Can you tell the difference between the composer Ludwig van Beethoven and a computer? Humor us for a moment – this is not a trick question. As artificial intelligence (AI) technology has found its way into every aspect of our lives, music is one of the areas where artists are making use of it.
The question is, how good is AI-generated music? Can it rival symphonies created by a genius like Beethoven? And would you recognize an AI-composed piece of music compared to one created by a human being? These are the questions we will try to answer in this article.
With the recent advancements in AI technology, it has become increasingly difficult to distinguish between music created by AI and music created by humans. AI-generated music has become more sophisticated, and it can now replicate different styles and genres of music with remarkable accuracy. However, many music experts argue that AI-generated music lacks the emotional depth and complexity that human-created music has.
People can often tell the difference between music created by AI and people, as AI-generated music may lack the subtle nuances and variations that come from a human musician’s interpretation and creativity. Additionally, while AI-generated music may be technically proficient, it may not have the same emotional impact on listeners as music created by humans, which often reflects the human experience and emotions.
For decades, artificial intelligence was the stuff of science fiction, something that movies were made of. Recently, though, technology has advanced to a point where extremely powerful computers are affordable even for smaller software developers. This sea change has put the development of AI-based apps and software within reach of many more people.
As a result, AI is becoming an integral part of our lives, and the art world is no exception. But how good are AI-based music generators really? Can they produce works that are indistinguishable from the productions of some of the world’s most renowned composers or most beloved bands? Before looking at the answers to those questions, it is worth taking a closer look at some of the leading AI-based music generation tools.
AI music generators are essentially pieces of software that produce music. To achieve that, they mimic human behaviors, such as learning and other cognitive functions. Most of today’s leading AI music generation tools are based on one of two principles: neural networks or machine learning.
Machine learning (ML) is perhaps the best-known subset of artificial intelligence technology. ML uses algorithms that continue to learn without additional human input. Once trained on an initial dataset, the algorithm then develops its own predictions in answer to a question.
In the case of music generation software, the algorithm would create a song or the overture to an opera, for example. If that piece of music is considered to be successful, the algorithm learns that its predictions of what makes a good song were right. As a result, it is likely to apply similar principles for its next task.
Neural network-based music generation tools work somewhat differently. Rather than training algorithms that learn from their own experience, neural network programming tries to mimic the way the human brain works.
When it comes to composing, (musical) humans have a sense of which sequence of notes results in a pleasing song or other work of music. To allow an AI to replicate this approach, scientists would recurring neural networks to train sequential models of AI.
You could be forgiven if this is starting to sound too complicated. For years, aside from a lack of computing power, the need for computer science expert training kept the field of AI small. As tools have developed, AI has become more accessible and democratized. For that reason, users no longer need to be an expert in machine learning to use AI-based music generators.
There are numerous AI-based music generators out there, aimed at professional music producers, amateur music enthusiasts, and AI fanatics. This selection is far from exhaustive, but we have aimed to cover tools with a range of different approaches and functionalities.
Our top five AI-based music generators include:
Amper Music
MuseNet
Magenta Studio
Brain.fm
Algonaut Atlas 2
1. Amper Music
Amper Music has become known as one of the most accessible AI music generation tools. It makes it possible for anyone to start using AI-generated music for videos or other projects. The cloud-based platform joined forces with image distribution company Shutterstock.
From its new home base, Amper allows users to create their own tracks with as little or as much input as they like. Users can choose to let the software compose the entire piece, or they can specify any number of criteria to put their slant on the piece of music.
Amper allows budding musicians to define keys, choose their favorite instruments, and determine the pace of the piece. All that is possible without an in-depth understanding of composition or music theory. Users simply need to be able to define what they like and dislike as Amper takes care of the hard lifting.
2. MuseNet
Right now, it is almost impossible to talk about artificial intelligence without mentioning OpenAI. The creators of the now-infamous ChatGPT software have also been working on a music generator. MuseNet is the result of their efforts.
