The Evolution of Generative AI for Music

It seems everywhere you turn various companies are popping up in the realm of generative AI for music like it’s ‘the latest thing’ – just in the past couple of weeks we’ve seen launches of both Suno and Udio . But generative AI for music is not nascent – it’s been around for decades in various forms; it’s just that over the past decade or so, the machine learning and compute power has caught up with the vision, making the process of leveraging computer technology to create songs on the fly much more seamless.

The history of generative AI for music involves a fascinating evolution of technologies that have progressively enabled computers to create music, either autonomously or as tools for human composers. Here's an overview of this development:

1. Early Experiments (1950s-1970s)

The exploration of generative music began with the advent of computers. One of the earliest examples was the Illiac Suite for string quartet (1957), created by Lejaren Hiller and Leonard Isaacson. This was the first piece of music composed using algorithmic composition techniques with a computer.

2. Algorithmic Composition and MIDI (1980s-1990s)

During the 1980s and 1990s, the rise of personal computers and MIDI (Musical Instrument Digital Interface) technology facilitated greater experimentation and development in algorithmic composition. Software like Csound, an audio processing and synthesis language, became popular among composers for creating complex sounds and music structures.?

This was the era when I really entered the space myself – I was working at Keyboard Magazine (and Guitar Player) as I was putting myself through college when MIDI first came on the scene in the mid-80s.? And a few years later,I joined Apple to launch its earlier focus on music and entertainment; our first foray into the space was launching the MIDI interface for the Mac in early 1988. That was my on-ramp to leading and overseeing the focus on for Apple music during my lengthy tenure with the company as we shaped and grew the market.

3. Machine Learning and AI Integration (2000s)

With the growth of machine learning in the early 2000s, researchers began to apply these techniques to music generation. Programs like David Cope's "Experiments in Musical Intelligence" (EMI) made headlines for creating music in the style of classical composers like Bach and Mozart, using rule-based AI systems.

4. Deep Learning Revolution (2010s-present)

The introduction of deep learning has dramatically transformed the potential of AI in music. Neural networks, capable of learning from large datasets of music, began to create more convincing and complex musical compositions. Notable developments include Google’s Magenta project, which uses TensorFlow to create tools and models for music generation, and OpenAI's Jukebox, a model that generates music, including rudimentary singing, in various styles and artist likenesses.

5. Commercialization and Accessibility (2010s-present)

Recent years have seen the commercialization of AI music technology. Startups and companies have developed products that help musicians with composition, accompaniment, and even performance. Tools like Amper Music (founded in 2014 & acquired by Shutterstock in 2020) and Jukedeck (founded by Ed Newton-Rex in 2012 & acquired by Tik Tok/ByteDance in 2019) offered platforms where users could create music using AI-driven systems, making the technology more accessible to non-experts.? I briefly advised another ‘early stage’ venture in the space at that time called Humtap with a similar orientation.

In the Fall of 2023, not long after Google unveiled MusicLM, a text-to-music tool that creates songs from a simple prompt, Paul McCartney used AI to pull out John Lennon’s voice for a new Beatles track (that otherwise wouldn’t have been possible). And Meta launched an open source music generation model (called MusicGen) that turns a basic text prompt into a respectable quality music sample.

--------

Far from being ‘nascent’, some would argue that the generative AI for music space is actually somewhat crowded these days. Let’s take a closer look at just a couple of the active companies in the space some of which are only a couple of weeks old. When analyzing generative AI music platforms like Boomy AI, Suno, and Udio, it's important to consider various factors such as their technology, user experience, functionality, and target audience. Here's a deeper comparison based on these elements:

1. Technology and Features

·?????? Boomy AI leverages AI to allow users to create songs quickly and with minimal input. Its technology focuses on instant music creation, with algorithms that handle everything from composition to production. Users can generate complete songs by simply selecting a genre or mood.

·?????? Suno uses AI to assist with the composition process, helping users create melodies, chord progressions, and even complete arrangements. Its technology is geared towards augmenting the creative process, providing tools that help users apply music theory effectively, thereby bridging the gap between creativity and technical knowledge.

