The most stunning Beach Boys AI I've heard yet

[guitarfool2002]

I don't think (or actually see where) anyone is looking for a theoretical or hypothetical endorsement from Brian on using AI in these discussions. The question was asking opinions if he would have used it had it been available 60-70 years ago, and in my opinion the answer would be a resounding yes, as would be the same answer if Brian were replaced by Lennon, Emerick, Townshend, Page, or any artist or audio engineer of note from the same era. How they would use it is one of the parameters, whether it would end up on released commercial material, if they would use it for demo purposes, or if they would simply tinker with it for fun...bottom line still being if they would use it or at least try it. How is that shifting the goalposts when it's a direct answer to the question which was open-ended and pure fantasy hypothetical to begin with? The actual facts and history we have of those artists from that era looking to new technology to expand their sonic toolbox and palette to create new sounds would hypothetically suggest they would of course try out new technology to further their art because they have a history of doing exactly that, and unless they were anti-tech Luddites who chose to record live to 2-track tape (or even to single-track mono disc) in eternity for whatever reason, they would try new technology if it would enhance their music.

[bonnevillemariner]

Had AI been around in the 60s, might Brian have considered using it?

Ok, I'll ask another question: If Brian was to listen to something Dae Lims made-- say his AI SMiLe-- what might his reaction be?

[HeyJude]

Had AI been around in the 60s, might Brian have considered using it?

Ok, I'll ask another question: If Brian was to listen to something Dae Lims made-- say his AI SMiLe-- what might his reaction be?

I dunno, but maybe it would be "Why? Didn't I finish the album in 2004, and didn't we also put out a Beach Boys version several years later?"

I'll have more to write when I get to my previously-mentioned thoughts on my recent re-listen to a bunch of that AI vocal stuff, but I will say that the stuff works much more as a "squint and imagine" exercise as opposed to a crystal clear, full-blown, high-fidelity detailed listening session.

I think Brian would be able to tell these things don't sound like him (or Carl, or Mike, etc.), and the question would be more whether he would be a nice, polite guy and say "Interesting, cool, thanks for your enthusiasm" or "Wow, that kind of sounds somewhat like us", or just blurt out that it doesn't sound like the Beach Boys individually or collectively.

[wavesoflove]

That's a great point and element in the process which I totally missed: Add every songwriter and songwriting team to the discussion, and if those writers were contracted to write new material for any name artists, whether it be Elvis, Sinatra, or whoever...they could demo a song with an approximation of the target artist's voice on the demo and present that demo to the artist, managers, producer, A&R, etc and have it sound close to what the song would be as sung by that artist.

Yes in that case I can absolutely see AI voice replication technology being used and embraced in those scenarios. Sometimes (or many times) the voice and the sound of that voice literally makes a song and a record greater than the song itself.

Imagine if the Colonel had got hold of this technology in the 1960's. We'd probably have twice as many bad Elvis movies, and i'm sure he would have exploited it even more once the King died.
There'd probably be a new Elvis album out every year til the mid 90's.

Regardless of your stance on ai covers, I assume most of us agree there needs to be some serious discussion on regulation with this tech.
Yes, it's a tool, but scammers are going to go wild with it, and that's whats most disturbing for me.

[guitarfool2002]

Had AI been around in the 60s, might Brian have considered using it?

Ok, I'll ask another question: If Brian was to listen to something Dae Lims made-- say his AI SMiLe-- what might his reaction be?

I think Smile may not be the best example to choose. To Brian, he finished Smile with Van Dyke in 2004 and they presented it to the world as both a fully-realized complete live piece and then a subsequent studio album of that same piece of work. It's done in his mind, and hearing newer interpretations of it probably wouldn't carry much weight other than to appreciate the musicality and execution of the work. And Smile still carries a lot of very heavy and personal emotions for Brian, who knows what feelings that would trigger in the man.

A better example might be the version of "Still I Dream Of It" that was done with an AI Frank Sinatra singing the lead over the orchestration. That vocal specifically is something Brian envisioned for the song originally, something he tried to make happen and wanted to happen but it sadly never did, and cannot happen in reality. I think an AI project like that - going on the making the impossible a reality mindset with a lot of imagination mixed in - is something he would be more interested in hearing because it's something he actually heard in his own mind when he was writing the song but something that can never happen physically. That kind of ethic is the best of this particular use of AI technology because it's 100% rooted in reality.

