• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

AI and Future of Music Production

This quote doesn't sum up my thoughts well (it's my bad). I've watched your video. Just like everything I've heard that's AI generated it's just really not engaging (euphemism for garbage).

Ai will have a very hard time exiting the '' uncanny valley '' and even when it will, it'll just be a very fast copyist <---- (that's more of a quote that sums up my message).
 
This quote doesn't sum up my thoughts well (it's my bad). I've watched your video. Just like everything I've heard that's AI generated it's just really not engaging (euphemism for garbage).

Ai will have a very hard time exiting the '' uncanny valley '' and even when it will, it'll just be a very fast copyist <---- (that's more of a quote that sums up my message).
Sure, it’s may not be there yet, but it’s really close, and given that it came basically came out of nowhere and moving really fast, it’s just a matter of time.

By this time next year, probably every new song will have some AI element, and you won’t know…
 
Sure, it’s may not be there yet, but it’s really close, and given that it came basically came out of nowhere and moving really fast, it’s just a matter of time.

By this time next year, probably every new song will have some AI element, and you won’t know…
Yes you're right, there'll be AI elements curated by a human brain. As i said AI is a powerfull tool, but very far from what i was sold, i think i won't be happy with it until it's renamed .
 
Yes you're right, there'll be AI elements curated by a human brain. As i said AI is a powerfull tool, but very far from what i was sold, i think i won't be happy with it until it's renamed .
It’s also called an Intelligence Quotient (IQ) regardless of its value ;)
 
I think I want to hear something like Tom Waits singing Nessun Dorma. That'd be awesome.
 

AES_AiMusician_Conference_1200x675_3.png.webp
 
Yes you're right, there'll be AI elements curated by a human brain. As i said AI is a powerfull tool, but very far from what i was sold, i think i won't be happy with it until it's renamed .
Renamed? I guess you mean that it's not intelligence. But artificial turf is not turf, it's something that mimics turf. There's no expectation that artificial anything (leather, meat, color, flavor, flowers) is indiscernible from what it mimics, so I don't see why AI is any different.

This quote doesn't sum up my thoughts well (it's my bad). I've watched your video. Just like everything I've heard that's AI generated it's just really not engaging (euphemism for garbage).

Ai will have a very hard time exiting the '' uncanny valley '' and even when it will, it'll just be a very fast copyist <---- (that's more of a quote that sums up my message).
I think it's already exited the uncanny valley, that's why everyone is stunned that it happened this quickly. If it can fool the average person, it's not in the uncanny valley.

At the moment, udio has a touch of the uncanny valley for me, personally, particularly in the vocals sometimes being "over the top", a bit of a caricature of the style, as compared with suno. But udio is still in beta, and is cleaner that suno. It won't take much improvement for it to be extremely convincing. A lot of musician already prefer it.

I've played some suno output for musicians, and they couldn't tell. My brother-in-law, and electrical engineer and musician on the side, had me sing on a song of his. On his birthday recently, I sent him a country-style suno song with the lyrics from his song. Suno kicks out two at a time, I sent him both. One was more authentically country, the other the modern country that's pretty much a basic rock tune with twangy vocals and the typical country harmonics. He said he liked the second one best, as being closer to his original. He asked if I sang everything on both. I told him everything was AI generated. I gave him another example of suno output, he asked who the female singer was. That's not the uncanny valley.

Also not in the uncanny valley: AI voices for dialog. AI recreations of celebrity voices. AI video generation...

You could find fault with this, but honestly, if this were an opening scene to a show set in the gold rush, with titling and opening credits, with music behind it...how much better would it need to be before you didn't notice that none of it is real?

 
I don't think this has been discussed. Major social media platforms license music from the owners. That incluses Tik Tok, YouTube, etc. Universal was unable to agree with Tik Tok on licensing terms. It is reported the issue was extremely low payments to the musicians and user-created AI content using the music.


They recently announced they have come to agreement.
 
Yes, tons of new music will be generated.

I'll have a listen, but at the end of the day, I'll surely still listen to my favorite five or ten albums of the 70s and 80s all the time.

I won't even have a listen. The five or ten (hundred) albums of the 70s and 80s will do :)
 
“BT UNDERSTANDS HOW PERFORMERS VIEW AND VALUE THEIR VOICES, AND SOUNDLABS WILL ALLOW UMG ARTISTS TO PUSH CREATIVE BOUNDARIES USING VOICE-TO-VOICE AI TO SING IN LANGUAGES THEY DON’T SPEAK, PERFORM DUETS WITH THEIR YOUNGER SELVES, RESTORE IMPERFECT VOCAL RECORDINGS, AND MORE.”

Put a different way, aging artists will be able to record notes they can no longer reach. Another singer will be able to perform a song in a different language, and the "artist's" AI voice will replace it, for sale in another market. And they'll use it to fix bad notes in live recordings.

(Sorry, but is the world clamoring for their favorite artists to "perform duets with their younger selves"? LOL)

:)
 
My friend, who can play a bit of guitar but is fairly good at computers, has dabbled with AI as a hobby, created this song. I don't know what programs he used. Maybe he did some mixing himself. I do not know.

It's a techno/trance version, pardodi, on a popular Swedish drinking song.

Helan går is a popular Swedish drinking song, or snapsvisa. Helan ("the whole") is an expression signifying the first (small) glass of spirit (commonly akvavit or vodka) in a series, and går means "goes (down)"; loosely translated as, "Bottoms up!" Thus, it is commonly sung as a toast, typically for the first glass of spirit at a seated dinner. The song has also become quite common in Finland, especially at academic dinners.


Fascinating what you can do these days. Generic techno beats in the song. No innovative music other than innovative in the sense of what a non-musician can do these days. The singing voice is definitely not his so it is processed in some way::)


The image on the video is most likely AI generated.

Edit:
I will ask him how much of the singing is AI generated.
...
Answer from my friend how he, or AI, or both made the song

AI didn't come up with the text. The ending with the echois mine. Otherwise, most via AI. But it's not just about pressing a button. Took probably 3 to 4 hours. You have to tweak and iterate like a thousand times before you find the right one. Can also control structure like intro verse chorus breakdown and so on.

I have used Udio. There is a free version. It was fun, try it yourself
.:)

 
Last edited:
After the flood of boring human made music the last 20years or so. We now get a even bigger flood of ai music thats even more boring. ;)
 
OK...I'm a songwriter and I can play keyboards well and several other instruments slightly (especially drums and guitar)...to be brief, I don't need AI, nor do I want something to generate music or lyrics or vocals for me. I'm not look to write the next hit, I only care about the journey of spark of an idea through developing it and recording it. But it's still interesting, and I'm not averse to using AI at some point or some degree. I wouldn't mind at all having AI listen to my song and say, "here are a couple of other ways you might try that phrasing", or harmony, or transition, and "let me play an example for you". Maybe I'll like it and use it, but more likely (and interesting) is if it doesn't exactly fit for me, but makes me think of other possibilities.

People in music will recognize this as getting input from a producer. One of the most successful names of all time is Rick Rubin, who doesn't play an instrument and can't run a mixing console. But as another big name producer said at a showcase I attended, "he has the gift of being able to listen as a fan." I know some of the stories, but I'll keep it short. OK, one: If you know System of the Down, Chop Suey, you know "...Father, into your hands, Why have you forsaken me? In your eyes forsaken me?..."—when stuck for inspiration, Rubin suggested that Serj open this book to a random page, drop his finger and sing what he landed on.

The point is, those with success, connection, or money, can get this kind of help. With that in mind, maybe getting AI help isn't so impure.

Also, most people don't know how many modern hits are recycled old hits. Even literal samples lifted and looped. Sure, sometimes it's just singing (or rapping) over old music, but sometimes you won't recognize it, it becomes something new and fresh. The youtube hit (2.2B views), Gotye - Somebody That I Used To Know (feat. Kimbra) [Official Music Video] is made of bits of several songs. the most recognizable was Luiz Bonfa - Seville, and the melody of "baa baa black sheep"on xylophone, along with lifted rhythm sampling.

The point is, the stuff that some people worry about with AI happens all the time, has for decades. That really cool drum intro to Zep's Rock and Roll is lifted directly from a Chuck Berry song. I'm not sure how badly I feel that normal folks might have access to such ability—even as a musician myself.

Out of AI song generators, I've mostly worked with suno. It was a little interesting to put in my lyrics and have it imagine a song in a style I request—mainly to hear how it might phrase my lyrics differently than I do. but it got a lot more interesting when they added audio input capabilities. It doesn't allow a whole song, it takes 6-60 seconds, then generates an additional four minutes starting from where the input ends. If an "ahh" is cut off, it might continue that vocalization in some way, so that if you connect the two, it's a seamless transition. Interestingly, if you put in music with a verse and a chorus, for instance, and you paste in the entire song's lyrics, it's smart enough to know which words you're already sang, and will pick them up in the second verse, for instance.

When I put in a dense, full developed song, it might not change it a lot. But it does deconstruct it and rebuild it, so parts that have a guitar melody might also be played without the melody, or with a melody with slight changes during other parts. The weirdest thing is hearing my own voice sing my own lyrics, but at times in different ways. Or repeat a phrase over and in a different way, to make the chorus a little different, for instance—it's not locked into only saying that words the same number of times that the lyrics dictate. In other words, as if a producer said, "why don't you try..."

I've also put in a theme on piano, and it gave me back some variations, at times expanded into a pretty reasonable piano piece that would certainly stand on its own for background to a video or presentation. That's the place that seems up for grabs by AI right away. Instead of licensing some music from stock music, why not pay a subscription to an AI site that you can feed a basic theme that you have in mind for a video backdrop, and have it complete and "record" the tune for you. Naturally, less music-adept folks would just describe what they want, until they get something they like.

The point is, this is already happening in a big way with photos for corporate presentations. Why go to Shutterstock and search for a picture, when you can say, "generate an image that conveys achievement of goals, for young adults"..."add sunshine"...
 
Back
Top Bottom