The Profitable Musician | Bobby Owsinski | Artificial Intelligence

 

Artificial intelligence is evolving and becoming ubiquitous at an extremely fast pace. It can be a bit confusing to follow what’s the latest on it and how it directly impacts your chosen industry. Bree Noble explores how this particular technology reshapes the music business with Bobby Owsinski, author of The Musician’s AI Handbook. Together, they explain the most ethical ways to use AI to create music, as well as some useful tools to help you out. Bobby also talks about the somewhat blurry lines that separate AI-generated music and copyright laws by explaining how to hone your creativity without putting yourself in a giant legal problem.

Watch the episode here

 

Listen to the podcast here

 

Discussing Artificial Intelligence And Copyright With Bobby Owsinski

Bobby’s Four Careers

I am so excited to be here with my friend Bobby Owsinski, also known as Bobby O. I think we’ve known each other for at least five years, and you haven’t been on the show. I was like, “What the heck? How did I not have him on the show?” I am so excited to have him here because we’re going to talk about stuff around AI. We’re also going to talk about some stuff around mixing because that is one of his biggest expertise but he’s diving into the world of AI.

He has a book out around that and music so we’ll get into all of that but first, since he’s never been on the show, I want to make sure that you guys know who Bobby is if you don’t know him already. If you’ve been on Amazon, you’ve seen that he has tons of books. Maybe you’ve heard of Bobby Owsinski before but Bobby, if you can let everybody know your background in music and how you ended up writing so many books around music and some cool experiences that you’ve had working with other people in the production and mixing world.

First of all, thanks for having me, Bree. I was wondering when you’re going to ask me so here we are. What I’m doing now is a fourth career. My first career was as a musician, and certainly, I was a player. I was playing four nights a week when I was still in high school. I have lots and lots of stage time. I did that until I was about 40 and decided that it wasn’t as much fun because by that time I was touring, after you hit 25 or so, it’s like, “This is work.” It’s not as much fun anymore. I was already working a lot in the studio and then I dedicated myself to being a producer, mixer, and engineer. I did that for a lot of years.

How did I get into writing? That was career two. How I got into writing was still as I was a player, I was on a tour bus and the bass player came on one day and said, “I got a job writing for The Music Paper.” The Music Paper was a weekly paper out of New York that had everything about music and all the clubs. It was oriented more toward musicians and artists, but it was also consumer-facing.

This is one of these things where you go to a supermarket and it is there for free. You just pick it up. It was a big deal at the time. He came on and said, “I got a job writing for The Music Paper.” I thought, “If he could do that, so can I.” I started to put feelers out to various audio magazines, and sure enough, I got a gig writing, first of all for Mixed Magazine. It was my very first article. It then exploded from there where next thing I knew I was writing for 12 or 14 different industry magazines, billboard, Grammy Magazine, variety, recording engineer, producer, EQ, and all those.

That was fun. I met a lot of people doing it. I did a lot of interviews. As a result, I met a lot of people that I hadn’t run into prior in the studio. I was still working in the studio, but then that opened some doors for me. What happened was I was a pretty good engineer, but I was not a good mixing engineer. This was pointed out to me over and over. My clients, the A&R people, and stuff, which is no fun. I knew I had to get good at it or else.

The thing about it is I knew all the best mixers. I knew them either from running into them in the studio or through interviewing them for one of these magazines. I thought, “Let me go ask them what they do and how they do it.” I interviewed 25 across the world and across different genres. I thought, “If I want to know this, I bet a lot of other people do too.” That became the basis for my first book, The Mixing Engineer’s Handbook, which is still my biggest seller.

At the time, people said, “You can’t write a book on mixing. It’s too subjective.” I thought, “There’s none out there, so let me try.” It turned out that there was a voracious appetite for that. It was immediately picked up by colleges all over the world. It’s still used in music production programs in colleges all over the world as a textbook. What usually happens, is you write a book and you go, “I’m never going to do that again.”

I have not done it again. It’s a big deal.

Since the first book had sold so well, the publisher was on me to write a second one, which became The Recording Engineer’s Handbook, and then a third one and a fourth. It got easier. By the time I hit the 20th book or so, it was fairly easy and fun. In total, I’ve gone through the process 50-some times because I’m in the fifth edition in some books and in the fourth edition in other books, which is like writing a new book every time you do it.

I think I have 27 different titles out there right now. It’s not all in the music business, but most of them are in the music business. I hit a ceiling because I thought to myself, “I can write another book but I don’t know that I’m going to make that much more money doing so. A lot of brain cells are going to die in the process. Let me see if there’s something else.”