[Dae Lims]

I don't think (or actually see where) anyone is looking for a theoretical or hypothetical endorsement from Brian on using AI in these discussions. The question was asking opinions if he would have used it had it been available 60-70 years ago, and in my opinion the answer would be a resounding yes, as would be the same answer if Brian were replaced by Lennon, Emerick, Townshend, Page, or any artist or audio engineer of note from the same era. How they would use it is one of the parameters, whether it would end up on released commercial material, if they would use it for demo purposes, or if they would simply tinker with it for fun...bottom line still being if they would use it or at least try it. How is that shifting the goalposts when it's a direct answer to the question which was open-ended and pure fantasy hypothetical to begin with? The actual facts and history we have of those artists from that era looking to new technology to expand their sonic toolbox and palette to create new sounds would hypothetically suggest they would of course try out new technology to further their art because they have a history of doing exactly that, and unless they were anti-tech Luddites who chose to record live to 2-track tape (or even to single-track mono disc) in eternity for whatever reason, they would try new technology if it would enhance their music.

Aside from doing the Beach Boys AI stuff as fun hypothetical scenarios, I am an active producer/songwriter in the industry. I use AI to demo songs for female artists. I'll also use it if a singer's (or my own) tone isn't quite nailing it for a demo. And it's not just me, lots of my peers are using it as well. I'd say it's totally possible BW and the likes would've used AI if it were around in the 60's. There's also such thing as "hybrid models", where for instance Brian could've combined his voice with, say, Mike's to get a punchier more nasal tone if he wanted.

[rab2591]

Yes in that case I can absolutely see AI voice replication technology being used and embraced in those scenarios. Sometimes (or many times) the voice and the sound of that voice literally makes a song and a record greater than the song itself.

Imagine if the Colonel had got hold of this technology in the 1960's. We'd probably have twice as many bad Elvis movies, and i'm sure he would have exploited it even more once the King died.
There'd probably be a new Elvis album out every year til the mid 90's.

Regardless of your stance on ai covers, I assume most of us agree there needs to be some serious discussion on regulation with this tech.
Yes, it's a tool, but scammers are going to go wild with it, and that's whats most disturbing for me.

The applications for this technology are so far ranging it's incomprehensible. Imagine wars being fought with AI computers that can calculate potential outcomes of battles by using information about Generals/culture/motivations/soldier-strength/etc - and creating battle plans with that knowledge. Imagine the technology where all the cameras placed at every stoplight in America can keep track of any person/car without human interaction. Imagine an AI that can take the sum of a person's entire web-browsing history and use it to create a psychological makeup of that person in milliseconds, and then using that information to help predict a person's actions/whereabouts/etc.

On the culture side, we've already had an AI track (maybe more than one) hit the music charts (and these tracks were not created with approval from the artist the AI was replicating). How many of us look at a video now and say "hmmm something isn't right with that" (look up the Sora video creator), or hear a narration on a Youtube video and it sounds soulless (because it's AI narrated)? Or how about the AI generated "art" that Adobe and other companies are promoting? While in practice it is a neat thing to mess around with, the end product just feels void of soul - there is something missing.

One thing I like about Dae Lims work is that there is such a human presence to it. He isn't typing a prompt into ChatGPT asking for a song to be created out of thin air, he is meticulously working at these tracks, and it shows. His work is one of the few things I actually like from this new wave of modern AI tech.

[bonnevillemariner]

I dunno, but maybe it would be "Why? Didn't I finish the album in 2004, and didn't we also put out a Beach Boys version several years later?"

I'll have more to write when I get to my previously-mentioned thoughts on my recent re-listen to a bunch of that AI vocal stuff, but I will say that the stuff works much more as a "squint and imagine" exercise as opposed to a crystal clear, full-blown, high-fidelity detailed listening session.

I think Brian would be able to tell these things don't sound like him (or Carl, or Mike, etc.), and the question would be more whether he would be a nice, polite guy and say "Interesting, cool, thanks for your enthusiasm" or "Wow, that kind of sounds somewhat like us", or just blurt out that it doesn't sound like the Beach Boys individually or collectively.

