March 24, 2019 at 08:45What is generative music?
This is a podcast with those who create content. The guest of the issue is Alexei Kochetkov, CEO Mubert , with a story about generative music and his vision of future audio content.
Alexey Kochetkov, CEO Mubert
alinatestova : Since we are not only talking about textual and spoken content, naturally, we did not ignore the music. In particular, a fairly new direction in this area. Alexey, you are the CEO of the Mubert project . This is a streaming service that creates generative music. Tell me how it works?
Alexey: Generative music is created in real time by algorithms. This is music that can be adapted, applied in any field, personalized, and so on. It is collected in real time from a certain number of samples.
A sample is a piece of music that every musician has the ability to record. That is, generative music is created from, as they say in English, human-made samples [samples created by man]. The algorithm analyzes them and creates a stream just for you.
Alina: Super. Music is created by an algorithm, an algorithm by humans.
It makes sense to talk a little about the background of this project, about its beginning. Why did you decide to do it? Was it connected with your musical interests?
Alexei: As they say, startups are born out of pain. I was running, and it constantly hurt on my side from switching music. At that moment, the idea came to my mind: why not create an application in which the samples will line up in an endless composition corresponding to the pace of my run. So the first idea of Mubert was born.
The team was assembled on the same day and began to create a product, which subsequently, of course, was made by several pivots. But the concept itself is the same that was created on the first day.
Alina: Did your musical past somehow influence your choice or certain steps that were taken during the development of the application?
Alexey: No. I have a musical jazz education, and here it helps a little. I know the notes, I know how to play the double bass and what the music consists of.
I was always responsible for the bass. In all the bands I attended, I always took the lowest frequencies and played the double bass, bass, and bass synthesizers. This does not help in working on Mubert. I just know about how the music works, I listen to it a lot, and I have a long time ago the belief that there is no bad music or bad taste.
There is a personal taste and personal approach to music. Each person has his own, and each person has the right to choose music and so show his taste.
A little knowledge of notes, harmonies and other things helps me. But in general, about fifty other musicians who are actively involved in the development of the interface, music ranking systems and artificial intelligence systems work on Mubert, in addition to me. These are people who constantly give advice and influence the way Mubert sounds today.
Alina: Is it possible to say that essentially generative is the type of music that harmoniously combines with other activities?
For example, usually writing a text or working to music is a pleasure for an amateur. Someone may get used to it, while someone else may not. Can algorithmized music give a synergistic effect, which, on the contrary, will allow you to enter a stream state?
Alexei: This is a hypothesis, and we are trying to test it.
They will soon read to the generative music - we are making a joint application with Bookmate. People run marathons to generative music, and this is the only application that gives you the opportunity to run and not change your pace for four, eight, sixteen hours and so on. They work and study to this music. It's probably a cool approach to music - to be a sponsor of your hobby. But this is a hypothesis.
Alina: And you check her collaborations?
Alexei: She is confirmed by the subscriptions and auditions that take place at Mubert every day. For example, meditation is the channel we buy the most.
Alina: And High, for what conditions, if not literally take it? (laughs)
Alexey: Resting, relaxing, feeling some kind of connection and so on.
Alina: Super. Please tell me, in your opinion, is generative music - algorithmic, repetitive, long-lasting - is it something fundamentally new or some continuation of ethnic, shamanistic and meditative music?
Alexei: This is a bit of a repeat.
Mubert started in 2000 when I actually dubbed Bomfunk MC’s from Monte Carlo [track] radio . As soon as he sounded on the radio, I continued to record on tape, until I recorded the whole side with this track. Then I did the same with the other side. As a result, I had a whole tape on which only Bomfunk MC's - Freestyler was recorded.
Mubert goes back to these times. A bunch of people use music on repeat. They include some track and work under it all day or play sports for some time.
Generative music in its current state does not fully have all the drama that a DJ can give. He real-time understands that now you need to raise BPM, now omit, expand harmony or narrow down. Generative music only strives for this.
And we are pioneers in creating drama in generative music, which we have learned to create endlessly long, smooth and understandable. Now we are learning to create drama in it.
As we recently showed at the adidas store. We created a DJ set without a DJ, and a lot of people danced beautifully to this music. It sounded at the level of German DJs, who, in principle, were the authors of the samples. But it was the set that Mubert created.
Alina: How does the algorithm work?
Alexey: The algorithm analyzes a lot of parameters: melody, rhythm, saturation, "fat" sound, instrument. His pace, tonality and so on. A bunch of parameters that are objective. Next, subjective parameters begin. This is a genre, activity, your taste. There may be location related options. When you want to collect, for example, a city stream, to understand how the city of Berlin sounds.
The AI system here is an accompaniment to ensuring that subjective parameters are executed. In order for you to receive music at the time of some activity that is based on your taste and on the things that you have already managed to show on this system.
Soon we will release an application in which you can like, dislike, “favorite” music and influence your own style. This will be the first application in the world in which there is no common chart. We do not even have in the database such a thing as a general chart of popularity or unpopularity of samples and artists. Each has its own chart, which contains combinations of parameters. Based on them, the system learns and creates your own soundtrack.
Alina: In fact, we are talking about the fact that for each user of Mubert there are many soundtracks to various aspects of his life.
Alexey: Yes. This is the first real personal streaming.
Alina: Super. You have already started talking about collaboration with adidas, but please tell us about collaboration with brands in general. How do they look?
Alexei: Music is the closest form of creativity to a person. Accordingly, if a brand wants to get closer to a person, it must be done through music. Not many people know about it yet, but those brands that know are already starting to do it.
For example, adidas holds pop-up parties that occur suddenly in some of their stores. They are not advertised. Other brands - sponsor theme parties.
Who should they move to, if not towards new technologies? They have two options: they either take the top DJ, or the top technology. If it is possible to combine this - as we did with adidas, when our samples were provided by one of the top producers of Berlin Atom TM- the person who created the electronics. Then the brightest spark is born, which sparkles so that the brand can express itself.
Alina: If we are talking about parties ... Naturally, there are a lot of people there. How does Mubert understand which music to make? How does personalization work in this case?
Aleksey: The party is customized to the party, the city to the city. That's all ...
Alina: Essence.
Alexei: Yes, an entity under which we can tune. Personalization goes from your time of day and day to some global things. As I explained: there are objective parameters, there are subjective ones. The set of subjective parameters is a genre, a city, you, morning. Anything. Objective - the saturation of the sound, its tempo, tone, gamma and so on. All those things that can be measured objectively.
Alina: What do you think, how will generative music and music as a whole develop? Will the algorithm replace the person-composer or DJ in the future?
Alexei: No way. DJ selector will remain. Steeper than a DJ, you can’t collect music - be it track or sample music. Previously, DJs were called selectors, and this work will remain because they collect the most “fat”.
The development of generative music will lead to the fact that it will appear on every phone, because it gives slightly different opportunities for adapting and personalizing this music. It will also feature collections of authors. For example, we will be able to exchange some generations and understand how you trained your Mubert, and how - I mine. It’s just like today with playlists, only at a deeper level.
Alina:It turns out that the future of generative music is a symbiosis of a human creator and an algorithm that analyzes everything that happens more deeply and accurately?
Alexey: Absolutely.
Alina: Super. And finally - our blitz of two questions. Music helps ...
Alexey: To live, to breathe.
Alina: The best track is the one ...
Alexei: Which “inserts”.
Alina: Class, thank you very much.
- listen in Telegram or in a web player
- subscribe to podcast in iTunes or on Habr
Alexey Kochetkov, CEO Mubertalinatestova : Since we are not only talking about textual and spoken content, naturally, we did not ignore the music. In particular, a fairly new direction in this area. Alexey, you are the CEO of the Mubert project . This is a streaming service that creates generative music. Tell me how it works?
Alexey: Generative music is created in real time by algorithms. This is music that can be adapted, applied in any field, personalized, and so on. It is collected in real time from a certain number of samples.
A sample is a piece of music that every musician has the ability to record. That is, generative music is created from, as they say in English, human-made samples [samples created by man]. The algorithm analyzes them and creates a stream just for you.
Alina: Super. Music is created by an algorithm, an algorithm by humans.
It makes sense to talk a little about the background of this project, about its beginning. Why did you decide to do it? Was it connected with your musical interests?
Alexei: As they say, startups are born out of pain. I was running, and it constantly hurt on my side from switching music. At that moment, the idea came to my mind: why not create an application in which the samples will line up in an endless composition corresponding to the pace of my run. So the first idea of Mubert was born.
The team was assembled on the same day and began to create a product, which subsequently, of course, was made by several pivots. But the concept itself is the same that was created on the first day.
This is music that does not have a beginning, an end, pauses, or interdesy transitions.
Alina: Did your musical past somehow influence your choice or certain steps that were taken during the development of the application?
Alexey: No. I have a musical jazz education, and here it helps a little. I know the notes, I know how to play the double bass and what the music consists of.
I was always responsible for the bass. In all the bands I attended, I always took the lowest frequencies and played the double bass, bass, and bass synthesizers. This does not help in working on Mubert. I just know about how the music works, I listen to it a lot, and I have a long time ago the belief that there is no bad music or bad taste.
There is a personal taste and personal approach to music. Each person has his own, and each person has the right to choose music and so show his taste.
A little knowledge of notes, harmonies and other things helps me. But in general, about fifty other musicians who are actively involved in the development of the interface, music ranking systems and artificial intelligence systems work on Mubert, in addition to me. These are people who constantly give advice and influence the way Mubert sounds today.
Alina: Is it possible to say that essentially generative is the type of music that harmoniously combines with other activities?
For example, usually writing a text or working to music is a pleasure for an amateur. Someone may get used to it, while someone else may not. Can algorithmized music give a synergistic effect, which, on the contrary, will allow you to enter a stream state?
Alexei: This is a hypothesis, and we are trying to test it.
They will soon read to the generative music - we are making a joint application with Bookmate. People run marathons to generative music, and this is the only application that gives you the opportunity to run and not change your pace for four, eight, sixteen hours and so on. They work and study to this music. It's probably a cool approach to music - to be a sponsor of your hobby. But this is a hypothesis.
Alina: And you check her collaborations?
Alexei: She is confirmed by the subscriptions and auditions that take place at Mubert every day. For example, meditation is the channel we buy the most.
There are three paid channels in total: Meditate, Sleep and High. High is dub reggae. The best-bought one is meditation, because in the process of meditation the music should not stop or change. Mubert does it.
Alina: And High, for what conditions, if not literally take it? (laughs)
Alexey: Resting, relaxing, feeling some kind of connection and so on.
Alina: Super. Please tell me, in your opinion, is generative music - algorithmic, repetitive, long-lasting - is it something fundamentally new or some continuation of ethnic, shamanistic and meditative music?
Alexei: This is a bit of a repeat.
Mubert started in 2000 when I actually dubbed Bomfunk MC’s from Monte Carlo [track] radio . As soon as he sounded on the radio, I continued to record on tape, until I recorded the whole side with this track. Then I did the same with the other side. As a result, I had a whole tape on which only Bomfunk MC's - Freestyler was recorded.
Mubert goes back to these times. A bunch of people use music on repeat. They include some track and work under it all day or play sports for some time.
Generative music in its current state does not fully have all the drama that a DJ can give. He real-time understands that now you need to raise BPM, now omit, expand harmony or narrow down. Generative music only strives for this.
And we are pioneers in creating drama in generative music, which we have learned to create endlessly long, smooth and understandable. Now we are learning to create drama in it.
As we recently showed at the adidas store. We created a DJ set without a DJ, and a lot of people danced beautifully to this music. It sounded at the level of German DJs, who, in principle, were the authors of the samples. But it was the set that Mubert created.
Answering a question, generative music takes its origins from repit and ends up with what we still cannot imagine.
Alina: How does the algorithm work?
Alexey: The algorithm analyzes a lot of parameters: melody, rhythm, saturation, "fat" sound, instrument. His pace, tonality and so on. A bunch of parameters that are objective. Next, subjective parameters begin. This is a genre, activity, your taste. There may be location related options. When you want to collect, for example, a city stream, to understand how the city of Berlin sounds.
The AI system here is an accompaniment to ensuring that subjective parameters are executed. In order for you to receive music at the time of some activity that is based on your taste and on the things that you have already managed to show on this system.
Soon we will release an application in which you can like, dislike, “favorite” music and influence your own style. This will be the first application in the world in which there is no common chart. We do not even have in the database such a thing as a general chart of popularity or unpopularity of samples and artists. Each has its own chart, which contains combinations of parameters. Based on them, the system learns and creates your own soundtrack.
Alina: In fact, we are talking about the fact that for each user of Mubert there are many soundtracks to various aspects of his life.
Alexey: Yes. This is the first real personal streaming.
Alina: Super. You have already started talking about collaboration with adidas, but please tell us about collaboration with brands in general. How do they look?
Alexei: Music is the closest form of creativity to a person. Accordingly, if a brand wants to get closer to a person, it must be done through music. Not many people know about it yet, but those brands that know are already starting to do it.
For example, adidas holds pop-up parties that occur suddenly in some of their stores. They are not advertised. Other brands - sponsor theme parties.
Who should they move to, if not towards new technologies? They have two options: they either take the top DJ, or the top technology. If it is possible to combine this - as we did with adidas, when our samples were provided by one of the top producers of Berlin Atom TM- the person who created the electronics. Then the brightest spark is born, which sparkles so that the brand can express itself.
For any brand, music is an information guide.
Alina: If we are talking about parties ... Naturally, there are a lot of people there. How does Mubert understand which music to make? How does personalization work in this case?
Aleksey: The party is customized to the party, the city to the city. That's all ...
Alina: Essence.
Alexei: Yes, an entity under which we can tune. Personalization goes from your time of day and day to some global things. As I explained: there are objective parameters, there are subjective ones. The set of subjective parameters is a genre, a city, you, morning. Anything. Objective - the saturation of the sound, its tempo, tone, gamma and so on. All those things that can be measured objectively.
Alina: What do you think, how will generative music and music as a whole develop? Will the algorithm replace the person-composer or DJ in the future?
Alexei: No way. DJ selector will remain. Steeper than a DJ, you can’t collect music - be it track or sample music. Previously, DJs were called selectors, and this work will remain because they collect the most “fat”.
The development of generative music will lead to the fact that it will appear on every phone, because it gives slightly different opportunities for adapting and personalizing this music. It will also feature collections of authors. For example, we will be able to exchange some generations and understand how you trained your Mubert, and how - I mine. It’s just like today with playlists, only at a deeper level.
Alina:It turns out that the future of generative music is a symbiosis of a human creator and an algorithm that analyzes everything that happens more deeply and accurately?
Alexey: Absolutely.
Alina: Super. And finally - our blitz of two questions. Music helps ...
Alexey: To live, to breathe.
Alina: The best track is the one ...
Alexei: Which “inserts”.
Alina: Class, thank you very much.
Our microformat on the topic of content marketing:What kind of office do you have?
Archetypes are enough : why do
PS In the glphmedia profile - links to all issues of our podcast.