MuseNet’s compositions are based on a deep neural network that was trained with data from online sources. According to its creators, it is able to integrate ten different instruments into its songs and can produce music in more than a dozen different styles.
Plus, MuseNet can imitate both contemporary favorites and some of the all-time greats like Mozart. However, despite its wide-ranging capabilities, this music generator does not yet allow users to compose pieces from scratch.
3. Magenta Studio
Magenta Studio is available to users courtesy of Google’s Magenta research team. The team constructed this music creator on top of its Magenta open-source library that offers access to different machine learning models and tools for art and music generation.
Although the tool is aimed at seasoned composers, it is user-friendly and therefore perfectly accessible for those new to AI music creation. Composers will find a range of accessible interfaces and instruments, including a step sequencer, a drum machine, or even a piano roll editor.
Google’s engineers also added previously trained models to Magenta studio to help users generate chord progressions, fresh melodies, and percussion patterns. The software can deliver any of those, depending on user input. Lastly, it is highly accessible, requiring no downloads and running from a simple browser window.
4. Brain.fm
While the previous music generators allowed users to create all types of music, brain.fm has more of a focus. This platform intends to help its users to become more focused, find it easier to relax, and improve their sleeping patterns.
Rather than instructing the software with a certain musical genre, users define the mood they would like to achieve with the help of the composition. Once users have defined the cognitive state they are looking for, brain.fm uses an algorithm, principles of neural network technology, as well as psychoacoustics to deliver the right piece.
Users can then alter settings such as speed or instrumentation to personalize their session further. The developers of brain.fm claim that their AI music generator has been proven to aid meditation, promote more restful sleep, and improve the attention span of listeners.
5. Algonaut Atlas 2
Algonaut specializes in developing AI-based music production tools, and Atlas 2 is no exception. Using machine learning algorithms, this software allows users to develop melodies, harmonies, and rhythms that are compatible with their existing music library.
One of Atlas 2’s key functions is the ability to understand what is in a user’s music library and match it. Users can import their own samples as a basis for Atlas 2 to work from, but the software also has a library of samples, loops, and sounds ready to use.
Although aimed at professional producers, the software is easy to use for anyone looking to simply have a go at AI music generation. Professional users will benefit not only from new inspiration but also from the ability to export MIDI files which can then be transformed further in any DAW software.
Can we tell the difference between AI vs human generated music.
It is too early to deliver a set-in-stone verdict on whether humans can detect music that has been generated by AI. Early studies seem to suggest that telling the difference is harder than most would think.
The developers behind the Amper music generator conducted their own research, supported by audio researchers Veritonic. They asked a panel of participants to distinguish between stock music, AI-generated music, and music composed by human intelligence. The overall finding was that the average person was unable to tell the difference. As if that result was not surprising enough, 25% of the panelists said they would value a piece of music in an advert more if they knew that it had been composed by AI.
Granted, perhaps a study conducted by the makers of an AI music generator has some bias. Therefore, it is worth looking at other tests. Customer service platform Tidio conducted a wide-ranging survey test comparing AI-generated art to human creations in 2022. Their research covered a wide range of AI-generated content, including AI-generation faces, music, and AI-written content.
Survey respondents stated that they found music to be the category where it was hardest to distinguish between human and machine. One of the reasons for the difficulties may lie in the panelists’ inability to separate composition and performance. A human-generated electronic dance music track, for example, was considered to be AI-produced by seven out of ten participants.
On the other hand, a song composed by an AI trained on Beatles songs was perceived by anonymous respondents as human-created. Perhaps the most surprising result of this research was an underlying sense of human inferiority. Participants attributed music they believed to be ‘too good’ or ‘too complex’ to AI, whereas more ‘chaotic’ pieces were ascribed to human composers.