I regret using Smile as an example in my question because both responses to it so far have focused more on my use of Smile as an example (it happened to have been the latest Dae Lims thing I listened to) than the actual question I'm asking. Let's forget I used that example. And I won't provide another one because that will get parsed to hell too.

[bonnevillemariner]

One thing I like about Dae Lims work is that there is such a human presence to it. He isn't typing a prompt into ChatGPT asking for a song to be created out of thin air, he is meticulously working at these tracks, and it shows. His work is one of the few things I actually like from this new wave of modern AI tech.

Exactly. There's some other cat out there dabbling in this with Beach Boys music and I couldn't get 30 seconds into the first track. When it comes to most things AI-- from writing to art-- I'm solidly in HeyJude's camp. It's Dae Lims' love for, and deep grasp of, the music and the band AND his musical abilities that make his stuff worthwhile to me.

[HeyJude]

I dunno, but maybe it would be "Why? Didn't I finish the album in 2004, and didn't we also put out a Beach Boys version several years later?"

I'll have more to write when I get to my previously-mentioned thoughts on my recent re-listen to a bunch of that AI vocal stuff, but I will say that the stuff works much more as a "squint and imagine" exercise as opposed to a crystal clear, full-blown, high-fidelity detailed listening session.

I think Brian would be able to tell these things don't sound like him (or Carl, or Mike, etc.), and the question would be more whether he would be a nice, polite guy and say "Interesting, cool, thanks for your enthusiasm" or "Wow, that kind of sounds somewhat like us", or just blurt out that it doesn't sound like the Beach Boys individually or collectively.

I regret using Smile as an example in my question because both responses to it so far have focused more on my use of Smile as an example (it happened to have been the latest Dae Lims thing I listened to) than the actual question I'm asking. Let's forget I used that example. And I won't provide another one because that will get parsed to hell too.

No, I totally get that you didn't mean to single out "Smile" and that you were just using it as a random example. I don't want it to come across as if I harped too much on specifically "Smile"; I think we all understand how that particular example would be unique in some ways.

But, if we're talking about any of the extant examples, I do think I'd fall back on what I said earlier:

I think Brian would be able to tell these things don't sound like him (or Carl, or Mike, etc.), and the question would be more whether he would be a nice, polite guy and say "Interesting, cool, thanks for your enthusiasm" or "Wow, that kind of sounds somewhat like us", or just blurt out that it doesn't sound like the Beach Boys individually or collectively.

[HeyJude]

This is an interesting little interview clip with Sheryl Crow. Listen to her describe when she's talking to a female writer who needed a male voice for one her demos, who paid a service $5 to create a fake AI John Mayer voice to sing her demo.

https://www.youtube.com/watch?v=T5DX5cB_m34

If you wonder what many, if not most (surely not all, but probably most) artists think of doing this, watch and listen to her reaction.

I'm not saying every application of this is the exact same scenario. But in scenarios like this, it's altering some fundamental aspects of how music is created, and I think artists that are sensitive to humans being the driving force behind that process are not embracing of using fake AI voices.

[rab2591]

In regards to if Brian would dig hearing these tracks, here is a quote from FatheroftheMan (who did the popular Bookoff Smile AI mix) that kinda relates to this:

I have a finished version of On A Holiday, with Mike and Al sharing the lead. On the middle section, the pirate rap, I had van dyke parks beat rapping the part I was so excited about it and couldn't wait to share.

I was encouraged to send it to the man himself, and I was excited to receive a reply. I told him that I wanted to make something special to celebrate my love of BWPS and the original tracks for the 20th anniversary. I made sure to let him know that if he didn't like it, I wouldn't release it as is.

Very respectfully, he responded that he didn't feel the track honored the legacy and that I should do something else with my talent (paraphrasing). He was very kind and I want to stress that I don't feel anything negative towards his response.

This stuff is cool for some of us as fans, but I think it's probably just plain weird to those whose voices are being replicated. In FOTM's defense, VDPs has been very vocal about not liking the sound of his own voice, so it could've been that and not the quality or idea behind the mix.

Brian was called "dog ears" by Mike because Brian could/can pick up on everything. He would probably instantly be able to tell it is fake - I'd imagine he'd really get a kick out of the idea of it, but would probably be instantly turned off by the sound of the fake voices. Many of us can look past the digital artifacts and vocal flubs heard in most of these mixes, but I think Brian would be less forgiving about that stuff.

[guitarfool2002]

We can't make blanket statements or assumptions about how artists feel overall about the use of AI voice modeling in these ways, I haven't seen and don't think there is a consensus that exists among artists and their opinions on and use of this tech. Each artist is obviously different and has different opinions of this tech. I'm sure if there are some that are totally against it, there are thousands more who are using it in some way. Suppose a singer writes a song with a specific legendary voice in mind to sing it, similar to Brian with Sinatra. Let's say that person thought of their song as something Marvin Gaye would sing - If the tech is available, and they know how to use or access it, I'm sure they would use the tech to put Marvin's voice on their song, even just to hear what it would sound like. Then consider the thousands of other legends that could be plugged into the equation, and how many musicians would want to hear what such a voice sounded like singing a song that they directly inspired.

That's where another separation comes in: Is it done for personal use or a love for the music, is it done for a demo purpose, or is it done for purely commercial reasons? Obviously if an artist is still alive and working, they would not want their voice used on a commercial project without their 100% consent, that's a no-brainer and I don't think any of what we're discussing crosses that line because it's not a project for sale. But then again, if random fans start using the voices of famous artists in non-commercial ways, and not selling the works for profit, who can control that until it crosses a line of either trademark/copyright or standards of decency in a legal sense? Anyone can try to make their own McDonalds Big Mac at home using recipes found online, and even serve it to people they know...as long as they don't sell it commercially. Anyone can make a poster using a famous copyrighted image or photo, and hang it on their wall...as long as they don't sell it commercially. All of the landmark court cases regarding home recording of copyrighted media, whether it was the Betamax case for video or the blank cassette suits for audio, came down to a person's ability and right to copy and use copyrighted material...as long as it's not sold commercially.

I'm just going out on a limb and suggesting there are probably a good number of artists working in the music business who have used AI in the ways we're discussing, and perhaps mostly for fun or to see what their songs would sound like if another artist was singing them. That's far different from Taylor Swift releasing a song that features an AI version of Prince singing a duet with her and doing it for profit, and not securing rights from Prince's estate before doing so.

Perhaps in the near future there will be court cases regarding AI similar to those regarding sampling in the 90's that literally changed the music business, or the Betamax cases which established the guidelines around home recording and copying of owned material in the 70's and 80's.

[guitarfool2002]

This is an interesting little interview clip with Sheryl Crow. Listen to her describe when she's talking to a female writer who needed a male voice for one her demos, who paid a service $5 to create a fake AI John Mayer voice to sing her demo.

https://www.youtube.com/watch?v=T5DX5cB_m34

If you wonder what many, if not most (surely not all, but probably most) artists think of doing this, watch and listen to her reaction.

I'm not saying ever application of this is the exact same scenario. But in scenarios like this, it's altering some fundamental aspects of how music is created, and I think artists that are sensitive to humans being the driving force behind that process are not embracing of using fake AI voices.

If you want an example of something that has fundamentally changed how music is created, one of many no less, look no further than digital recording technology. It has changed nearly every aspect of how music is recorded and presented to listeners. Modern listeners are now conditioned to hear things which are often physically impossible due to the ease of editing, fixing, and correcting elements of the music. The notion of recording a song with a "human" feel or with imperfections is all but gone from popular music, unless there is an artist with an analog fetish who only records as if it were 60 years ago in an old studio. Drummers sound perfect thanks to beat mapping, singers hit every note perfectly thanks to Melodyne and other pitch correction software, guitar and bass parts are not only played perfectly in time but they're sterilized by editing out all the squeaks and noises inherent in the instrument itself...the list goes on. It's at the point where many younger listeners are so conditioned by this sterilization and fixing that something played 100% real and live can sound "off" or wrong, and that also has removed the notion of a real groove or a pocket from the process, since a beat mapping software will take all of the ebb and flow and variation out of a drum part in order to quantize it to sound perfect.

So based on that, and all of that too is what Sheryl Crow has used on her own music I'm sure, again the question is how is AI as a new tech any different? We're already hearing a majority of pop music that has been altered and sterilized by existing technology to make it sound different than when it was recorded. Some including me would say the same radical shift in the creation of music happened when audio engineers learned how to edit tape seamlessly. It created a product that was not a pure or true reproduction of a musical performance, and also led to some of the greatest popular music ever recorded.

And don't get me wrong, of course I share similar concerns about the AI technology and where or how far it could go, but ultimately it's a tool just like digital recording, sequencing, editing, and quantizing is a tool that has fundamentally altered the process of making and recording music, and shows no signs of disappearing.

[rab2591]

So based on that, and all of that too is what Sheryl Crow has used on her own music I'm sure, again the question is how is AI as a new tech any different? We're already hearing a majority of pop music that has been altered and sterilized by existing technology to make it sound different than when it was recorded. Some including me would say the same radical shift in the creation of music happened when audio engineers learned how to edit tape seamlessly. It created a product that was not a pure or true reproduction of a musical performance, and also led to some of the greatest popular music ever recorded.

And don't get me wrong, of course I share similar concerns about the AI technology and where or how far it could go, but ultimately it's a tool just like digital recording, sequencing, editing, and quantizing is a tool that has fundamentally altered the process of making and recording music, and shows no signs of disappearing.

I really liked her line "the thing that creates art is the human experience, not a computer's experience" but then I, too, immediately thought about how literally everything in modern music is done with the aid of computers...And most egregiously the use of fiddling with someone's vocal to make them "pitch perfect". So are we actually hearing a 100% real voice anymore? Most likely not, when it comes to any top 40 hit. We are hearing someone's voice being filtered through a computer with pitch correction. It is 95% their voice, but it's not a 100% real vocal anymore - and it's because of a computer.

Also, in the context of AI, when I first heard the AI 'Thank Him' or the coda on the AI 'Holy Man' it felt like when I first heard 'Wouldn't It Be Nice' in stereo for the first time. And I'm not exaggerating that - it was that same endorphin rush when I hear a gem from the real Beach Boys for the very first time. And this is likely because I know that 'Thank Him' and 'Holy Man' are real songs and Dae Lims is just cleaning them up (by replicating vocals) and adding some Beach Boys-like harmonies. But the fact that a computer helped make that and my brain enjoyed it, that experience cannot be discounted. So as much as I want to agree with Sheryl (because I think at heart her quote above is very true), I also can't ignore that my brain felt a meaningful connection with a fake Brian Wilson vocal.

What a crazy time to be alive.

[HeyJude]

I can only speak for myself, but I don't see this as such a hazy sort of situation.

I just don't subscribe to the slippery slope that the invention of the microphone or tape splicing or outboard effects are the same thing as artificially generated vocals.

This about the intent and actual actions on the artist's end, and then on the listener end it's a matter of what we know we're listening to.

I think the vast, vast majority of people who CONJURE this stuff up, who write it and then workshop it and perform it themselves, have the reaction Sheryl Crow had in that video. You kind of either get it or you don't. Yes, artists are often sensitive people. But if you are one, or if know one, or even if you just study one (which I think BB/Brian fans do), this should be clear, and is something that by and large is a virtue.

Just like people who actually draw/paint, etc. artwork are rightfully put off by the lazy people who type a description into an AI art program trained on a MILLION PEOPLES' ARTWORK, and then spit out their "masterpiece."

People can come up with exceptions and scenarios where some people are recording things so inorganically that everything is augmented by software or effects, etc. But at the end of the day, Brian Wilson (or whomever) singing a vocal and then autotuning it up the wazoo and adding 97 plug-in effects is still a *human being* driving the task.

Somebody dumping a thousand examples of Brian vocals into a computer, and through machine learning crafting a fake Brian voice, and then singing to guide that artificial voice; that is not the same thing. Saying it's the same, or anywhere near the same as splicing tape or the advent of digital recording is, to me, like saying orange marmalade and orange house paint are the same, or similar, or that one is just a progression of the other.

