Automated (Artificial Intelligence) Music Compostion / Production Arrives - Problem?

While I’m not going to post any links or cite any names, about a week or so ago, a tech company announced the release of a new platform that uses artificial intelligence to produce original music for the corporate and hobbyist video market. Songs can be created “on demand” that match the video length and you can choose from a handful of genres and moods. Obviously, the music is computer generated, but it does use samples for live instruments, including guitars that appear to be based on MIDI grooves. The music is royalty free, said to never duplicate itself, and is very cheap. I tested this app and it does what it says. Songs are generated very quickly and if you don’t like the result, you start over until you get what you like. The company has a team of musicians that handle the music theory and (I suppose MIDI) playing. Programmers created the application using artificial intelligence based on said music theory that pieces together what are most likely pre-sequenced parts. We’re talking about almost instant custom music that is fit to the video length and can be drag dropped into a video time line from a desktop after download.

OK, now… the songs sound like crap to me, and I doubt I needed to tell you that. But, here’s the conversation I want to start:

Is this a new direction we’re about to see that’s going to further complicate this highly over-saturated market with falling prices and 100 new “composers” every 30 seconds? As we all know, the “first” anything generally isn’t that advanced. Over time, this could get much better. They’ve indicated that they plan to expand their genres, moods, production techniques, etc. Of course, this won’t be anything more than a lower tier royalty free music service for a quite some time, in my opinion. BUT, isn’t that where you guys are right now? Could this be a serious problem within the next five years if this catches on and the music improves? Sure, musicians will always value the human touch and we can be extremely picky, but how important is that to music buyers looking for cheap music for a corporate video?

The company claims to be targeting Google / YouTube as a future partner along with other companies that would use this type of technology within their applications, for example: Vimeo, Animoto, Daily Motion, and the like.

I doubt we’re going out of business anytime soon and the higher level markets aren’t going to be threatened by this. But, I’m interested in any thoughts out there. Please look at this with the assumption that the technology will improve and consider the benefit this could offer music users in this market versus the issues it could cause for us musicians and the smaller RF and PRF marketplaces and music libraries. This is one company, but there are others working on this as well. Hopefully, we can go further in this conversation than things like, “You’re crazy. I heard it and it’s garbage.”

Is this a threat and is it another sign that the production music market at this level is falling further into chaos? Or, is this no worse than just having yet another small music library in the sea of mediocrity that already exists?

I think it will be a game changer. Not necessarily this company, but some other at some point. AIs will definitely come and stay on every possible level of production and interaction. If AIs can drive cars, they can most definitely compose inspiring corporate music :smiley: However, I don’t believe this will happen in the next 5 years. In 10, maybe, in 20 most definitely.

I think it will lead to human composers doing more high end custom stuff that is not as easily replicated by generated MIDI sequences, but it will probably cause many of the composers currently gaining reasonable income from “basic stuff” to run out of business.

But in general level, I don’t think anyone truly understands how AIs will change the world. It might very well be that in twenty years time there is not a lot of work to be done by humans anymore and it’s at the same time terrifying and amazing.

2 Likes

Solution is simple, find John Connor and help him destroy the Skynet. On the other hand, it could be potentially a serious competition, because every other track on this site sounds pretty much… well “inspirational”. Not to mention that they can’t fix the search engine. So, the so called plan B if we fail to destroy “Skynet”, we should simply join it, and be the people who create midi for it :D.

2 Likes

Well i doubt its truly AI, for me it just mix’n’match scripts, it can’t analyse a track and make decisions…yet.

This issue has been addressed a few years ago i’m forgot where is it now, but here we are still selling and survive once again.

I don’t believe in the Als generative hype as humans consumers of the final product we are supposed to be. Many years ago when a “XXXsoft” was released there was a concern here about how this could affects the music production scene. At the end XXXsoft is great but just as a tool to help musicians, music teachers, students to get a sketch of the track but don’t expect it to create something cool and impresive for a video presentation.

AI is real and it’s coming. I think music is always going to be intimately connected with current culture and trends. AI compositions may very well be efficient in many cases but as long as 60% of the audience (rough estimate) will be on the lookout for hip new alternative music, reprogramming the AI to focus on “that latest sound” will probably take longer than it takes for a professional musician. At least as long as we don’t reach “singularity” and the AI figures out how to adapt and improve itself faster than human kind can keep up with :wink:

When hip tips over into mainstream, and ultimately nostalgia, new things must be invented to capture marketable attention. When AI can do that too, what’s left for the buyer (assumed human) to decide is ultimately if human musicians are worth supporting - simply for it’s own sake. Most of popular music is currently sold by integrating personality, image and hype into the game. The music itself is already more or less just a raw material we sculpture into culturally relevant and useful forms.

Will AI music intensify competition? Yes, especially for low-engagement background stuff.

Will the human race step up to the challenge? I hope so :sunglasses:

1 Like

Humans will always beat the artificial intelligence in my opinion when it comes to some genres like orchestral or music for picture based on motion, feelings, etc.

I cannot imagine a robot composing a soundtrack for a movie then seeing at the end credits “Music by Automated Music Creation Software”.

Certain genres of music are in “danger” because can be easily created by robots (EDM like house, techno, trance, dubstep, hip hop, etc) but other genres like symphonic orchestra or various acoustic genres can be difficult to reproduce by artificial intelligence.

Robots at the end are programmed by humans.

1 Like

Time will tell, but if a “real sentient AI” will be created, I think it can handle classical emotional music as well. After all, there’s no magic behind it, it is still quite reasonable amount of notes being played on a few instruments with some details added (dynamic volume changes and expression on violin). Only thing to do is to make the AI learn by “listening” to what emotional music is all about and after enough of iterations it should be able to produce reasonable compositions that match the input.

I think musically naive (both in composition and quality) users of AI software who wish to perform media production will struggle to get what they want in a timely manner. They can iterate easily for new tracks but fine tuning will be a challenge (think of how much fine tuning and improvement work we do that relies on experience). So it probably will be a struggle to get exactly what they want in a timely manner.

What I think might be more of an AI impact to the stock music business is “time-to-market” for the music author community. If todays music authors start using AI tools tomorrow to get a ballpark track quickly, and then fine tune it to their liking and upload, production rates will sore even more than they are today. Much like in-the-box tools, samples, and midi have enabled us to generate tracks more quickly than full recording, out-sourced mixing, and mastering, AI might give authors a very fast starting point for a specified genre and length.

Thoughts?

We’ve been able to work in a similar manner without AI and without “composer tools” for a long time and at nearly the same speed.

Let’s face it, every possible chord progression and rhythm considered pleasing to listen to has already been used an almost infinite number of times. What makes things “original” is how you combine those elements. That is how AI most likely operates and so do we…

A very well versed musician has a built in catalog of popular chord progressions, rhythms, production styles and riffs they can mix and match at any time. It comes from many years of playing music and can easily be expanded by just listening to and playing new music. But even those less experienced can get over “writer’s block” by taking bits and pieces of different existing songs from an endless supply of sources, altering each selected component slightly, then mixing and matching them. Often, this is done without even realizing it. Doing it consciously expedites the writing process dramatically. A composer just has to admit to himself that at this point in history, none of the basic components of a song will be original (and still be popular). Even production styles are widely copied to the point where one sometimes ends up becoming a new genre. Every style and genre can be approached this way, especially genres with instrumentation and production elements that normally don’t vary all that much.

The above being said, I personally don’t think AI will help a composer do things he can already do. The biggest benefit will be for people who cannot create music due to a lack of skills and/or tools. It’s something that is more likely to replace us rather than assist us, especially if technology advances to a point where AI can instantly do what I explained above. If it reaches a point where a customer can drag and drop multiple songs they like to an interface and ask the application to compose based on those songs, it’s going to out pace us very quickly. Already, there are production music libraries that have drag and drop search engines that will find songs in their library that are similar to what a client wants. It’s just a matter of time.

I thought that AI composition was already in use with creating tracks of Inspiration - it just seems that the same button keeps getting activated.

1 Like

Too funny Graham!:smile:

Actually, I think, after listening to AA generated tracks our AJ Inspiration tracks will sound more primitive )))

Well, I’d say a lot of music these days is sort of manufactured anyway, using the same progressions, simple melodies, and production, with a lot of “composers” imitating each other and what’s popular, in order to make a buck. In this kind of environment, I think AI can succeed. Consider how commodified music has become over the past 100 years or so. Music used to be a somewhat personal expression where sometimes you were lucky enough to find a patron and make a living, and sometimes not. Wagner lived a life of staggering debt, but at least he was able to get his musical ideas heard. With advancements in technology, music has become a commodity, not much different from an orange or a banana, and whenever technology industrializes something, it tends to replace human labor except for where it absolutely can’t. Once something becomes commodified in this way, often the only way to make money from it is to focus on quantity over quality. That being said, a computer algorithm will probably never be able to score things in a precise or unique way, and neither will it be able to come up with something new that is of any real value. Consider for example, the problem of melody; take five notes, vary their duration, the intervals between them, their melodic content contrasted with the harmonic content of the piece, how that harmonic content resolves, ect. These are things that a human brain needs to develop or have a knack for, because you can’t program creativity. You can program randomness, and even ordered randomness, but not worthy creative endeavor. I strongly disagree with the notion that every good sounding musical idea/progression/ect. has already been tried, it’s just that people have developed strong preferences that generally put them off from hearing different or new ideas - or even old ideas presented in new and interesting ways - and the new ideas that do tend to leak-through are somewhat gimmicky bullcrap - dubstep I’m looking at you - but everyone will parrot them because it becomes popular and everyone wants to make money. This is further reinforced by markets like AudioJungle who explicitly control the content of their market, as opposed to say, operating a completely free market, further emphasizing certain preferences and trends over others. If someone truly feels that there is nothing original to do with music, they have likely become disillusioned, and should take a break from stock-music and popular music for awhile to write some music for themselves, just for their own creative satisfaction, and to get back in touch with why they probably became a musician or composer in the first place. If you are already making the kind of music that 1,000,000 other people are making, then you are already competing with 1,000,000 other entities for a share of the market, so what difference does it make if you now have to compete with technology? Either way, as long as the human population keeps rising, your market is going to become increasingly saturated over time. The choice that people who work in creative mediums are going to have to start making is, “Do I put myself in a cage with all the other mice, and hope that I’m somehow able to beat them all to the water bottle, or do I sit outside of the cage and risk everything, in the hopes that I can maybe pull a few of the mice out with me?” For most of human history, being an artist or musician was unprofitable. When recording technology came a long, there was about a 50 year period where being a musician suddenly become stunningly profitable, but now that advancements in digital recording technology have caused the price of music production to plummet, and the internet has made the cost of information cheap, we are slowly moving back towards music being unprofitable again. The only reason it hasn’t completely collapsed yet is because that same technology that has made it easy for people to create music has simultaneously made it easier for people to create and distribute video content (and similar things) that require music. However, over-saturation is going to happen, it’s going to keep happening, with or without new technology, and how someone chooses to deals with that problem is entirely up to them.

1 Like

Artificial intelligence is not the same as just saying something was generated in a random manner by a computer. AI has the ability to take everything a human knows about music theory, combine it with knowledge of the history of popular music trends over generations, and analyze what sounds, phrases, intervals and playing styles will convey which emotions. We do the same thing, only much slower and how well we do it varies between individuals based on skill level. AI can ingest information fed by multiple sources relating to music as it evolves in real-time and do it 24/7 with no distractions. We can’t. The goal of AI is for it to continue learning with no need to continue programming it. At some point, it could be more “creative” than humans because it has the capacity to store and utilize more information relating to music quicker than we can, thereby having the ability to generate new ideas at will. Plus, it will still be able to create based on trends of the present or the past.

Here’s the thing. Creativity means new ideas, but that doesn’t mean they have to be weird and unpleasant new ideas. You can waste time trying to reinvent music, or note that most musicians considered to be extremely creative still use standard chord progressions and write within genres. Creativity comes in the way theses elements are combined. There are numerous standard progressions, rhythms, playing styles and genres from which to choose. The more familiar you are with them, the less limited you are. No matter how “different” you try to be, the truth is that it means nothing to anyone except you if others do not enjoy your work. Why people have certain preferences doesn’t matter. They have them and a working musician’s job is to serve those preferences. This being the case, even AI will most likely continue to work within those parameters found to be most profitable. However, it will have the ability to create new ideas based on trends, both past and present, much faster than we will. It could also learn the likes and dislikes of an individual user.

You aren’t competing if you’re deemed unnecessary and eliminated. The saturation issue is what is killing production music, and it’s doing it at a rapidly increasing speed. Yes, it is due to technology, which is exactly why I think AI will eventually take over this low budget market and might go higher. Currently, in this market, we have a zillion bedroom composers using DAWs and uploading massive amounts of music to sites a music buyer has to search, totally ignoring the bulk of it. That’s a waste of time and effort, not just for us musicians, but also for the music user. AI eliminates that by talking care of the music theory the music user may not know and the playing / recording / production skills he may not have. Those things are the only reason they need us. If they can leverage these things using an intelligent system that can generate new music all day long using current trends, use top production methods, render full masters and rarely repeat itself from a platform within their workstation, we’re toast. And we probably should be toast at that point. How long before that happens is a function of the speed with which AI technology improves.

Again, this is most likely some years away. But it might be a good idea to consider seeking higher ground if you’re a musician who composes low budget production music. I think most areas of music composition will remain safe for a very long time, but this specific market… I don’t know…

1 Like

Yes, but we aren’t talking about actual AI here, we are talking about simple algorithms masquerading as “AI.” We are likely still many decades away from anything approaching true artificial intelligence - I doubt it will happen in our lifetimes. Now, how close do you have to get to true AI to compose quality music? Depends on the system, and if you are just arranging mixing/matching loops, ect, and a lot of that content has to be generated beforehand - algorithms have to be continuously updated, new ones programmed, instruments/phrases recorded, ect. It still requires a good deal of human effort to get something like this working, and one starts to wonder about the cost/benefit ratios of these things, and if these companies can actually be profitable, especially considering there’s an endless supply of human composers willing to work for close to nothing, and an already saturated market they are attempting to break into.

I didn’t say that they had to be, by new ideas I simply meant ** not the same exact thing everyone else is doing**, but I’d argue most composers that are held in high regard today are notable precisely because they changed or innovated something within their respective style, or were directly or indirectly responsible for creating a new genre or sub-genre. Humans are really good at mimicry and imitation, and everyone apes some other musician or composer to an extent because you obviously can’t start from scratch, but that doesn’t mean we need sooooooo many people who are willing and able to produce soooooo much similar sounding music, and often times not even end up with anything to show for it. If you are able to at least partially balance what’s in demand/popular with the kind of music you’d actually like to be making, then at least personally you can get something out of it.

So what? I know this is AudioJungle and everyone is trying to sell music, but if someone is only making music 100% for others or 100% just to make money, and never for themselves, I’m not sure why they bother at all. Might as well get a job as a dishwasher or grocery store clerk if you are going to spend all your time in the service of other people’s preferences…although I do realize people get into music for all sorts of different reasons, and not necessarily as an outlet for creative expression or to convey meaning or for the joy of it. Personally, I can’t do it man, I’ve tried, and I just can’t do it. My hats off to everyone who can do it - people will always vary a bit when it comes to personal integrity and what they feel comfortable doing for money.

Yes, but as long as there is a market for music you are never going to be able to completely eliminate the human factor. Stock music is already cheap as hell, and not even worth doing financially for a lot of people, and yet they still do it. Why? Because they like doing it. I think we have to also look at this from a buyers perspective - what does a buyer really care about? One would assume quality at the lowest price possible. Is AI really going to be able to offer a significant leap in quality at significantly lower prices than what the market is already sitting at? That’s the question, how cheap would they need to go? $1 a track? $2? It can’t approach zero, because the cost of paying someone to program and develop this AI, maintain a website, turn a profit, run a business, ect needs to be factored in. Music is something that a lot of people are so passionate about it, they are willing to do it for free or next to free, and this is precisely part of the reason why we have an over-saturated market to begin with, and why people are already willing to accept such low pay for their music. How low you can set the pay-scale and still have humans producing music? Probably close to 0, so I’m just not sure how much this kind of technology will ultimately have an effect on market dynamics. being a professional musician/composer isn’t a great position to be in, because most people are attempting to turn a profit off of something they actually enjoy doing and in which there is an overabundance of other people doing the same, so you are are already starting off from a position of poor leverage. Balance that out with the fact that most non-musicians don’t really know much about music and more-or-less just listen to or buy whatever happens to be presented to them, and you have a bit of uphill battle before you either way.

From a monetary perspective it’s a waste of time, but if you are someone who just enjoys making music and the type of music you are making, then it’s never really a waste of time, at least not from the author’s perspective.

I don’t know man, I sort of agree with you. You can probably program a computer to generate something in the style of say, “Bach,” or “Kurt Cobain,” or whoever, but you definitely have to spend time and money doing it, probably more time and effort than it’s worth. You need a team of talented programmers who have good ears and/or also happen to be musicians and are able to stay on top of changing trends. It will be interesting to watch this dynamic play out for sure, I think there will end up being more of a market for it in terms of custom work and not necessarily even stock music, because likely a lot of people will be attracted to the idea of creating a custom score without having to deal with a human, but even that might be too much effort for some people, and they might still just want to sift through a few tracks until they find one that works.

OK. I think we’re in different places on this and will remain that way, which is cool. People don’t always agree on everything Let’s let others chime in.

People will always buy a track that “resonates” with them. Others may buy a generic corporate track just to get the project finished because the deadline is upon them. There is a “text to speech” technology trying to eliminate the human voice over. It just won’t happen in my opinion because people seek out the “human emotional element”. It may be OK in some applications. For example, blind people or people who have serious vision disabilities can have the paper or a book read to them. Robotic voices will not put VO folks out of business because advertisers want a specific “Character/ Actor” to deliver their message based on the “human emotional” element. The same will always apply for music.

Computer generated music has been around for years. There is a program called “Band in Box”…Just type in chords, select a style, click play… you are hearing music…It will even auto generate a solo/ melody. I actually have used it as a production tool many times, but ultimately you have to merge your own personality and performances into the final track to create something that will resonate with the public and sell. I am an optimist, sure we may have to compete against robotic computer generated music some day…and that music will sound pretty good…but ultimately the human being will create better, more interesting tracks.

Garage Band, and Acid are similar concepts, I have created tracks with those programs before (really just for fun and to experiment while on an overseas flight for example) and have sold those tracks for high prices and have seen them get used on TV shows as background cues. Even when you are just “assembling sections” and not “composing” with your own two hands on a piano, you still have to have a sense for what works, and what will resonate with buying public.

It is true that it will take brilliant composers and brilliant tech minds to program computer generated music. The real question is why would they even want to create such a sterile offering to the human race? People always are looking for that “new Beautiful voice” (Adele) (Lorde)…People will always want to watch other human beings perform (All the world is a stage) So music and people go hand in hand. Don’t worry about it. It’s a conveyor belt…tracks get on the belt …ride their way down, then fall off, some can get back on when “oldies” suddenly are needed.

Regarding generic corporate tracks. It’s interesting how we all say “everything sounds the same, everyone is copying everyone, etc. etc.” I too made those statements when I first came here, but then I said to myself “try it some time”. And I say that to other writers with the same remarks “That corporate sound is just so lame and boring and repetitive, 3 or 4 chords over and over and over” Once you say to that dismissive writer “Give it a shot, let’s see what you have in your bag of tricks”…Things can change quickly. Creating a great corporate track IS a tricky craft and art form to master. This is why some guys sell thousands and others just a handful. Every production decision you make is so important for those tracks. What drum sound should I use? what shaker should I use? Which piano? How much reverb? EQ treatment, compression treatment, guitar tone? Tempo? All of those decisions are incredibly important and only humans can make those decisions. Some here at AJ really know how to make a track that resonates with corporate folk, others are still trying to master that style.

In closing, I have thought about this dilemma a great deal over the last 5 years, In the end, all I can do is analyze the facts: my tracks that sold the most are those that contained real human performances. A human actually played the drums, guitar, piano, and bass. A human actually sang. I am still shocked by my best seller on this site - a mariachi piece, it’s interesting though how every note in the piece was played by a living and breathing human being…not on keyboard either…a real, human guitarist, bass player, trumpet players, and a percussionist performed live in the studio on that piece. That says a lot about what the buying public really wants and we all need to give more of that to them.
Also I once read an article that humans prefer to rumble and dance to imperfect/ non quantized “human rhythm” verses perfectly quantized “computer rhythm”…So stop worrying! and start playing your instruments more often and ask friends to perform on your tracks if you can’t play the instrument you need on your track. Computers will NEVER be able to post score precise film cues in a scene.

1 Like

Just to be clear of what is being discussed, the topic doesn’t relate to all forms of computer generated music or any existing products. I’m specifically talking about artificial intelligence, which can duplicate the learning process of humans from ingested data sources and play back results using both high quality synthesis and high quality virtual sampled instruments that incorporate the same articulations a human can create. We’ve all heard very convincing orchestral mock-ups and music from most popular genres involving virtual acoustic instruments that can even fool most musicians. I’m talking about a system that uses AI to compose without just spitting out a pre-programmed patterns and can reproduce human playing techniques. This is specifically addressing the low budget RF market, no other forms of music.