11 Startups Using AI to Compose Music
Table of contents
It’s a common and generally accepted misconception that artificial intelligence (AI) will only automate and revolutionize repetitive and menial jobs – like HR and recruiting. In our article “What Jobs will Robots and AI Do in The Future?”, we looked at some jobs that may already have casualties from the AI revolution. As we would expect, most of them are administrative or repetitive, requiring little specialized knowledge. Then we looked at 10 Startups Coming for White Collar Jobs and changed our minds. AI is now encroaching upon areas of human creativity we would all considered talent-based – like logo design. Of course some people are now questioning if talent is a real thing, and we can tell you after long and illustrious careers in the corporate world, that yes, talent is a real thing that few people actually possess.
Music is another genre where some true talent can be found. Just spend some time watching Prince perform if you think that music doesn’t have talented people. Unfortunately there’s another act coming to town that may never stop becoming more and more talented. While would-be musicians are debating whether it’s possible for a program to write music that’s indistinguishable from a human, AI is already having a field day. Since as early as the ‘50s, experimental composers have been using randomized statistical models to write music. Even David Bowie used a custom text randomization software called Verbasizer in some of his work – like Hallo Spaceboy. Let’s take a look at 11 startups that are using AI for composing music and augmenting tomorrow’s musicians.
AI Musicians that Compose Music
Founded in 2014, New Yawwk Startup Amper Music has raised $9 million to develop a cloud-based platform that composes, performs and produces unique AI generated music to accompany any content. Essentially, Amper is a technologically advanced version of production libraries that contain pre-created, licensable music for the film and gaming industries. The idea behind the business model is that generic, non-exclusive scores are not enough, and the traditional search and licensing process is frustrating. That’s where Amper Music comes in. Their team created AI algorithms that collaborate with the user, making it easy for anyone to generate scores based on musical genres (like classic rock, cinematic, or 90s pop), moods (driving, happy, reflective, etc.) and length. The pro version of the application allows for further customizations like instruments, tone, time signature, or key. The app comes with an Application Programming Interface (API) that can be used to embed it into other software along with connectivity to Adobe Premier Pro. Amper will use its latest funding of $4 million received March this year to expand internationally and double its employee base in the US.
Speaking of cheezy Euro pop, we now move across the pond to look at London startup Jukedeck which has raised $3.4 million to develop an AI music composer similar to Amper. The team feeds its neural networks huge chunks of musical data to analyze, then the AI learns to write original music, like these examples. Jukedeck’s compositions cost $1 for individuals or small businesses, but are free as long as the company is credited. Larger clients pay around $22 per track, and there’s also a monthly subscription model. As a comparison, stock music for a 45 second exclusive track for a commercial costs $925 on stockmusic.net. Jukedeck has a diverse customer base across 169 countries including large names such as Coca Cola and Google. They’ve created more than 500,000 tracks already.
Founded in 2012, San Francisco startup Melomics Media has raised an undisclosed amount of funding to develop an autonomous AI music composer. There is no human intervention in Melomics’s songs at all, since they’re created from scratch on an Iamus supercomputer. In 2010, Iamus composed the first contemporary classical piece created by a computer in its own style, titled Opus One. Melomics later extended the supercomputer with the Melomics 109 computer cluster that produced the first popular genre AI album, 0music. The AI has released roughly 1 billion pieces since then and all of them are available royalty-free in the public domain. Melomics’s relaxing music has been used in a stress reduction clinical trial as well in 2014, funded by the Spanish Ministry of Science and Innovation. Releasing music for free and not making any money on it is super popular these days.
Inspired by the movie “Her” and founded in 2016, Luxembourg startup Aiva Technologies has received $768,000 in funding to develop an AI script composing emotional soundtracks for films, video games, commercials, and other entertainment content. The AI was taught using the works of classical composers and released its first oeuvre, “Opus 1 for Piano Solo”, in early 2016. Since then, Aiva became the first AI algorithm to be recognized by the French Author’s society SACEM, released an album titled “Genesis”, wrote music for Nvidia’s keynote introduction, and produced the main theme of the video game Pixelfield. Aiva’s services are aimed at corporate clients and prices are not made public.
Staying with the Eurovision theme, Berlin startup Melodrive was established in 2016 and has taken in an undisclosed amount of funding to build AI algorithms that compose music. In the words of the company “Melodrive is an AI music engine that empowers users – no matter their level of musical knowledge – to create immersive soundtracks at the click of a button.” Their white paper on “How Music Can Boost User Engagement by 40%” implies that everyone should be using their tool to create unique music in seconds. Whenever they launch their product, we may generate some soothing music to accompany our articles. Wouldn’t that be engaging?
Lastly, some folks out of San Bernardino, California have developed an app called Algotunes that creates AI music on the fly. You login to the site using Facebook or whatnot, then choose some variables like “style” and “mood”. One minute later, a musical score pops out which you can then download and use for free or pay a small fee to license. We tried it and let’s just say, free was a bit expensive for what came out of that thing. Try it and tell us what you think.
(Editor’s Note: To wrap up this section we’d like to send out a few messages to some other AI music startups out there. Groov.ai, we know you exist. We didn’t “miss you”, it just that you’re in the early stages of development and there’s not much to say about you. AI.Music, you may have raised 5 million British pounds but we couldn’t decipher anything from the vague description you gave of your product. To those other music startups we didn’t include, fire your PR people and give us a much smaller amount of money instead for some proper coverage.)
AI That Complements Musicians
Not all AI applications in music involve pure composition. Some other startups have recognized the need to offer services to the David Bowies of the world, providing tools that use AI to help musicians compose better music. Here’s a look at a few startups doing this.
Founded in 2017, Brisbane, Autralia startup Popgun is a relative newcomer to the AI music scene having completed the Techstars Music accelerator program the same year Amper Music did. They received an undisclosed amount of seed funding early this year. The team uses deep learning to build an AI musician that learns from live human performances and, in theory, will be able to complement human players in real-time, practically jamming together with them. Popgun’s first project was AI Alice, who can predict what a musician will play and accompanies him or her, even improvising upon the human player’s score. The platform will be launched later in 2018.
Founded in 2013, San Francisco startup Humtap has raised undisclosed funding to develop an app geared towards social media sharing of your music videos. The team is one of the members of Abbey Road Red, the musical incubator established in 2015 at the legendary recording location of the Beatles and Pink Floyd. Humtap users can choose a preset musical style courtesy of the AI’s database, then tap and hum their way through anything, and the algorithm will turn it into pretty decent music complete shite. The newly created song is added to a smartphone video, ready to go viral on Facebook. As the other startups above, Humtap sidesteps the issue of copyright as the tune is made by the user and doesn’t use any pre-recorded material. The team just showcased their newest feature, Arabic music generation, in Dubai a fortnight ago. Frankly, this may be one of the dumbest most creative applications of AI we’ve seen to date.
Founded in 1999, New Yawwk startup Realtime Music Solutions has raised an undisclosed amount of funding to develop musical enhancement tools for professionals. Their flagship product called Sinfonia is the industry leader in orchestra enhancement. It complements live professional and amateur musical theatre performances when the performers cannot provide a full band for whatever reason. Sinfonia has been used in over 100,000 performances, including ones on New York’s Broadway, London’s West End and with Cirque du Soleil troupes. Their other tools include a rehearsal app, a line-learning app, and a keyboard programming tool.
(Editor’s Note: We couldn’t find any mention of Realtime Music Solutions actually using AI on any of their company collateral. In fact, they don’t even appear to be a startup. However if they’re willing to fork over some cash, we’ll help them to plaster phrases like “cognitive symphonics” and “machine music” all over their website – and then raise a ishtload of funding shortly afterwards).
In case you want to jump on the bandwagon and use AI to create your own music, we found something from zee Germans called Ludwig Music Software which has created software to compose music in collaboration with AI algorithms. Users need to play or upload a melody, choose the musical style from the 50 available, and Ludwig finds the proper chords, writes all parts of a professionally sounding band, and accompanies the piece. Founders Matthias Wüllenweber and Frederic Friedel started the company as a side project stemming from chess software when they realized that the algorithms applicable to chess also work for music composition and arrangement. You can get the software for free when their website is up again.
Conclusion
It’s not only startups dabbling in this space. Google’s Magenta project and Sony’s Flow Machines are both looking to use AI to compose music. There are also a handful of independent university research projects in the field, like MorpheuS, ChucK, or the Computer Music Project at Carnegie Mellon University. While the knee-jerk reaction of artists and producers might be “AI’s going to steal my job”, most of these companies highlight the collaborative aspect of their tools, elevating rather than replacing human creativity. No point in fretting over it, instead just should listen to this other Bowie masterpiece that uses the Verbasizer. Feel better now?
Share
If you’ll be updating this list – probably, it’s time to add there Musical Blockchain, a first true AI-Composer, psychology-based – which is one-year old – so old news with time-proof 😉
Info about technology/solution and example applications in music, art, education, urban/city development, social stuff, medical research and more – at musicalblockchain.com