One last example to illustrate how close AI-based music generators are to human creations. Sony Computer Science Laboratories created an AI to try and create music like that composed by Baroque composer Johann Sebastian Bach. Bach’s music is known for its complexity. The scientists developed a neural network called DeepBach and trained it on the chorales of the composers.
The creations by DeepBach had 50% of listeners convinced they were hearing the real thing, whereas 75% identified the real composer correctly.
Artificial intelligence is establishing its place in the art world, including image generation through the likes of Stable Diffusion and music. Where exactly that place will be is not yet known. Amper’s test mentioned above compared AI-generated music to stock music, which the majority of people would use as background music in everyday life rather than something they listen to attentively.
Perhaps this is where AI-generated music is headed, or perhaps machine-made scores will become the mainstay in fields like advertising. In those cases, music is one factor among many, but it is not the main act, so to speak. Saying that, there is no reason why the role of AI-generated music should not expand with advances in related technologies.
Autonomous vehicles, the future of healthcare – no matter where you look, AI is making an impact on our quality of life. Whether it is by taking care of repetitive tasks or creating sought-after online content, greater segments of society will likely become more dependent on these technologies.
At the same time, AI will no longer be the exclusive remit of someone who is an expert in systems design, but it will become more accessible to everyone. As a result, its role in real life will expand further. What that means for music remains to be seen.
TwoSetViolin. “Can You Tell the Difference between AI and Human Composers?” YouTube, Video, 23 Sept. 2020, https://www.youtube.com/watch?v=PmL31mVx0XA. Accessed 24 Mar. 2023.
Introduction
Can you tell the difference between the composer Ludwig van Beethoven and a computer? Humor us for a moment – this is not a trick question. As artificial intelligence (AI) technology has found its way into every aspect of our lives, music is one of the areas where artists are making use of it.
The question is, how good is AI-generated music? Can it rival symphonies created by a genius like Beethoven? And would you recognize an AI-composed piece of music compared to one created by a human being? These are the questions we will try to answer in this article.
With the recent advancements in AI technology, it has become increasingly difficult to distinguish between music created by AI and music created by humans. AI-generated music has become more sophisticated, and it can now replicate different styles and genres of music with remarkable accuracy. However, many music experts argue that AI-generated music lacks the emotional depth and complexity that human-created music has.
People can often tell the difference between music created by AI and people, as AI-generated music may lack the subtle nuances and variations that come from a human musician’s interpretation and creativity. Additionally, while AI-generated music may be technically proficient, it may not have the same emotional impact on listeners as music created by humans, which often reflects the human experience and emotions.
Table of contents
AI music generating tools
For decades, artificial intelligence was the stuff of science fiction, something that movies were made of. Recently, though, technology has advanced to a point where extremely powerful computers are affordable even for smaller software developers. This sea change has put the development of AI-based apps and software within reach of many more people.
As a result, AI is becoming an integral part of our lives, and the art world is no exception. But how good are AI-based music generators really? Can they produce works that are indistinguishable from the productions of some of the world’s most renowned composers or most beloved bands? Before looking at the answers to those questions, it is worth taking a closer look at some of the leading AI-based music generation tools.
Also Read: 12 Apps and Tools To Make Music With Artificial Intelligence
How AI Music Generators Produce Sounds
AI music generators are essentially pieces of software that produce music. To achieve that, they mimic human behaviors, such as learning and other cognitive functions. Most of today’s leading AI music generation tools are based on one of two principles: neural networks or machine learning.
Machine learning (ML) is perhaps the best-known subset of artificial intelligence technology. ML uses algorithms that continue to learn without additional human input. Once trained on an initial dataset, the algorithm then develops its own predictions in answer to a question.
In the case of music generation software, the algorithm would create a song or the overture to an opera, for example. If that piece of music is considered to be successful, the algorithm learns that its predictions of what makes a good song were right. As a result, it is likely to apply similar principles for its next task.