Lo and behold, I was asked to audition for Lynda.com and Lynda.com is now LinkedIn Learning because it sold out a few years ago. I auditioned twice, once on camera and once doing voiceovers. I kept on saying, “I have millions of views on YouTube for stuff like this.” I said, “No. You still got to do this.” That turned out to be twenty-some courses for them. After a while it was like, “I can pretty much do this myself.” That became career number four and something I’m in now. Between writing books and doing online courses, that takes up a lot of my time these days. That’s my background[Ma1]  in a nutshell.

I think the mixing thing, I’m glad you didn’t agree with them and say, “Yeah, it’s too subjective,” because it is subjective but you have to have some concrete things to start from and that’s why the book is so helpful, I’m sure.

It turns out that’s one of the things I’m good at that I didn’t realize it was good until I started to do this but it’s taking up two subjects and making them easier to understand and putting them in a step-by-step basis. Again, it’s something I never expected that I could do, and it’s a superpower in a way. That’s helped me sustain a career.

Mixing Mistakes And Tips

I’m sure that’s why you decided to write on AI because that’s one of those things too that’s super confusing and elusive and that kind of thing. We’ll get into that but I wanted to stay on the mixing a little bit longer because you’re here and you’re such an expert in this. What are those major mixing mistakes or things that people who are starting to learn how to mix struggle with?

There are a number of them and the biggest one is paying attention to equipment more than the space that you’re mixing in, which is critical because you can’t mix it if you can’t hear it. That’s the number one thing. If you can’t do anything with your space, there are now good programs that simulate mixing spaces with headphones. The Slate VSX is very good. Waves have Abbey Road Studio 3, so you can use any kind of headphones with that. That works well.

The Profitable Musician | Bobby Owsinski | Artificial Intelligence

Artificial Intelligence: When mixing, pay attention to the equipment rather than the space you are doing it. You cannot mix it if you cannot hear it.

 

Spend a little money on that, and that will take you a long way. The next thing is not learning how to listen either loud enough or soft enough. What I mean by that is, that this is a trick I learned early on. If you don’t mix loud enough, then you can’t hear what’s going on with the low-end, at the base. Usually, if you’re mixing too quietly, you’ll end up getting the wrong balances down there. You don’t have to do this for long. A couple of minutes is enough, but you have to move some air if you’re doing it on speakers in order to hear what’s happening. That’s the first thing.

The very last thing is a trick I learned from talking to all these great mixers, and it’s to turn the level down as low as it will go so you’re at whisper level or below as the final step in your mix. Things will jump out at you that you’ll go, “That’s too loud,” or you’ll go, “I can’t hear something.” It will allow you to get the very last balances together, and this will allow your mixes to translate across other playback systems as well. It’s such a simple technique, but it makes a huge difference. Those are a couple of things right now.

You don’t need complicated equipment to do those things. That’s what’s so great about knowing the tricks of the trade or whatever. As you said, people think that it’s all about the equipment, but sometimes it’s about the space and knowing what to do to find the things that might not jump out at you otherwise.

If you go look at my YouTube channel, there’s a video up there. It’s from a lecture that I did in Vancouver at a college. I can’t remember the name of it. It was How to Improve the Sound of Your Room for $150 or Less[Ma2]. This is assuming that you feel comfortable with using a hammer and putting together your own sound panels, essentially. If not, you can buy them very easily now, and you don’t need many to make a big difference.

One of the things that people confuse here is that “If I put these things up, it’s going to isolate everything,” or, “I’ll get all that acoustic foam and I’ll put it all over the place, and then I can yell and nobody will hear me. I won’t hear the gardener outside,” and all that but there are two different things. Isolation is entirely different from acoustic treatment. You have to get your arms around the fact that isolation is expensive and there’s no cheap and easy way to achieve that. However, on the other hand, you’ll find that acoustic treatments for not a lot of money can make a big difference.

The Profitable Musician | Bobby Owsinski | Artificial Intelligence

Artificial Intelligence: Isolation is expensive. There is no cheap and easy way to achieve it. On the other hand, you can make a big difference with acoustic treatment without spending a lot of money.

 

That’s something so helpful for people like me because this is the room I have. I don’t have a fancy studio.

Again, I wouldn’t have said this a few years ago, but you can do good work on headphones these days because there are these software programs that simulate you being in a real room which is fantastic. It may cost you $200 or $300 to do.

Do you need to have a high-quality set of headphones to make that useful?

It helps, but it’s not required. High quality means a $1,000 set of headphones. No. If you have some reasonably good $100 headphones, it should work.

