Over the last 1.5yrs, I've worked with my colleagues on Melodrive, an AI music engine that enables people with limited to no musical skills to easily create highly adaptive music. The music engine uses AI to generate music from scratch, in realtime. The music smoothly transitions between different emotional states. We're planning on releasing the engine for free to indies. This means that indies can quickly get a soundtrack, which is more dynamic than anything that exists today on the market - composers included! 

We’ve just released four Melodrive tech demos. These showcase Melodrive's AI music potential across a range of experiences from VR to music branding. You can download the demos directly from our website here: http://melodrive.com You can also sign up to become an alpha tester.

Hope this may be of interest to you!

While I appreciate all of the effort and systems that go into AI generated music, the actual music itself rarely works for me as a listener or as a player. AI generated music could work in smaller doses, mixed in with more traditionally composed music. But for most games I know of (and most players I've talked to) seem to want more structured soundtracks that can red thread the main theme (or themes) into various points in a game. Best of luck though! Sorry to not be more excited about it but that could just be my own personal bias. 

Just now, nsmadsen said:

While I appreciate all of the effort and systems that go into AI generated music, the actual music itself rarely works for me as a listener or as a player. AI generated music could work in smaller doses, mixed in with more traditionally composed music. But for most games I know of (and most players I've talked to) seem to want more structured soundtracks that can red thread the main theme (or themes) into various points in a game. Best of luck though! Sorry to not be more excited about it but that could just be my own personal bias. 

Hey nsmadsen, thanks for the thoughtful feedback! We're actually addressing the issue of coherence and self-reference throughout a soundtrack. I very well know that so far most of the generative music systems lack on that point. To address that, all the music that Melodrive generates is built around the idea of a musical theme. A musical theme can be implemented into a number of different styles/emotions (in realtime) and attached to game objects and locations. Moving forward, we'll also try to implement concurrent musical themes in the same musical setting, if say for example, the boss and the protagonist, with their unique themes, are both in the same scene.

Apart from being AI researchers, we all have a strong musical background and are trying to have our AI "think" as an opera/video game composer would think. We definitely don't want to replace composers, rather we would love to arrive at a point where the composers can direct the AI in order to create highly adaptive music which is not possible today, given the obvious limited resources of a composer. We call this deep adaptive music. With deep adaptive music, human composers/devs can provide the core musical ideas to the AI, which then implements these in realtime in order to fit the non-linear experience at any time. Hope this clarifies our approach

10 minutes ago, Melodrive said:

We're actually addressing the issue of coherence and self-reference throughout a soundtrack. I very well know that so far most of the generative music systems lack on that point. To address that, all the music that Melodrive generates is built around the idea of a musical theme.

Interesting! Do you have any samples of this that we can hear now? Or is this still in development? 

You can take a look at our demos at  http://melodrive.com. There are also short video trailers on the website, so you don't need to download them. However, the demos are relatively simple use cases, so they don't really show the full range of possibilities of the system. (We kept them simple, as we already found that people outside of video game music has a difficult time grasping what we're doing!). "Oceania" and "SoundSlides", show the emotional change in the same style with same musical theme. With "WorldMusic", you can listen to the same musical theme passed through different styles. In "LightSynth" you can see how users may directly affect the emotional state of the music, always based on the same musical theme. Unfortunately, we don't have a demo that showcases multiple musical themes.

I think the concept is a wonderful idea but the examples that are on that demo reel need a LOT of work. I seems that it would benefit you to have some long and in-depth conversations with people who understand how music structured and how arrangements are built. Going off what I'm hearing on that video, your  AI has got to be infinitely more refined before its going to become a usable tool, not to mention a sellable one. 

Kudos for making the effort to explore the use of AI in music. Keep at it! I hope you guys get there  Neat project.

I checked out your Oceana demo and I very much liked it! It is a very fascinating concept and none of the melodies generated by this system seemed out of place to me. It was simple, ambient and tonal, the way I would expect from an underwater scene.

My guess is nsmadsen did not actually listen to the content before commenting, as your primary point in this software is to allow indies without a budget to hire a professional composer or with little access to stock resources to still have interactive music in their game that adapts to the player's movements and the game content itself.

I noticed how the tones changed with the shark and the jellyfish and I am wondering have you pre-programmed midi melodies to control instrumentation and respond to in-game cues with a sequence of "danger" melodies when danger occurs and "mystery" melodies when mystery occurs, etc? Or are the melodies actually generated algorithmically from scratch as well?

The latter would be very impressive, but either way, I think this is a very cool project and I wish you success with it!

19 minutes ago, Jordan Winslow said:

My guess is nsmadsen did not actually listen to the content before commenting, as your primary point in this software is to allow indies without a budget to hire a professional composer or with little access to stock resources to still have interactive music in their game that adapts to the player's movements and the game content itself.

Actually, I did listen to several of their videos. But, as I stated before, I just don't usually like AI generative music. AI tends to make decisions that I don't care for and prefer the touch of human composers. AI generated music does have it's usefulness but it also has some strict limitations. Some of this is most certainly personal bias but it seems to be shared by many composers and gamers I've talked to over the years about this very topic. 

6 hours ago, Jordan Winslow said:

I noticed how the tones changed with the shark and the jellyfish and I am wondering have you pre-programmed midi melodies to control instrumentation and respond to in-game cues with a sequence of "danger" melodies when danger occurs and "mystery" melodies when mystery occurs, etc? Or are the melodies actually generated algorithmically from scratch as well?

 

The latter would be very impressive, but either way, I think this is a very cool project and I wish you success with it!

Many thanks for the nice words! As for the melodies, as well as for all the other musical elements, they are all generated algorithmically from scratch in realtime. We don't use any pre-composed midi melodies.