The New Economics of Audio Stem Separation: Driving Revenue for Artists, Producers, and Publishers

You can fly your DAW just fine. So why would you need a solution to separate your audio stems? Dig deeper into audio stem economics, and you’ll get your answer.

Software platforms that tackle this task have been on the market for several years, but the latest entry into the sector, AudioShake, saw room for improvement. Using artificial intelligence (AI) to separate songs into stems, the company has seen a quick uptake for its enterprise platform, AudioShake Live, which is currently in use by all three major label groups. Top publishers including Hipgnosis, Primary Wave, Spirit, peermusic, Downtown, plus distributors like CD Baby, Sun, and CODISCOS are also on board.

Earlier this year, they followed up with the launch of AudioShake Indie, providing indie artists and producers with a fast, affordable way to create song stems and instrumentals. The focus: empowering these users with new monetization opportunities and uses for their recordings.

Credit AudioShake’s ascent in large part to their advanced AI, capable of producing broadcast-grade audio stems that are, for instance, ideal for responding to a music supervisor’s creative brief on a moment’s notice. Also note the sharp instincts of Co-Founder Jessica Powell, who had previously been head of communications for Google and a member of its management team. AudioShake co-founder Luke Miner is the former head of data science for fintech company Plaid.

Why did this tech-savvy pair aim their skills into the audio world, and what are the new revenue opportunities that stems open up for artists, producers, publishers and labels? Powell provides a fresh perspective on sonic entrepreneurship in this comprehensive Q&A.

Jessica Powell is CEO and Co-Founder of AuioShake

Jessica, what got you interested in audio stem separation, and co-founding Audioshake? How did being VP of Communications at Google lead you to this corner of the audio industry?

I feel like sitting out in the Valley you hear these really tidy origin stories of how things come about. It was more of a collection of experiences. When I was a teenager playing bass I was learning the standards, but I was really into punk and I wanted to figure out how to play the bass line to “Holiday in Cambodia” and “Waiting Room.”

I wished I could pull apart those songs. Punk can get pretty gritty and you get the feel of it sometimes, but not necessarily the technical aspects — although sometimes being technical is quite the opposite of punk! Then years later, I ended up living in Japan and I’d go karaoke there, and I would be like, “What if you could just have karaoke for anything in the world?”

The final piece of it was that my first job — even though the vast majority of my career was in tech —  was actually in music. I worked for the Paris-based performance rights organization CISAC.

When I left my job at Google, I knew I wanted to get back into something, more creative, something in the arts. My co-founder, Luke Miner, and I — everyone on the team —  is a musician. We were very interested in working with music and we wondered if AI had gotten to the point where it could break a song apart in a cleaner way than it had done in the past. People have been working on this problem for well over a decade, but the problem is it always produced a lot of artifacts — that’s what really drew us to the problem.

What were some of your early experiences in designing the AI engine behind Audioshake?

The first song we tested it on was a Smiths song, and Morrissey’s voice sounded demonic. Our first version of it was terrible, but there was something in it. We could tell that we were doing all these things wrong, but that at the same time that the separation was much cleaner than it had been in the past.

We spent two years just building the tech and then we launched it in July of 2021. Our first platform was for rights holders, major labels, and most of the top publishers in the US that range from A&R people to sync people.

The Business Model: Empowering Artists

What made you decide that stem separation would be a profitable business model potentially?

This was a passion project where we thought that yes, there probably was some money to be made. But it wasn’t like we woke up and said, “You know what? The next big business to be created in the world is in stem separation.”

It was just something that we felt very strongly interested us intellectually. And being musicians, caring tremendously about art and helping artists make money, it was about trying to do it in a way where you put the power in their hands. This rather than something where their art is taken from them. All of that was what drove us to creating it, and drives us.

There was never an Excel spreadsheet that said, “This is how much money we could make.” I think if we had done that we probably would not have started Audioshake, because there are probably easier businesses.

What we did try and figure out, in tech terms, is what you’d call the minimum viable product (MVP): What is it that we would need to release to have a sense of whether what we were building was actually useful for people?

We said, “OK, what is it that we’re building exactly? You can separate a song, but how are you actually delivering that to people? And what does that look like? And who is your initial customer? Who is it that you eventually want to serve? And how do you reach all these people?” That was much more rigorous.

AudioShake bases its stem separation on advanced AI.

How Stems Open Up New Revenue Streams

Something I’m sure that readers will be interested to know is, “How can I make more money by having better control of my stems?”

There are current uses, and then there’s future uses. Opportunities today are in sync licensing, for independent artists through to major label artists. It is a growing segment of the music industry, and music supervisors almost always want the instrumentals to be delivered — they always want to be able to work with as many elements as possible.

Potentially, the lack of an instrumental can lose you that deal, because there’s usually sync teams pitching hundreds of songs.  There’s usually many good candidates for a commercial or a movie, or whatever the media might be.

When we started talking to sync departments at publishers and at labels, we learned that 30-50% of the requests that come through, wouldn’t be able to be fulfilled because of a lack of instrumental. So I think for artists being able to have, at a minimum, the instrumental for their song is very important and opens up sync licensing.

From there, if you’re starting to deal with artists who perhaps are a little bit more popular, where people are looking to remix those songs acapella, for instance, having the full stem set is very useful. In addition to sync licensing, we’ve seen interest in having stems available for remixes and spatial audio.

What’s a way that you’re seeing quality stem separation be of use to record labels?