Utilizing AI In Music

Thank you for those tips. I want to jump into AI because I know it’s such a hot topic right now. I know that you wrote a book on it. What made you get interested in AI and how it relates to music?

The reason why I write all of my books is the fact that if there’s something I don’t know, I want to learn about. Also, maybe selfishly or egotistically so, if I want to know this, a lot of other people will too. It’s turned out to be the case but like I say, sometimes I think, “Maybe that isn’t,” but in this case, it is certainly because there’s so much going on with AI and it can be very confusing. It’s a little less so in the last few months, but it can be daunting. It could be very confusing what you think AI can do for you and how it can do it. I wanted to write a book that covers everything and it covers everything from the basics, “What all of the buzzwords mean? Where do they apply and where they don’t?”

Also, get into things like AI copyright, which is a field that slowly but surely is working itself out but there are a lot of gray areas. Also, get into things like, “Let’s talk about AI for music generation, for creating music. Let’s talk about AI for helping us with compositions and lyrics. How about AI for the audio portion?” It’s because there are some good AI plugins that work well. How about AI for mixing and mastering? Finally, how about AI for marketing? When it comes to marketing, it includes AI videos. AI creates videos for you. AI does things like release schedules and writing bios for you. AI that does branding or those things.

Album art and all that stuff.

One of my students sent me a copy of his CD, and on the cover, it was all done by Midjourney[Ma3]  so it can be done and it was nice too.

Three AI Myths

What I love is that you created a handbook for musicians that covers all the ways that they can use AI and there are so many websites, summits, and things like that all about AI. It does cover a lot of that marketing stuff, but they don’t cover all the music-related things like you said, the production and the copyright. That’s a crazy one. We’ll get into that. I want to start with what are the myths that people believe about AI or the things that confuse people about AI that your book helps to dispel.

Before I did the book, I sent a survey out to my list, which is quite large, and I asked them about this. What scares you? You think of various questions. Two things came back. The first thing was, “I think AI is going to sap my creativity,” which is not the case at all. AI is good at big things, but it’s not good at nuances. Humans are good at nuances. It’s not going to take your creativity but there are certain things they can do very well.

The second thing was AI is moving so fast that whatever I learned now about it is going to be obsolete tomorrow, which is not the case either. Sure, there are some things that are going to change and improve, especially when it comes to AI programs, but that’s not necessarily the same thing as learning about how it works because the basis for AI goes back to 1954. Many of those things, the terms, and the concepts have been around for a long time. This is not new stuff. If you learn those basics, it’s something that will still be worthwhile going forward in the future. Those are two of the big ones right there.

The basis for AI goes back to 1954. Learning its basics is worthwhile going forward into the future. Click To Tweet

Here’s a third one. People are under the impression that if they go to one of these AI music generators, it’s text to music. They put in “Make me a Michael Jackson song.” It’s automatically going to come out like a Michael Jackson song because they hear these clones. They don’t understand how those clones are made. Most of the good clones of Sheryl Crow, Michael Jackson, Drake, and Taylor Swift, most of those are done manually.

What they’ll do is there’s someone that’s programming it, not the AI. They’re programming it the way we normally do on the computer using loops and samples. Maybe they’ll take a vocal sample of Ariana Grande and drop it in. It sounds like her song, but it’s not. Only a small portion of it is generated by AI. We’ll get the impression that if I put in “Make Me a song that sounds like Drake,” they’re going to instantly get it. It doesn’t work that way.

I think people think that’s going to make so many musicians obsolete but as you said, they still have to build the whole thing and then maybe they put something in there that’s like, “What can I tweak about this to make it sound more like Ariana Grande,” and that’s where the AI helps you, but you still have to build it.

I’m not saying that you can’t build it within. There are very good AI music generators, but it takes time. It’s not something you put the text in and it spits it out. You have to tweak it and tweak it. It’ll take hours to get what you want. Now, I’ve gotten to the point where every time I’ve tried to do that for a music cue or something for my podcast, for instance, I get frustrated. I go, “I can write this faster than it’s taking me. I can do this and be done.”

That makes sense. Maybe people who aren’t musicians think that they can use it because they don’t have the musical creativity to do it themselves. Maybe that will help them. Maybe they can come up with something that’s similar but I think if you are a musician and you have those skills, you’re right. It’d probably be easier to do it yourself.

I think that’s what a lot of these are aimed at. They’re aimed at consumers who don’t have the experience that musicians do. Musicians grow up playing and practicing all the time, and they put their 10,000 hours in. Most of us have, and more so it’s apples and oranges.

AI tools are aimed at consumers who don’t have the experience that musicians do, especially when they grew up playing and practicing all the time. Click To Tweet

AI And Copyright

Let’s delve into copyright because that does get messy. In my experience, copyright is the melody and the lyrics of a song but then when we get into AI, are there some other pieces that are involved in copyright?

Before I wrote the book, I consulted with four of the very significant IP attorneys who dealt with and worked in this all the time and one proofed the whole chapter. I knew it was right after he did this and I didn’t want to get sued. That was the other thing. Here’s the thing. There’s a lot that’s in the dark about this, but there’s one thing that has been defined over and over. The copyright office has said many times now that you have to be human in order to get a copyright. AI cannot obtain a copyright.

That’s a big thing. Now, if you as a human use AI to create something, where does that stop? Where does AI stop in a human begin? That has not been worked out yet, but if it’s 100% generated by AI, it cannot be copyrighted. That’s the first thing. The second thing is copyright is so misunderstood that even some of these AI platforms have it wrong. An example would be, they’ll say to you, “You can do this, generate this AI music, and it’s free,” but you can’t use it anywhere.

You can’t post it on social. You can’t post it wherever or on YouTube but if you pay $10 a month, then you can post it but we own the copyright. If you pay $20 a month, you can post it and you own the copyright. Now, the question comes in. Do they have the right to own the copyright? I’m led to believe, no, that’s not the case because it’s AI-generated. How can they own it? As I said, even on some of these platforms, there’s confusion, and we’re going to see multiple lawsuits on this soon. We’re already seeing a lot of lawsuits having to do with AI. The biggest one was the Sarah Silverman one. Are you familiar with that?

No.

Sarah Silverman and five other writers sued OpenAI, which was the parent of ChatGPT saying that it used our books in order to teach ChatGPT what to do. They’ve lost in court on most counts so far. The reason why is to think of yourself. When you write a song, is it 100% yours from the standpoint or is it creativity? No. You’ve borrowed it from other people. You’ve borrowed bits and pieces. “I like the turnaround. I’m going to use this.” “These chord changes are very cool. I’m going to use this.” “This phrase is cool. I’m going to change one word.”

That’s exactly what virtually all of the AI large language models do. They’re good at pulling stuff in, but they don’t regurgitate most of them anyway. They’re working well. They don’t regurgitate the same material. If they did, that’s copyright infringement but if it’s brand new stuff, it’s not. As a result, there are some lawsuits that are going in favor of the AI platforms, and that’s one of them.

It feels to me like it’s almost like what we would do as creatives. It does it a million times faster because it’s doing what we would do. As you said, we’ve got influences. We hear things. Sometimes it’s totally subconscious that we borrowed something and that’s what’s happening there. It feels like it’s stealing because it’s doing it so darn fast that a human could never do that.

If it spits out something exactly like something that’s copyrighted, that’s infringement but it doesn’t do that most of the time. There have been cases where that’s happened, but with most technology, there are glitches that happen.

I know that traditionally, by the books, or whatever, copyright for a song is the melody and the lyrics but then there’s all these other things around it that make it a song like the chords and all that. For example, if you ask the AI to make something that sounds like a certain band. Let’s say Tom Petty and the Heartbreakers because I heard one of their songs the other day.

I’m like, “They have a sound.” From the first note, you know it’s them. You feel like, “It is stealing from them. If you ask it to make a song that sounds like Tom Petty and the Heartbreakers because that’s their thing. That’s their brand. That’s their style and we’re stealing that but there’s nothing that you can pinpoint that is a copyright.

Yes, but I don’t know if you’re familiar with the Blurred Lines copyright suit from a few years ago.

I heard about it, but I didn’t hear exactly what happened.

This is exactly along those lines. Blurred Lines was the Robin Thicke song. It was a hit that came out, and it sounded a whole lot like the Marvin Gaye song.

They didn’t take that sample. They were imitating it.

Yes, they’re imitating the sound. The jury sided with Marvin Gaye’s Estate saying, “Yeah, you copied this sound.” This made IP attorneys everywhere crazy because they said, “This shouldn’t happen. You’re not supposed to be able to copyright that.” The copyright law is saying you don’t yet the jury chose to believe them. That’s one of the problems that you have when you go to the jury trial. You never know which way they’re going to go.

So many rap artists used to sample and now they try to reproduce maybe a sample that they would’ve wanted to use that they can’t because that would cause a lawsuit.

One of the reasons why you see up to twenty different writers on songs now, especially hip hop songs is to cover your ass type of thing where if someone is in the control room while they’re making the song or in the room while the song’s being created, they give them credit. The reason why is, “You can’t come back and sue us.” The big problem and this is a problem if you’re doing sync. There are so many writers with different publishing companies that those songs are impossible to sync.

What about us as indie artists? We put our music on Spotify, on YouTube, and all of that. How do we protect ourselves from being infringed upon copyright-wise by AI?

Again, understand that I’m not an attorney. I’m not giving legal advice to anybody. I can’t do that. I can give you my opinion, just to be straight on that. It’s difficult to do that but I wouldn’t worry about AI. I’d worry about some bad actors because already we’ve been seeing people that are absconding with a title and saying, “This is my title,” and putting it up. What they’ll do is they’ll upload it maybe a little faster or, a little slower with a different title but your music.

We’ve seen a lot of that happening on Spotify and YouTube. They do it in such a way that if they change enough, the copyright protection won’t find it. They turn around and they’ll put a take-down notice against you saying, “This is my song. It’s not your song.” That’s the worst. That’s not AI doing that. That’s some people in Russia that are trying to make some money.

AI’s Creativity And Limitations

That’s so frustrating. I’m a musician and I want to utilize AI to help me write to be more creative. How can I do that and what are the limitations that AI can’t help me with?

I think AI is excellent at lyric generation. Now, not that it’s going to write the whole song for you or even a big part of it, but it’s excellent for ideas. You might get a phrase out of it that you never would’ve thought of otherwise. A phrase that I’m talking about. If you are absolutely hopeless at lyrics, then it’s going to come up with something better than you ever could. You rarely see Moon June Spoon and stuff like that. It’s reasonable in many cases.

I think that lyric generation and lyric ideas are good. There are some very good AI programs that are good at coming up with different chord changes. You give it the chords that you’re working on and it will suggest other chord changes, and that’s excellent. Scaler 2, which is a very deep program, but on a very surface basis is very good for that.

For example, if you don’t know much about music theory, you learn to play guitar yourself and you’re basically writing every song in 1-4-5-1 [Ma4] kind of thing, it’ll give you more chords to substitute.

Yes, way more and good ones too. Whenever I played with it, I go, “I should have thought of that. That’s fantastic.” Billy Joel often comes up with changes like this new song that he put out. I listened to some of the changes that he made and the turnaround. I was like, “This is great. This is very unusual.” Elton John does that too.

I’d have to mess around with that. We get stuck in the same ways.

You can go to any of the AI generators and see if there’s something that pops out. That’s a good idea. That’s something you wouldn’t have thought of. Give it some parameters and see if there’s something that you’ll go, “That’s cool. I like that.” Now, here’s another limitation of AI that I didn’t go over. It’s a big one. This is coming from a technical standpoint. In the studio for the last many years, the resolution that we record at is 96K/24-bit[Ma5]. This has been standard. If you’re doing something for a record label, that’s what their delivery specs are going to say. We want the masters at this particular frequency and bit depth.

This has been a standard for a long time. AI cannot generate at that resolution. The best it can do is CD resolution 44.1/16.[Ma6]  That’s the absolute best it can do. There are a few that can do 48K, but none can do anything higher than that. Not yet. The reason why is it takes a huge amount of horsepower. Think of it like this. When you go to an AI in say Midjourney for instance, and say, “I want a picture of Bree and Bobby speaking on a podcast.” It will come up with something, but it does it once. That’s it.

If I ask the new Sonos that came out, “Do a Bree and Bobby a video of us on the podcast.” It’ll come up with something, but it’s doing it at 30 frames a second, 30 times a second. With music, you’re asking it to do 44.1000 times a second. It requires a lot of horsepower to do that, and that’s why we’re not seeing higher resolutions but it’s big when it comes to, “Can I use this professionally?”

Is that because it’s doing it over the internet? In the old days, we had software on your computer and then you do something online.” If you had a system that was on your computer and was using tons of hardware and all that, could you do it?

Possibly. We haven’t gotten to that point yet or maybe you don’t. You have to be looking at the tech side of things but NVIDIA, which makes GPUs or Graphics Processing Units, speeds up graphics, now they’re making custom AI chips to speed up AI and this could help things. We’ll probably see add-ons to computers pretty soon that will be specifically for this. As soon as we go to, “Let’s take the AI off of Worldwide Web and let’s put it onto our local computer,” that’s coming.

Recommended AI Apps For Music

What are some apps that musicians should look into for music-related stuff and even marketing-related stuff? Did you put those in your book? Because I’m sure they’re changing constantly? I was wondering how you keep up with that.

I looked at over 200 different AIs, which was very daunting, I have to say. A lot of it runs together, but there are a couple of good ones that do stick out. For music generation, it’s worth checking out AIVA.ai. That makes some nice music. It does orchestral stuff, and you’ll go, “This sounds like the real thing.” Again, that’s fun to play with. On the audio side, there are lots of good ones made by only a few companies. You have Focusrite.

There are only a handful of companies that do it. iZotope is one that makes a lot of AI stuff. They’ve been doing it for years and they’ve never used that as a marketing. Even now, they don’t do that but it’s very heavy into AI. What we find is you can have automatic EQ, for instance. It will figure out the EQ or the compression for you. It will figure out the limits. It will figure out which tracks are clashing with one another and EQ both of them or all of them so they don’t do that anymore. That’s AI at work, and it does a good job of that.

I wish I had that back when I first started trying to produce my own music. It would’ve made life so much easier.

It’s very helpful. There are limitations to that as well. You have to tweak it.

You still have to listen. It can give you suggestions, but you still have to listen and go, “That’s not quite right.”

There’s no AI, for instance, for mixing because usually it’ll mix 4 or 5 tracks or 4 or 5 stems. That’s all it’s capable of doing. For instance, for song lyrics, there’s something that I like. It’s a song lyrics generator. It’s very simple and to the point. That’s Rytr. It has many uses but one of the things that it does have is a lyrics generator that’s particularly good. ChatGPT is still good. It will give you what you want, but you have to be very specific. This is another limitation of AI. Some people cannot express themselves what they want well.

The Profitable Musician | Bobby Owsinski | Artificial Intelligence

Artificial Intelligence: There is no AI for mixing because existing tools for it are only capable of handling up to four or five tracks or stems.

 

“I thought I told it to do this,” but you weren’t specific enough.

What happens is that AI wants you to be specific. That’s a limitation. I think that’s going to be worked out soon. It’s a UI thing more than anything. One that will blow your mind is Suno.ai a music and production generator. That’s good. There’s a new type of digital audio workstation called a GAW. It’s a Generative Audio Workstation. There are a few of those that have come out. polyMorph is one. Audio Design Desk is another one. The third one is from RipX. They came out with a brand new one, that even works on Apple Vision glasses. It’s the first one that will do that.

For lyrics, LogicBalls is a good one. For EQ, compression, and stuff like that, the smart comps that we have are called Sonible. It was one of the first into AI so much so that their engine is used by other companies but they make many different ones. They make very simple EQ, compression, limiting, and inexpensive. All of these are inexpensive, by the way, and they have a higher-end one that you should be a techie to use. There’s that and then Focusrite is another big one.

For noise reduction, there are tons of them out there that are good. One I use all the time is Accentize called dxRevive. It’s fantastic. You can have the crappiest dialogue you can think of. Somebody in a cave talking on a cell phone from three feet away and it will clean it up and make it sound like it’s in a real studio. I don’t know how it does it, but it’s fantastic. There are things like that.

Also, mastering. AI mastering is good. This is going to hurt mastering engineers because it’s good. Is it as good as a mastering engineer? No, but it’s close. I’ve done ABs [Ma7] where I’ve taken a master’s from a mastering engineer and I’ve tried to duplicate it. It was close enough that you’d go, “I can use this.” There are tons of these now.

Is LANDR using AI now?

Yeah, and they’ve been doing this longer than anybody. One of the reasons why they get better is because of the training. I think they said there are over five million tracks that they’ve done. That’s a lot of training.

They were smart to get into the business early, even though they weren’t using AI at first. I’ve heard some LANDR ones from back when and they are not very good. Now, it’s been training and better and better. They have all that data.

eMastered is another one. CloudBounce is another one. We have the ones that are from various manufacturers. Slate has their own online mastering. That’s pretty good. Waves have online mastering. Plugin Alliance now has its own online mastering. All of these are good. The last three that I mentioned, they’re done in such a way that they’re close to a mastering engineer. It’s worth the money.

Let’s talk about video. One of my favorites is Rotor AI, and that’s a good one. Maybe what’s better is Kaiber[Ma8]. It is very good. One of the things I like about Kaiber is right up front, it tells you, “You own all the copyrights. We don’t own anything.” You can’t always be sure about some of these that I’m telling you. You have to go look at the terms of service to see what they say. Not so much the audio ones. For the audio ones, the copyright is always yours. It’s not a problem but for the marketing ones, for instance, you have to worry about the video generators, the graphics generators, all those, and the branding ones. Adobe now has Firefly. That’s worth checking out. That’s fun for graphics.