Neural network-based music generation tools work somewhat differently. Rather than training algorithms that learn from their own experience, neural network programming tries to mimic the way the human brain works.
When it comes to composing, (musical) humans have a sense of which sequence of notes results in a pleasing song or other work of music. To allow an AI to replicate this approach, scientists would recurring neural networks to train sequential models of AI.
You could be forgiven if this is starting to sound too complicated. For years, aside from a lack of computing power, the need for computer science expert training kept the field of AI small. As tools have developed, AI has become more accessible and democratized. For that reason, users no longer need to be an expert in machine learning to use AI-based music generators.
Also Read: These Pieces of Music Were Created Using Artificial Intelligence (AI)
Top Five AI-Based Music Generators
There are numerous AI-based music generators out there, aimed at professional music producers, amateur music enthusiasts, and AI fanatics. This selection is far from exhaustive, but we have aimed to cover tools with a range of different approaches and functionalities.
Our top five AI-based music generators include:
1. Amper Music
Amper Music has become known as one of the most accessible AI music generation tools. It makes it possible for anyone to start using AI-generated music for videos or other projects. The cloud-based platform joined forces with image distribution company Shutterstock.
From its new home base, Amper allows users to create their own tracks with as little or as much input as they like. Users can choose to let the software compose the entire piece, or they can specify any number of criteria to put their slant on the piece of music.
Amper allows budding musicians to define keys, choose their favorite instruments, and determine the pace of the piece. All that is possible without an in-depth understanding of composition or music theory. Users simply need to be able to define what they like and dislike as Amper takes care of the hard lifting.
2. MuseNet
Right now, it is almost impossible to talk about artificial intelligence without mentioning OpenAI. The creators of the now-infamous ChatGPT software have also been working on a music generator. MuseNet is the result of their efforts.
MuseNet’s compositions are based on a deep neural network that was trained with data from online sources. According to its creators, it is able to integrate ten different instruments into its songs and can produce music in more than a dozen different styles.
Plus, MuseNet can imitate both contemporary favorites and some of the all-time greats like Mozart. However, despite its wide-ranging capabilities, this music generator does not yet allow users to compose pieces from scratch.
3. Magenta Studio
Magenta Studio is available to users courtesy of Google’s Magenta research team. The team constructed this music creator on top of its Magenta open-source library that offers access to different machine learning models and tools for art and music generation.
Although the tool is aimed at seasoned composers, it is user-friendly and therefore perfectly accessible for those new to AI music creation. Composers will find a range of accessible interfaces and instruments, including a step sequencer, a drum machine, or even a piano roll editor.
Google’s engineers also added previously trained models to Magenta studio to help users generate chord progressions, fresh melodies, and percussion patterns. The software can deliver any of those, depending on user input. Lastly, it is highly accessible, requiring no downloads and running from a simple browser window.
4. Brain.fm
While the previous music generators allowed users to create all types of music, brain.fm has more of a focus. This platform intends to help its users to become more focused, find it easier to relax, and improve their sleeping patterns.
Rather than instructing the software with a certain musical genre, users define the mood they would like to achieve with the help of the composition. Once users have defined the cognitive state they are looking for, brain.fm uses an algorithm, principles of neural network technology, as well as psychoacoustics to deliver the right piece.
Users can then alter settings such as speed or instrumentation to personalize their session further. The developers of brain.fm claim that their AI music generator has been proven to aid meditation, promote more restful sleep, and improve the attention span of listeners.
5. Algonaut Atlas 2
Algonaut specializes in developing AI-based music production tools, and Atlas 2 is no exception. Using machine learning algorithms, this software allows users to develop melodies, harmonies, and rhythms that are compatible with their existing music library.
One of Atlas 2’s key functions is the ability to understand what is in a user’s music library and match it. Users can import their own samples as a basis for Atlas 2 to work from, but the software also has a library of samples, loops, and sounds ready to use.
Although aimed at professional producers, the software is easy to use for anyone looking to simply have a go at AI music generation. Professional users will benefit not only from new inspiration but also from the ability to export MIDI files which can then be transformed further in any DAW software.
Can we tell the difference between AI vs human generated music.
It is too early to deliver a set-in-stone verdict on whether humans can detect music that has been generated by AI. Early studies seem to suggest that telling the difference is harder than most would think.
The developers behind the Amper music generator conducted their own research, supported by audio researchers Veritonic. They asked a panel of participants to distinguish between stock music, AI-generated music, and music composed by human intelligence. The overall finding was that the average person was unable to tell the difference. As if that result was not surprising enough, 25% of the panelists said they would value a piece of music in an advert more if they knew that it had been composed by AI.
Granted, perhaps a study conducted by the makers of an AI music generator has some bias. Therefore, it is worth looking at other tests. Customer service platform Tidio conducted a wide-ranging survey test comparing AI-generated art to human creations in 2022. Their research covered a wide range of AI-generated content, including AI-generation faces, music, and AI-written content.
Survey respondents stated that they found music to be the category where it was hardest to distinguish between human and machine. One of the reasons for the difficulties may lie in the panelists’ inability to separate composition and performance. A human-generated electronic dance music track, for example, was considered to be AI-produced by seven out of ten participants.
On the other hand, a song composed by an AI trained on Beatles songs was perceived by anonymous respondents as human-created. Perhaps the most surprising result of this research was an underlying sense of human inferiority. Participants attributed music they believed to be ‘too good’ or ‘too complex’ to AI, whereas more ‘chaotic’ pieces were ascribed to human composers.
One last example to illustrate how close AI-based music generators are to human creations. Sony Computer Science Laboratories created an AI to try and create music like that composed by Baroque composer Johann Sebastian Bach. Bach’s music is known for its complexity. The scientists developed a neural network called DeepBach and trained it on the chorales of the composers.
The creations by DeepBach had 50% of listeners convinced they were hearing the real thing, whereas 75% identified the real composer correctly.
Future of AI in music
Artificial intelligence is establishing its place in the art world, including image generation through the likes of Stable Diffusion and music. Where exactly that place will be is not yet known. Amper’s test mentioned above compared AI-generated music to stock music, which the majority of people would use as background music in everyday life rather than something they listen to attentively.
Perhaps this is where AI-generated music is headed, or perhaps machine-made scores will become the mainstay in fields like advertising. In those cases, music is one factor among many, but it is not the main act, so to speak. Saying that, there is no reason why the role of AI-generated music should not expand with advances in related technologies.
Also Read: AI Generated Music from Audio Wave Data
Conclusion
Autonomous vehicles, the future of healthcare – no matter where you look, AI is making an impact on our quality of life. Whether it is by taking care of repetitive tasks or creating sought-after online content, greater segments of society will likely become more dependent on these technologies.
At the same time, AI will no longer be the exclusive remit of someone who is an expert in systems design, but it will become more accessible to everyone. As a result, its role in real life will expand further. What that means for music remains to be seen.
Also Read: MusicLM and AudioLM Google’s Text to Music and Audio Tool
References
ColdFusion. “A Deep Look at A.I. Generated Music.” YouTube, Video, 22 Sept. 2021, https://www.youtube.com/watch?v=EyeW_axUEQU. Accessed 24 Mar. 2023.
—. “Human or AI: Can You Tell Who Composed This Music?” Futurism, 16 Dec. 2016, https://futurism.com/human-or-ai-can-you-tell-who-composed-this-music. Accessed 24 Mar. 2023.
TwoSetViolin. “Can You Tell the Difference between AI and Human Composers?” YouTube, Video, 23 Sept. 2020, https://www.youtube.com/watch?v=PmL31mVx0XA. Accessed 24 Mar. 2023.
https://www.tidio.com/blog/ai-test/%23ai-music. Accessed 24 Mar. 2023.
Share this: