Hacker News new | past | comments | ask | show | jobs | submit login
Ask HN: Recommendations for AI generated music software?
205 points by andrewstuart on July 25, 2019 | hide | past | favorite | 83 comments
It's surprisingly hard to find a solution for these simple requirements:

I want to create AI generated to use as soundtracks on videos - so there must be no question about royalties or copyright

I want to be able to specify how long the music should go for.

I want to be able to save the music.

It would be great if I could have some control over the pace/style of the music but I'm not a musician so lots of options would be detrimental. I don't want to compose music - I want it made for me.

I'm happy to pay, but the prices must be clear and up front and not require me to email the company to "make an enquiry".

I've done a fair bit of searching but not found anything to make AI music that meets these needs. It's strange given there is so much interest in AI I would have thought there would be plenty of good products available to buy.

Ideally I'd like to make music such as ambient music, similar to Brian Eno, Jean-Michel Jarre, Phillip Glass, but it would be great just be able to explore a range of types of music to find what I want.

Can anyone recommend any good AI music generators?




What you want is a plain old music generator app. Search for "generative music". Drop the Ai, it's overkill. Also with the amazing tech these days anyone can be a musician. It's basically push-botton.

Bloom on the iOs app store. https://apps.apple.com/us/app/bloom/id292792586

Someone's cloned Bloom: https://github.com/generative-music/blossom

Wotja too maybe. Here's their FAQ about rights: https://intermorphic.com/wotja/faq/#faq-recording-ownership

Wired can point you in more directions: https://www.wired.com/story/generative-music-apps/


I just want to highlight that Bloom was specifically created by Eno, whom you mentioned. He considers the generation his art. But the github repo 'clone' is under MIT so just use that.


> Drop the Ai, it's overkill.

Well, if you do want to use AI, this project is totally open source, "real" AI and pretty damned easy to use,[0] and generates something that's pretty darn close to SOTA as far as actual "musical" interest goes! Unfortunately, it seems to need Python 2. It would be really nice if someone forked the project (it's pretty researchy/proof-of-concept right now!) and committed to maintaining it for the foreseeable future, to run on whatever supplants Python 2 as the "standard" platform for Theano, when that version of Python goes EOL.

[0] https://github.com/hexahedria/biaxial-rnn-music-composition


While really really cool I don't think this is polished and directable enough for OPs needs.


>Also with the amazing tech these days anyone can be a musician.

If they’re musicians, George Jetson ran a factory.


I understand your point. Yes, push button music is a scourge ;)

Though, let's draw a line between someone who can make music for unobtrusive background applications (like OP) and someone who can make music for entertaining lots of people, perform it in person, and do both for decades while keeping a high level of quality (like Bruce Springsteen or Björk).


Thank you for sending me down this rabbit hole.

Eno's Teropa site is amazing, and I've played with the generative drum machine for hours.

https://teropa.info/loop/#/stochasticdrummachine

I recommend people start at the beginning, you progress with space bar. https://teropa.info/loop/


> Also with the amazing tech these days anyone can be a musician. It's basically push-botton.

And this is why it's getting harder and harder to find decent electronic music that doesn't sound like a chimp threw its shit at a wall; because some wanna-be with no musical skill can download a few loops, layer them up in their cracked FL Studio (or whatever) and over-compress the shit out of it for their "mastering". imo electronic music was much better when there was a barrier to entry in so much as you needed about $5k worth of equipment before you could produce something semi-decent.


I think there's just more noise tbh. The overall output of good music is without a doubt much higher without the 5k barrier to entry - you just have to search harder to find it


I fell ya, and used to think the same way too.

I've even spent that same $5K and got that equipment.

However, there's something missing from weekend bedroom producers, the hunger to be the best and dedication of their entirely life to it.

How? Open mic gigs, playing festivals, being friends with other producers, having lots of charisma, cranking out the tunes and adapting until one hits that sweet spot over and over again.


Not "AI" generated, but generated nonetheless, this HN thread from a few months back should be pretty helpful :

"Generative.fm – Endlessly unique ambient music" https://news.ycombinator.com/item?id=19397817

Be careful: this thread is great, I spent a full evening digging into the various links people posted. You've been warned.


I run a company that tries to solve your problem!

We haven’t done a big public launch yet - but we use AI techniques to compose work for existing clients (think businesses + Youtube video creators).

We try to take all friction out of the process.

Here’s how it works: You send link to your video. Tell us what emotions you want / any requirements (we have a simple form). Our AI engine then structures and creates a song exclusively for you (no royalties / copyright). We then have our in-house composers make final adjustments and sync it perfectly w/ your video. We price a lot less than normal composition studios due to the AI piece.

Trying to talk to as many customers as possible. If I could ask you a few questions or if you want to be a beta customer please email me at: leagueduels@gmail.com I’d greatly appreciate it!


FYI: Our typical rate is $200 for original soundtrack for up to 3 minute video + $25 for each additional minute. Includes the composition and syncing to the video with intros / fade-outs, etc. Comes with exclusive life-time license to the track. We currently have clients paying at many different price points, so still trying to figure out the right one.


What if a client does not like the result?


We deliver a demo first. Clients can then request any changes before we lock it in as final. Clients can also send links to music they like and we try to achieve a similar sound.

It's a good point though - we work on a small scale now. We haven't thought through all the details once we start taking new clients. One free revision seems like our most likely approach.


If you have to employ composers, why do you need AI? I’m guessing you can hire 3 composers for the salary of a single ML engineer.


I'll hazard it's because they can produce more volume if they're just doing corrections and syncing


http://aiva.ai is exactly what you’re looking for.


Not what you're asking for, but I think you should take a look at band-in-a-box (https://en.wikipedia.org/wiki/Band-in-a-Box).

When I studied music this was very interesting to me. A lot of music composition is fill-in-the-blanks / formulaic, and this generates melodies and countermelodies / accompaniment following a common (user selected) chord progression, that sounds correct and would probably pass as an assignment from an undergraduate music composition student.

There's no AI as far as I know, but if there was a Turing test for music, this would pass.

Coupled with a bit of arrangement and an original melody it can be quite nice.


There's musenet[0], but it seems like there's only an API to the web generator[1] that might work for what you need?

[0]: https://openai.com/blog/musenet/ [1]: https://github.com/bmckalla/musenet-api


If you use Ableton Live you should check out Magenta Studio https://magenta.tensorflow.org/studio/ableton-live/


Holy crap I’ve been wanting to make something just like this. Awesome to see that it exists.


Uf damn, just seen that one of the companies thatdid this https://www.jukedeck.com/ is not doing it any more.

They had everything for free, and the music seemed perfect for background music for videos, etc... You could configure parameters, including the time and get an interesting track out.


Yup, Jukedeck solved exactly this. Looks like they got bought by ByteDance very recently (https://musically.com/2019/07/23/ai-music-startup-jukedecks-...)!


Not sure if this helps at all, but here's one that was featured here on HN not too long ago:

https://generative.fm/


Thanks for sharing this, I'm the author of Generative.fm and the music systems on it.

Just wanted to pop in to tell OP that there's a recording feature built in to the site, and everyone is free to use the recordings for whatever they'd like - I just ask that you give attribution (it's licensed under CC-BY 4.0). Though, there's no configuration and I don't use AI, if that matters.

EDIT: I forgot to mention it's free, and all the music is ambient.


The web-site shows an empty page at this moment.


Would you mind telling me which browser you tried with?


It is due to certain settings in the browser profile that I use for everyday surfing. Without giving too much information, let's just say that although some things are blocked, popular music and video streaming web-sites still work, so I guess this one should also work. It works in a clean profile with default settings in the same browser.

There are some errors like this:

Loading failed for the <script> with source “https://platform.twitter.com/widgets.js”. generative.fm:1:1

Loading failed for the <script> with source “https://www.googletagmanager.com/gtag/js?id=UA-134692042-1”. generative.fm:9:1

TypeError: this.context.createConstantSource is not a functionmain.7acf3eaa3b54511db064.js:7:5340


Check out Amper:

https://www.ampermusic.com/

My good friend created the tech. They use a lot of custom acoustic instrument samples and some sound synthesis driven by Haskell and SuperCollider. No neural nets or machine learning involved (which is what most people seem to mean by "AI" these days), it's mostly rules based, with a large and byzantine rule set which has been cultivated, so some may call it Algorithmic or Generative music.


So close, it says on the front page that it can be used on videos!

Amazingly, the first link is one so you can listen to the output!!

There's a product link! ..... which leads to nothing about a product.

And there's no pricing information at all.

It looked really promising but I had to conclude there's nothing there to buy.


You can try Amper for free here: https://www.ampermusic.com/lp/royalty-free-ai-music-1/


Looks cool, but has a contact form instead of a price sheet.


Not quite "AI generated", but Dunc's Algomusic[0] is free. I don't think you have THAT much control over how it sounds beyond picking different seed inputs. I find it especially great for weird/unsettling background music. Not sure if there is a built in way to export tracks but there is probably some way to use another app to record it.

[0] https://marasmusine.itch.io/duncs-algomusic


To add to this list, I haven't used it but I've recently listened to a few demos of the commercial orb composer [1]. I personally think that the melodies and hooks of these music generators usually sound a bit lame, but orchestration and accompaniment somehow make up for it.

https://www.orb-composer.com/


> I want to create AI generated to use as soundtracks on videos - so there must be no question about royalties or copyright

> I want to be able to specify how long the music should go for.

> I want to be able to save the music.

Product question @andrewstuart - is it important to you that the changes in video scenes line up with the beat of the music?


>> is it important to you that the changes in video scenes line up with the beat of the music?

Well yes in an ideal world it would be great to feed in a video and have appropriate music created to match scenes etc but I can't imagine how that would work without fiddling on my part, and it's hard to imagine it being cheap. I want to do this cheap.


Royalty-free soundtrack music is incredibly cheap. It would be faster, better, and cheaper to buy what you want.


How cheap?


Hi HN. There is cool app that lets you do just that, it's called Lonofi. They are currently in open Beta test and you can check them out here on https://home.lonofi.com

It works sligtly differently than the other AI music website:instead of just choosing a style an pressing play to get random music, here you actually choose the sounds and instruments you want to hear, and an AI mix them together intelligently to create an ambiance that play forever without repeating itself.

Given it's AI, it's all royalty free. The download function isn't available now, but if you create a scene and want to download it, just ask the guys, they will let you do it for free at the moment (in exchange for a good review) .


Its implementation may be a bit academic for what you're looking for, but Watson Beat (https://github.com/cognitive-catalyst/watson-beat) can generate music given a seed melody a number of configurable parameters.

All output is in MIDI files, and by default need to be glued together manually (though I suspect this could be automated).

I've also wondered how interesting this could get if it could be configured to dynamically feed the generated melodies back in as the seed, effectively evolving the music over time while remaining somewhat familiar from one adaptation to the next.


David Cope's Common Lisp one? http://artsites.ucsc.edu/faculty/cope/cmmc.html


Hi!

I am currently working on the AI-generated music problem

Every output on the site is unique and public domain, feel free to ask me about any questions

Link: https://hookgen.com/


The red herring here is AI. I believe most everything you’re trying to accomplish can be done using a popular and longstanding application called Band in a Box, from a company called PG Music.


This[0] is not a music generator, but a very handy tool for music makers. It extracts vocals and/or produces a karaoke version of a song. It claims it is using AI and I suppose it does so because the results are astonishingly good. It's not just the removal of the "center" channel which most vocal removers are doing. It literally pulls the vocal track out of the mix like magic.

[0]: https://phonicmind.com/


A bit late to the party, but I wrote a generator for a project last year running on vanilla JS with a simple HTML front end. There is no recording feature, but I'm sure you could just record the playback externally.

https://github.com/gee842/genmusic

hosted here: http://genmusic.byethost11.com/genmusic/


at the suggestion of another user, I changed hosts to github pages, rather than a shady web-host.

https://gee842.github.io/genmusic/dev/index2.html


https://boomy.com - try the “ambient” filter in beats, or anything in meditation / relaxation style


As with most of the options suggested in this thread, no pricing and no explicit definition of whether it can be used on videos and no clear definition that I can save output of a specific length.

I get the impression that most of the music software companies are run by people who are very musically/engineering oriented, but not very connected to the business side of things.

I shouldn't have to send an email to a company saying "how do I buy this?"

Without any obvious product or service for sale, it's not discernible from a hobby project.


Boomy's EULA [0] is a fairly easy read that covers the business side of things.

> In consideration of free use of Boomy, any arrangement, sound recording, composition, and/or musical work accessed through the use of the Boomy platform, shall be the property of Boomy Corporation as a work-made-for-hire.

> The user may not edit, remix, or create derivative works of the song outside of the Boomy application without purchasing ownership.

> The user may not distribute the song to streaming services without purchasing ownership.

> The user is permitted to purchase the ownership from Boomy for a fee outlined in the Library section of the Boomy app.

[0] https://boomy.com/eula


https://boomy.com/about

can change length via editor


In order to use an AI generator, it has to be fed with music that is in the style you like so that a model is created and trained. The question of copyright and plagiarism in such a case is not really clearly defined. The patterns it will generate will not be same as existing ones, but can sound or seem similar. As if the "AI composer" was influenced by the original song composers.

Has anyone any comments regarding such issues?


One option is Wolfram Tones (tones.wolfram.com/generate). It uses cellular automatons (so not quite AI but achieves your goal) to generate music.

- You can change the type of music (Jazz, Classic etc.), select instruments and so on

- You can download the music as well

The only downside is that you can only get a maximum of 30 seconds. What you can do however, is go to their free programming lab and write some code to generate any length that you might want.


I found an app called Mubert which seems close to what I want, but as with all of these generative/AI music things, they seem to carefully avoid giving straight information about price and the licence for using the music on videos.

Startups: please make something that I can pay money for that lets me make music clips and use them on videos, that I can buy at a clear price without sending an enquiry email.


Perhaps this product by Teenage Engineering would wet your whistle. https://www.nytimes.com/2019/07/25/magazine/electronic-music...


I was going to type Jukedeck. Thought I'd visit the site first. Offline. Apparently TikTok has bought it!


Is there anything that can generate music for in the style of a particular musician? It would be cool to generate a clip that sounds like a particular artist or band that convincingly sounds like them jamming / working on a new song, but it's actually some made up AI generated song.


It's worth considering the complexity of this problem and how notable and popular services have approached this https://en.wikipedia.org/wiki/Music_Genome_Project


Not a generator, but maybe inspiring nonetheless: https://daily.bandcamp.com/2018/01/25/music-ai-coding-algori...


There is Ludwig by ChessBase, which fits the AI requirement. You will have to find a way to come up with the primary melodies, though: http://www.write-music.com/


Maybe Mubert is what you are looking for? https://mubert.com/

It has a lot of channels including ambient music.


http://tones.wolfram.com/generate/


You can try my webapp, https://www.muzoti.com


Slightly orthogonal to your request, but I'm not quite sure what you get by choosing machine generated music over human generated music WRT copyright. The machine generated music will be considered a derived work under copyright, so you're going to need a license for it anyway. You could just obtain music from a human and get a license as well.


I don't think this is true. The output of a program is not considered as a derivative work of that program. E.g. Is a photo taken by a digital camera considered a derived work from the camera firmware?


That's a different situation. The digital camera is not creating the work. It's analagous to a compiler generating code. It used to be the fashion of compiler vendors to not give you a free license to the derived work that the compiler generates. GCC even has a separate clause in its license to allow the output not to be GPL. See this license: https://gcc.gnu.org/onlinedocs/libstdc++/manual/license.html


Hello from Amper! You can do exactly this with Amper, and you can try it for free: https://www.ampermusic.com/lp/royalty-free-ai-music-1/


Code parade made a pretty good music generator that is to your specifications https://youtu.be/l-ZZ8dkJVOI

And yes, it is true AI.


The quality of music in their YouTube videos is pretty awful. Do they have better samples suitable as video soundtracks?


You should be able to download the software and generate it yourself!


I prefer listening to some samples before I install anything. Have you generated anything good with their software?


Good chiptune video game music. You can learn more about the process here: https://youtu.be/UWxfnNXlVy8


Yep, that's the video I watched before I said the quality of music is awful. Anyway, I'm not sure how is "chiptune video game music" relevant to what OP is asking for.


They asked for a program that generates music with AI, which this is.


>for these simple requirements

What gives you the idea that these requirements are simple for an AI to live up to? I'm only a novice in AI, but the product you're asking for seems very complex.


It is pretty simple:

The app lets me make music clips of a certain time duration, save them and use them on videos royalty/license free. The app gives me some control over pace/style of music. I pay either a once off or monthly fee to use it.


Simple to describe, not accomplish. Think about the sheer amount of genres/sub-genres, what you're describing can probably be done in a few popular genres by some ML algorithm.. But are the results good enough?


OP wants to make ambient music, much simpler than generating music with lyrics.


Shameless self-plug, I maintain a Spotify playlist that is optimized for getting me into "Flow." I use it for coding/thinking. It has 677 tracks so far. See here: https://whoami.sh/thought/flow-playlist

https://brain.fm/ also has some good stuff. Some of it is tailored for focusing and works pretty well as coding music. I believe their stuff is AI-generated.


That looks like a nice playlist (I know lots of the artists and will have to check out some of the others), but it's completely unrelated to someone asking for a way to generate copyright-free music to use in their own videos.


Dude, even though it was a bit off topic, I have been listening to this playlist for around an hour and I'm pretty stoked on it! Thanks! Funny, I already had quite a few of these songs in a playlist of my own :) Great stuff.


Hi, thank you for compiling and maintaining the playlist. It's been weeks since I first discovered it and it has replaced brain.fm for me.


Used your playlist for some minutes and it definitely helped me being more focused, thank you.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: