This is very interesting, I'm sure many of you are aware of its existence already but as usually I'm behind the curve - DALL-E is an art generator which somebody won an art competition with, to the anger of other contenders.

https://www.creativebloq.com/news/ai-art-wins-competition

It also has the facility to analyse existing images and expand them:

https://www.pocket-lint.com/apps/news/162539-dall-e-outpainting-classical-paintings

I can imagine the next iteration of AI for music production doing something similar, if anything it may be easier than the visual realm because there are already established rules and conventions for it to follow. I like the idea of feeding one of my existing tracks into it and seeing what it comes up with. Maybe it will be able to recognise the creative elements of a track and use those for reference rather than simply identifying the key and tempo. Come up with a basic melody and chords and whatever else and let it do the rest.

I know it doesn't exist for music but I think it probably soon will. Thoughts?

https://openai.com/blog/dall-e/

 

AI is coming for sure.
Can AI compete with a truly talented person in this realm? No
To make the transition more seemless they have dumbed music down to idiot level, make it perfect, quantise the life out of it. Pigeon hole it into genres.
Why? it makes it easier for AI to compete.
All those music files you upload to the net, or all your music you send to online mastering etc, seems like a perfect dataset for AI to learn from to me, much in the same way social media is a dataset for AI.

At the end up, music is about being human, and expressing being human. It has soul.

 

The future depends on how good melody/riff generators have become. I'm out of the loop but do you know any good generators that are able to take a melody and create different variations of that melody instead of creating a new melody/riff everytime. Does something like this exist right now? I am aware there are melody generators which spit out a new melody/riff every time you click the "generate" button, but is there a software/AI service that acts like your band member that helps you with the melody you have already created and just create variations of your melody.

I would love to know what is the best melody/riff generator right now according to you and the rest of the forum members?

 

The thing I have observed so far with Virtual Software Synthesizers is the "Random Generator" button. I have never used it.

I can imagine that soon there will be virtual software synthesizers based on AI,
future market, then every uncreative user can feel like a great artist.

At that time there was from the company Superwave the. Equinoxe, Equinoxe is a powerful virtual analog synthesizer
that recreates the works of Jean Michel Jarre. This is already a small miracle, a real artist plays live with his band
and hardware on the big stages of the world and his work can be partially transformed in a software program.

The other big change in the past was samples. So the analog becomes digital. Is also cheaper to produce, there are also
no storage costs and delivery costs. From there the AI will come and if it is well made - so well fed with lots of data,
the AI will make many artists unemployed. who wants to work against the AI. What the many new unemployed will
do then is written in the stars. I think it kills a lot of creativity and we will increasingly become stupid consumers.

Make already some stickers, AI free zone. AI no thanks, AI kills creativity. Robots have to stay outside.
AI leads to more loneliness and social coldness.

A joke at the end or not: These thoughts were written by my AI writing robot.

 

This app is a free random midi melody generator which you can use to create melodies. The melodies are procedurally
generated and are royalty free. You are free to download the midi files and use them in your own music.
https://dopeloop.ai/melody-generator/?s=6860476082221891

Riffer Sequencer by Audiomodern
https://www.pluginboutique.com/prod...omodern-Riffer-Generator/?a_aid=5aff606c9e208

 

Audiomodern Playbeat (AI-powered intelligent plugin)

With Playbeat, Audiomodern apply an algorithm-driven approach to drum sequencing. On the surface, the plugin functions like a lot of other software drum machines. Users can import samples and sequence them across an eight-track step sequencer. There are parameters for varying pitch, level, flams, pan position, etc.

Where Playbeat is unique, however, is in its algorithms that can be used to randomise or remix a pattern based on current groove. This is where the ‘intelligent’ element of design lies – rather than simply randomise hits and parameters, it makes creative decisions inspired by the existing pattern. Buy: Audiomodern Playbeat 3(opens in new tab)

 

The most exciting AI app for me is Riffler in iOS, I bought it when it only produced guitar riffs - but really well, better than anything I've heard - and the creator has worked on it a lot and now it plays many more instruments and styles. It's also coming in a VST version.

 

I don't like AI.

 

OpenAI, creators of DALL-E, have gotten destroyed by Stability.ai since they've released a similar model but 100% uncensored/open source. It's ridiculously good for txt to img. Reddit: https://old.reddit.com/r/StableDiffusion

That same team is developing a similar diffusion model AI tool for music called HarmonAI. That'll be the thing to watch.

I've had a lot of fun w/ OpenAI's Jukebox which you can run using a Google Colab notebook. But that's more continuing WAV/produced songs vs. writing songs.

Good discord server that's been for Jukebox, but has a section now for early discussion about "HarmonAI"

https://discord.com/channels/766622617393430559/766622617393430564

 

Been using different AI tools around 3 years now.
This year the text-to-image blew up and is starting to be everywhere.
I'm not worried.
While it does the images better than most of us, it still needs guidance and selection of the human eye. Same thing with the AI music generators I've tried.
While the audio is still pretty rudimentary, it will evolve and fighting against it is quite futile, like it or not. So better to adopt the tools early to get to know the usage and how to include them into your own workflow and taste.
It's was the same thing with fighting against photoshop, digital painting and analog versus digital daw's. Use what you need, leave the rest out and don't get bitter the world has always evolved people like it or not.

 

You would have to go 20 years ago and ask yourself if there is any program capable of generating melodies, and the answer would be no, now go 20 years ahead and ask if any AI is capable of creating an interesting melody, the answer would be another.

But that already exists, but in a different form.
The autotune live. And they think they are good singers.

 

The author of Riffler posts updates on twitter, very interesting developments...

https://twitter.com/rifflerapp?lang=en-GB

 

I respectfully disagree.

See the thing about AI is that it isn’t just a tool like photoshop or a DAW. Photoshop still requires a learning curve, you can’t just go on photoshop with 0 experience and say a few words and have things happen exactly how you want them. The fact is, this will set the bar super low in terms of effort, it will no longer matter if you spent a long time working on a piece of music or art because an AI will do it much cheaper.

when people say “oh it’s just another tool, it’s just evolution”

it really isn’t… the fact is most people/companies who pay for original images, original music, are going to be sold a subscription model to these AI programs once they’re mature enough, making real human artists the second option almost every time.

The problem is most artists who don’t see AI as a threat are thinking like artists and creators not like companies or businesses.

think about it from a capitalist business perspective. Why would you hire an artist to paint a logo for your company for $5,000+ and negotiate when you can just subscribe to an AI image generator program which only charges you $20 month? You really think they’re going to pick the artist first? Hell no.

hiring an artists takes time, you need to interview, sign contracts, review, make changes, etc. While with the subscription model it will just be like “click this button for a random cool image” and then BAM they find what they’re looking for. And if they don’t, then and only then they will probably reach out to a real artist.

AI subscription models will come for other industries too, they haven’t even taken on music generation fully on yet. I know there’s a lot of cool generator tools and music AI things out there but as far as I know they’re NOT mature and or trained with the same amount of data sets as dall-e or mid journey.

You say you can adopt these tools but you don’t realize the whole purpose of these tools is to make them so that any one can use them even a child. You once the subscription models start rolling in they’re really going to make ease of use a priority.

again, I’ll set an example… why would any company or business hire me or you to produce an original music track when they can just download the new update for their AI generator app that makes it even easier to generate the perfect music for their type of business? You think they’re going to need us for that? Lol , dall-e can basically be used by a 5 year old today, now imagine in 2 years once these programs mature and then the tech gets applied to subscription models, and then they start focusing on catering to businesses, etc.

In my opinion, subscription models + AI = genocide for independent artists.

 

couldn’t agree more.
-also written by AI

 

Very true, but even bad singers still have to push breathe out of their lungs.

AI is the equivalent of having a voice box installed on a robot and if you don’t like the voice you press a button and you’ve got a whole new singer.

 

Singing will be the final frontier for AI, the big stumbling block - every other instrument will be done and dusted within a decade or two, maybe sooner.

 

AI makes it possible: clone voices for a few dollars a month
Five years ago, the first machine learning systems made it possible to imitate human voices.
These programs are now available for little money.

On the other hand, other tools will sooner or later allow more or less simple voice theft. Cloned voices can sound very real, so that audio tracks produced with them can hardly be distinguished from real studio recordings. The most noticeable thing about them is that they sound a bit too perfect, don't contain any throat clearing or noise, and the imitated persons speak in an unnatural way "as printed", depending on the text that is slipped underneath.

Source German:www.heise.de/hintergrund/KI-macht-es-moeglich-Stimmen-klonen-fuer-wenige-Dollar-im-Monat-6144114.html
Translate with Google: https://www-heise-de.translate.goog...l=auto&_x_tr_tl=en&_x_tr_hl=de&_x_tr_pto=wapp

 

Here is a link to Nvidia's language research, very advanced...!
I can't answer the question of whether technology is making us lazier and stupider.

Nvidia: New AI voice sounds almost real
German Original: https://the-decoder.de/nvidia-neue-ki-stimme-klingt-fast-wie-echt/
English translate: https://the--decoder-de.translate.g...l=auto&_x_tr_tl=en&_x_tr_hl=de&_x_tr_pto=wapp

 

Music is way too complicated to be replaced by AI. It may be possible at some time, but it's gonna require so much computing power(and thus energy) that it'll be pointless. It's if we talk about fully produced pieces, and not just some random ass notes that sound "musical" together.

BTW, where AI really could help is automating routing work. That's where its development should've been directed from the very beginning.

 

I agree with you, singing will be harder and so will music but at the same time, if you think about it the top songs of today are very simple in nature. Not saying top charts are “good” songs, but mainstream music is becoming more and more simple to the point where top chart artists liked drake don’t even bother to have good mixing in their album because people just don’t care about the sonic quality of it as long as it relatable and or comes from a big celebrity.

Case and point. This “artist” is literally topping everyone right now and honestly, it could be anyone singing, the auto tune is so heavy that you could probably switch the “singer” and no one will notice. Also, the mixing is terrible but check it out. 1.1 Million views in 6 days.

While I don’t think AI will be able to compose an orchestral piece any time soon, I do think they will be able to replicate mainstream rap music like this since it’s rather simple. (Not hating on rap btw)

As far as the data churning and power consumption goes, you’re also right about that. However , you gotta remember that quantum computers were created for this exact purpose, we have yet to see how quantum computers will influence the next 10 years.

What I predict will happen is that everything will become more and more meta. Everywhere you look people will be trying to sell you something but it won’t look or appear like they’re selling you something. Take TikTok “influencers”… most people trust in them because they think they’re just like them because they’re posting from their phone and they talk just like them, they think “oh man this person is just like me, an average joe. They even dress and talk like me”

but most people don’t even realize that almost everything they see now on social media is a paid ad disguised as information, news, comedy, etc.
AI and quantum computers are going to accelerate this process.