We hear a lot about Midjourney. What do you think about that?

I think it’s clunky and the reason why is you have to use a Discord group to do it.

I’ve heard that. That’s so weird.

I don’t like Discord, first of all. That’s the first thing, but second of all, it’s clunky to do what you have to do. There are better ones. Leonardo.ai is good. Now, the open AI is DALL-E, DALL-E 2, or DALL-E 3 and that’s getting good. That’s the same people as ChatGPT. It used to be you had to go to a DALL·E site, now you could access it from ChatGPT, and that’s the cool thing. It saves you a little bit of work. Even Getty Images now has its AI generator.

Cleanup is an image generator. Both of those are separate. They’re cool for doing what you think. For instance, with Cleanup, you could easily take the background out of the picture if you want or change it to a different background or upscale it so you have this crappy low-resolution video or photo and it will upscale it so it looks like it’s high res somewhat.

How good is it? Could I say, “Take the headphones off of Bobby?” Could it do that?

Yeah, no problem. As a matter of fact, I have an AI course as well. It’s AI for Music Production. I used this picture of me and Ken Scott together in Nashville. Ken Scott is a famous legendary engineer. I went in and I erased Ken. I had some glasses hanging out of my pocket. I erased that. There was a funny thing in the background that was coming out of my head. Not only does it erase it, it replaces it so it’s seamless.

If you’ve taken a picture and you’re like, “I look good in this picture but the background is so weird.” You could fix that.

There are lots of them that will do that now. Image generator’s a cool one for that, but there are tons of good ones.

Do you use anything to help you with your marketing? You use it to help you write emails or even courses. I see stuff out now like that.

Using AI For Marketing

Yes and no. First of all, let me say one more. Nightcafe is great. The reason why Nightcafe for graphics is good is because it does access all of these different generators. Also, DALL-E. Not Midjourney, but some of the other big ones, for instance. From one space, you can access a lot of different engines and that’s cool. Nightcafe is a fun one. What do I use? When I wrote my AI book, I used ChatGPT as an editor. I didn’t use it so much to generate it. I did on occasion, and it would mostly be something like this. “Can you say this clearer than I did?” It would add some clarity, but I go back and forth between that and Google’s Bard, which is now called Gemini, and then even Bing AI.

ChatGPT is good as an editor. It can add some clarity to your work. Click To Tweet

I would go back and forth between them all to see what would happen but acting as an editor, it did a good job. I find whenever I try to write something with it, once again, we’re back to, “I’m a good writer. I could write it faster than you write it for me, and I have to edit it,” which is what the big problem is. I’m continually asking it to do a blog post for me and am continually frustrated. I was like, “I have to rewrite this whole thing. This doesn’t sound like me.”

That’s the issue. The question is, “Can we train it to sound like us eventually,” because that’s what I get. Even captions for social media, I’m like, “I would never say that, or, “This is so hypey,” or something.

I’m working on something right now on that and see if we can train it in my voice. Now, that being said, for podcasts, I’m going to do an experiment and do a podcast completely with an AI clone of my voice. I know it can be done, and I’ve played with it already. It doesn’t get some of the technical inflections or technical points that I’m trying to train it on, but once I do that, that’s an experiment. I’m going to do it.

With that being said, what drove me down the route was I had a listener who did it for me and sent me the audio of me talking about something I never talked about before which was scary. I said, “Please don’t let this out in the wild.” He said, “No, I just want to show you.” I said, “Tell me what you did and how you trained it,” and he did. It doesn’t take much. You could train a voice clone with as little as 30 seconds worth of material.

AI Voice Cloning

Now, it doesn’t get good with 30 seconds, but it will sound like you. It takes at least 30 minutes to an hour of very clean audio so it can get your inflections but that’s not a big deal from podcasts especially. You probably have plenty of it and you pull that out and you upload it. ElevenLabs is a good one for that, and it will do it. There’s also a lot of very good AI voice cloning, and there are a lot of pro-songwriters that are now using it.

The reason why is if you want to get the song to Taylor Swift or somebody like that, and you’re male, you don’t want to sing a song in your voice. Many times, you can’t get a female voice to sound the same. There may be a clone that sounds closer, and if whoever it is you’re trying to sell the song to, if the song already sounds like them, they’re going be more inclined to buy it. This is used more than you might think professionally.

It’s going to put demo singers out of business. I hope not totally. I love doing demos. I still do them on occasion.

It’s going to make it harder in some ways. Again, there’s a technical part to this that many people can’t jump over. It’s the limitation.

On the other hand, there are so many great songwriters out there that their barrier is they can’t afford to spend $1,000 for every song on a great demo that’s going to show off their song. If this could help them overcome that, we could have more great songs in the world.

We can’t have too many of those.

Using AI For Podcasting

This has been so good. I did want to ask you though, podcast-wise, I’ve heard the Riverside podcast. It is great as far as taking out any background noise on the fly and all that. I wondered if Zoom has built that in because we’re on Zoom and I did a podcast earlier this morning on Zoom, and he’s like, “The fire alarm is so loud,” and I could not hear it at all. I was like, “That’s crazy. I wonder if Zoom is taking that out.”

Let me ask you a question. While we’ve been doing this, my housekeeper is here. There’s a washing machine. There are all sorts of noises.

I’ve heard nothing.

Then it’s working.

I bet you they’ve built it into Zoom now.

With that said, I do use dxRevive all the time. It’s the cheapest. It’s $30 or something so it’s not a big deal, but it’s also on these things that you put it on the track and it does it. You don’t have to do anything but it’s mostly because I get people that are talking to me in their kitchen and going all over the place. The audio is awful. It’s usually the higher up in the audio business they are, the worse their audio is. What I have to do is clean it up and this does it pretty well.

Closing Words

That’s true especially if they’re not normal podcasters. They don’t have a little separate room and all that stuff and all the right equipment so that makes sense. If you want to get to the experts, sometimes you have to do your podcast on their phone. I’ve had that before where it’s like, “This person’s not set up, but I want to hear what they have to say.” You shared so much great stuff, and thank you for being so specific in giving all those tools and everything. I hope everybody who’s reading will go out and get your book on Amazon. Tell them how they will find your books. How do they find you online?

The Profitable Musician | Bobby Owsinski | Artificial Intelligence

The Musician’s AI Handbook: Enhance And Promote Your Music With Artificial Intelligence

It’s easy to find me in my books. It’s BobbyOwsinski.com. If you go there, it points everywhere. It points to my blogs, podcasts, and books. You can read excerpts there if you want. All the books are on Amazon or Apple Books or any place that you want to get a book both in electronic version or in print version. This particular book is The Musician’s AI Handbook, and it covers all of those things that we talked about. It’s a little bit for everybody.

Again, you mentioned that everything has changed so much and some of these platforms have changed and upgraded or whatever. What I’m planning on doing, and there’s one coming up soon, is doing a webinar that goes over all of this stuff. This is for anybody who buys it, and I’ll also send a PDF out if I have your information and give you an update on the book. That’s easier than sometimes doing the whole book itself. It’s faster.

Go to BobbyOwsinski.com, get on his email list, and he can keep you updated on all the coolest AI because that is not my forte. That’s why I bring people like Bobby onto the show so I can learn and so I can get this information out to everybody. Thank you so much, Bobby. This has been so great. It’s way overdue for sure because I know I was on your show a few years ago. It never happened for whatever reason so I’m glad I finally did that. I knew this would be the perfect subject for us to talk about so thank you so much for sharing all your expertise.

Thank you for having me, Bree. It’s always a pleasure to talk to you.

 

Important Links

 

About Bobby Owsinski

The Profitable Musician | Bobby Owsinski | Artificial IntelligenceProducer/engineer Bobby Owsinski is one of the best selling authors in the music industry with 24 books that are now staples in audio recording, music, and music business programs in colleges around the world, including The Mixing Engineer’s Handbook, The Music Producer’s Handbook, and The Musician’s Ai Handbook.

He’s also a contributor to Forbes as a category expert on the new music business, his long-running production and music industry blogs have won numerous industry awards, and he’s appeared on CNN and ABC News as a music branding and audio expert. Bobby’s highly-rated Inner Circle podcast is now in its 10th year, with more than 500 episodes that feature mover and shaker guests from all parts of the music industry.

Release Concierge Release Planning & Budgeting Checklist

RELEASE YOUR NEXT MUSIC PROJECT WITH CONFIDENCE

Struggling to keep track of all the steps you need to take in order to plan your next release? Release your next music project with confidence with our Release Planning & Budgeting Checklist.

FREE GUIDE: 39 LITTLE-KNOWN INCOME STREAMS

These are NOT the obvious, industry-backed income streams like streaming royalties, album sales, publishing royalties, and concert ticket sales. These are out-of-the-box solutions that will transform a Starving Artist into a Confident Creator who has ENDLESS possibilities from monetizing their music.