the other day, I talked about how “the age of combination, which will be ushered in by generative AI, will turn all of mankind into DJs”. (Not the combination of disc sound sources as in the past, but the combination of separated instruments and other sound sources in the future?)

Until now, it has been considered impossible to separate music that has been mixed. It is, by analogy, like separating mixed colors into the colors they were before they were mixed, and then returning them to their original multiple colors.

But as far as digital is concerned, I suppose it is possible to separate and restore what has been mixed. But to do so requires tremendous computational power.

【Google Develops Quantum Computer That Can Perform Computations That Would Take 47 Years Instantly】
https://nazology.net/archives/129181

Although quantum computers are not used in the above sound source separation technology, it is easy to imagine that the evolution of technology and computing power has been amazing.

The computational power of current computers (smartphones) is also increasing rapidly, which is probably why they are able to perform complex and advanced calculations and processing such as sound source separation.

However, according to the quantum computer article above, quantum computers do indeed have great computational power, but quantum noise can sometimes cause them to make inaccurate calculations.

This may sound strange,

Even if a quantum computer instantly derives a calculation that would take a supercomputer 47 years to do, it would still be 47 years before we could answer whether it is really correct or not, after having the supercomputer do the calculation. LOL!

When I read these, I thought,

The technological advances, such as generative AI and quantum computers, are amazing, and I don’t know what the right answer is and how to get around as a musician and as a person, which makes me think about many things, though,

In the first place, as with a quantum computer, there is no easy way to match the answer, so we have to accumulate what we think is the answer at this moment in time.(Even AI generation and sound source separation have various merits, demerits, pros and cons, so there is no correct answer at this point.)

Such was my thought once again.

We also often say that the answer is not outside of us, but within us.

I am almost swept away by various technologies and topics, but I eventually return to my starting point, as if I were trying to place emphasis on my inner self.

Well, but we are entering a great era in many ways!

See you then.

This is a screen shot of the music software I am currently working on. Usually, a musician’s job is to mix various instruments and sounds to complete a piece, but I wonder if this music will be separated by someone someday. I wonder if separating as well as mixing will become one of the jobs of musicians in the future.