·?????? Udio provides detailed tools for mixing and mastering, making it a powerful choice for finishing tracks. Its AI analyzes audio tracks to optimize sound quality, balance levels, and apply mastering effects to enhance the overall production.

2. User Experience

·?????? Boomy AI offers a straightforward and accessible interface, ideal for beginners or non-musicians. Its rapid generation process appeals to users who want quick results without delving into the complexities of music production.

·?????? Suno might require a bit more musical knowledge to fully utilize its compositional tools, but it also offers educational components that can help users learn as they create. This makes it a good middle ground for hobbyists or semi-professionals.

·?????? Udio tends to be more complex, suitable for professionals or serious enthusiasts who have a good understanding of audio engineering. Its interface and tools are designed to offer precise control over the audio production process. That said, as a non-musician, I was delightfully surprised at how quick and easy it was to create an original song that actually sounded pretty good on the fly!

3. Functionality

·?????? Boomy AI is highly functional for generating a wide variety of music quickly. It's less about control and more about exploration and automated creation.

·?????? Suno provides functionality that supports both creativity and learning, allowing users to experiment with complex compositions using AI-supported suggestions.

·?????? Udio excels in post-production, offering functionality that can elevate a good track to the level of a professionally polished song.

4. Target Audience

·?????? Boomy AI targets casual users, content creators, and social media influencers who need a quick supply of custom music.

·?????? Suno is aimed at musicians, music students, and hobbyists who want to deepen their compositional skills or break through creative blocks with AI assistance.

·?????? Udio is best for music producers, sound engineers, and professional musicians who require advanced tools for finalizing their tracks.

Effectively, choosing between Boomy AI, Suno, and Udio depends largely on where one is in their music creation journey and what aspects of the process they value most. Boomy AI is best for rapid creation and simplicity, Suno for enhancing musical creativity and learning, and Udio for professional-grade production quality. Each platform serves its niche effectively, making them valuable tools in the evolving landscape of AI-assisted music production.

SUMMARY

As instruments, recordings, and synthesizers launched and evolved over time, they impacted and amplified the number of both creators and consumers of music. Now with generative AI music tools available, lines are blurring between artists, producers, and consumers. Now consumers with no musical knowledge, training or experience can create original songs virtually friction-free on the fly and effectively are becoming members of the creator economy as well as fans.

While the quality of the output of such consumer-generated tracks could be argued, by dramatically closing the gap between idea and creation, AI is a powerful tool that allows more people than ever to create music while at the same time enhancing the creative possibilities for existing professional artists and producers.

As AI music technology advances, it also raises ethical and creative questions about authorship, creativity, and the economic impact on musicians. The boundary between human-created and machine-generated art becomes increasingly blurred, leading to discussions about copyright, the nature of creativity, and the role of artificial intelligence in art.? I’ll have more to share about this over time.

Meanwhile, Generative AI for music continues to evolve, pushing the boundaries of what machines can achieve in creative domains and transforming how music is composed, performed, and experienced.

There’s a lot more to share on this topic – for those who are seriously interested and want a deeper dive, I recommend these articles for starters for a deeper dive:

*The Future of Music: How Generative AI is Transforming the Music Industry (by a16z team)

*New AI-powered ‘instant’ music-making app Udio raises $10m; launches with high-profile backers

*Suno is a music AI company aiming to generate $120B/year; but is it trained on copyrighted recordings? (Op-Ed by Ed Newton-Rex, a veteran expert in the Gen AI for music arena)

*Yes, Udio’s Output Resembles Copyrighted Music Too ? (this is also an Op-Ed by Ed Newton-Rex)

?

Let’s Keep in Touch!

At my site, KelliRichards.com you can sign up for my monthly newsletter, peruse my archive of articles, and learn more about us working together. On the site, click on Connect, then scroll down to Join Our Mailing List in the footer. If you like what you read, please spread the word with your network and share the news. Thanks for being a part of our community!

?