Labels are using AudioShake to pull songs apart that don’t have their stems. It actually still happens quite often today that stems aren’t passed on in a catalog, or they’re passed on but they’re incomplete. And depending on the age of the song, you could be dealing with mono-track recordings, analog tapes that are lost or damaged, or — more recently — just issues with plugin compatibility and people not being able to open the files.  So they’re using AudioShake to pull the song into stems, and then a remix or a spatial mix could be created from that.

Where I think things are going to go in the next couple of years is AR/VR experiences, with spatialized audio and adaptive music. Imagine if you were wearing some kind of fitness device, you’re walking and listening to an Ariana Grande song, and you’ve got not just a BPM shift, but a genre shift if you move from walking to running. Or in a video game, imagine a much more dynamic, almost customized audio experience where the audio changes based on what your avatar was doing, and that entire environment is stem-based.

Social media — like YouTube and TikTok — is where the creator tools are going to get better and better, to the point that users can do some pretty cool stuff with audio editing in those apps — just like we already can do with video and images. So I think there’s just going to be more and more opportunity with stems as well.

What does this mean from an artist’s perspective?

That there’s going to be so many more opportunities for your music. You’re not just going to be dependent on streaming and live. There’ll be these other areas where sometimes it might be more background or part of a larger gaming experience. But it’s a way to expose people to your music, for you to make money, and to drive traffic to the original music.

Why AI, When a DAW Will Do

A lot of SonicScoop readers at this point are reading this and thinking, “I have the stems of what I’ve produced at my fingertips. So how can I benefit from this type of platform?”

If they have their stems, they should use their stems. I’m not going to tell someone they should come use ours instead of theirs — use the real stems!  That said, with instrumental and acapella stems, it’s usually pretty hard to tell the difference between ours and stems that are created in a DAW. If you were looking for something super quick, you could use AudioShake because it [takes] 15 seconds.

However, I would say that I imagine that actually most of your readers probably do have project files that have been lost. In fact, I don’t know a single producer that does not experience that. And even if it is standard practice for them, working professionally, to hand over stems, they probably have a lot of interesting creative projects of their own where they didn’t take the time to produce the stems yet, because it’s time-consuming and annoying to do.

What we’ve found from folks coming through the website — for example there’s a producer who works for big artists like Diplo — and he’s been using it largely for lost files. But he’s also talked about using it as a creative tool, to go into old projects , quickly pull them apart and listen to, “OK, what was going on in the drums?” Not necessarily something that would end up being a final product, but a pre-production exploratory element.

When Audioshake decided to tackle this problem, what made you decide to focus on AI as the solution? What was the process for actually applying AI to stem separation?

Luke led data science at a company called Plaid, and we both come from tech backgrounds. The area that our team is very experienced in is artificial intelligence within music. So that was a natural progression for us.

We trained our model on thousands of real stems. Think about your Android or iOS phone — if you go into your photos and search for the word “tree” it’ll return photos that you’ve taken of trees. It’s not because you ever told Apple that those were photos of trees, but because Apple’s AI has trained on thousands of photos of trees.

In a similar spirit, we trained on thousands of stems, like vocals and drums. We teach the AI what those stems sound like, which then allows it to take a song that it’s never heard before and pull it apart.

Entrepreneurs and Audio: The View from the Valley

Speaking from an entrepreneur’s perspective, what would you say has been both the most rewarding and the most challenging aspects of getting AudioShake off the ground?

I think what is so hard about being an entrepreneur is that there are always so many people that will tell you that your idea is bad and that it’s not going to succeed  — and they’re doing that with great intentions.

You almost have to be slightly sociopathic in not listening. You have to just make sure that the voice in your head is louder than the voices around you so that you can continue to push on. That’s true not just in business, but in creative projects too.

Then there’s raising the funds to be able to carry out your business plan — and then whatever funds you have, you may still have less than the competition. So I think with any project that you’re going into, a lot of it is about, how do you prioritize from a financial or human resources perspective knowing that you can’t do everything?

You can’t have all of the resources, so how do you just stay super-focused? One thing I’ve noticed is that people will come to us with great ideas about stuff we should be doing. They’re great ideas, but they’re different businesses. Simply staying really focused is a big challenge.

What are your tips for people to maintain that focus, and stick with it when things feel challenging?

It’s remembering what you set out to do. It doesn’t have to be some cheesy mission statement, but something like, “I started this company/project to do this thing — not to do ‘B,’ ‘C’ and ‘D,’ but to do ‘A.’ Just remind yourself of that. Be able to look at the bigger picture, recognize that it’s going to take a while, and be patient yourself.

Pursuant to that outlook, what’s going to make you feel like AudioShake is a success. What’s your barometer for that?

We would really like to be able to help power a wave of new audio experiences in the world. It does not matter to us whether you necessarily know our brand or who we are. If we can help fuel this kind of creation, and make it easier for artists to drive revenue from their works in a whole bunch of different ways, we will see that as success.

The coolest part is seeing the creative things that people do with AudioShake. One of our customers is (music intellectual property rights company) Hipgnosis in the UK, and they are actively working on how to best represent this super-iconic catalog that they represent. Whether it’s Rick James or Fleetwood Mac, they’re saying, “What can we do to reinvent these songs for a new generation? How can we change that up or remix it?”  Those are really fun projects to see.

— David Weiss is an Editor for SonicScoop.com, and has been covering pro audio developments for over 20 years. He is also the co-author of the music industry’s leading textbook on synch licensing, “Music Supervision, 2^nd Edition: The Complete Guide to Selecting Music for Movies, TV, Games & New Media.” Email: david@sonicscoop.com

Please note: When you buy products through links on this page, we may earn an affiliate commission.