[rab2591]

Just like people who actually draw/paint, etc. artwork are rightfully put off by the lazy people who type a description into an AI art program trained on a MILLION PEOPLES' ARTWORK, and then spit out their "masterpiece."

This line got me thinking about if AI had been used to restore classic works of art. And after looking it up, it has.

What Dae Lims is doing is not typing a description in and a computer is spitting out the result. Rather, it is what you say "somebody dumping a thousand examples of Brian vocals into a computer, and through machine learning crafting a fake Brian voice, and then singing to guide that artificial voice"

Using the example of 'Thank Him': that is Dae Lims using this technology to "restore" a very weathered piece of art. And just like restoring a painting (where the strokes are not real but simulated), Dae Lims is using the "strokes" and "color schemes" of Brian's voice to restore the vocal. It's not real, but I don't see why it's "absurd" to enjoy the result of this.

Are we to look at a piece of fine art, restored with AI, and not find beauty in it? Is it not still art? The intent behind the art is still primarily with the original artist, a computer is merely helping us see something closer to the artist's original piece. Or lets say someone, out of sheer curiosity, wanted to see 'Starry Night' by Van Gogh, only done more in the style of his 1886 work 'Le Moulin de Blute-Fin' - would that be absurd if people enjoyed the AI result?

I don't see why it's any different for the music of The Beach Boys. Again, your initial argument is that no one should be looking to AI Beach Boys to hear "more" material when there are thousands of hours of real Beach Boys music yet to be discovered. But if someone wants to hear a "restored" version of 'Thank Him', why is this more absurd than listening to the static-sounding real demo? I think they both have a place in The Beach Boys world - and clearly many people on Youtube and on the forums agree.

[HeyJude]

Just like people who actually draw/paint, etc. artwork are rightfully put off by the lazy people who type a description into an AI art program trained on a MILLION PEOPLES' ARTWORK, and then spit out their "masterpiece."

This line got me thinking about if AI had been used to restore classic works of art. And after looking it up, it has.

What Dae Lims is doing is not typing a description in and a computer is spitting out the result. Rather, it is what you say "somebody dumping a thousand examples of Brian vocals into a computer, and through machine learning crafting a fake Brian voice, and then singing to guide that artificial voice"

Using the example of 'Thank Him': that is Dae Lims using this technology to "restore" a very weathered piece of art. And just like restoring a painting (where the strokes are not real but simulated), Dae Lims is using the "strokes" and "color schemes" of Brian's voice to restore the vocal. It's not real, but I don't see why it's "absurd" to enjoy the result of this.

Are we to look at a piece of fine art, restored with AI, and not find beauty in it? Is it not still art? The intent behind the art is still primarily with the original artist, a computer is merely helping us see something closer to the artist's original piece. Or lets say someone, out of sheer curiosity, wanted to see 'Starry Night' by Van Gogh, only done more in the style of his 1886 work 'Le Moulin de Blute-Fin' - would that be absurd if people enjoyed the AI result?

I don't see why it's any different for the music of The Beach Boys. Again, your initial argument is that no one should be looking to AI Beach Boys to hear "more" material when there are thousands of hours of real Beach Boys music yet to be discovered. But if someone wants to hear a "restored" version of 'Thank Him', why is this more absurd than listening to the static-sounding real demo? I think they both have a place in The Beach Boys world - and clearly many people on Youtube and on the forums agree.

I'm not like a painting/art aficionado, but from everything I've seen, painters/illustrators and other visual artists are emphatically against using AI *in any way*, including restoring or "completing" pieces. Like, exponentially more than even musicians/singers from what I've seen. Probably because, right now, it's easier for some schlub to type stuff into a prompt and spit a "finished" piece out.

If you want like an A+, pristine example of "MISSING THE POINT", check this out:

https://www.smithsonianmag.com/smart-news/keith-haring-painting-artificial-intelligence-180983563/

A short excerpt:

The year before he died of AIDS-related complications, the artist Keith Haring created a unique work known only as Unfinished Painting (1989). In its upper-left quadrant, black and white lines form stylized patterns on a purple background. Streaks of purple paint trickle down onto the otherwise empty lower-left quadrant; the right half of the canvas is also blank. Haring intentionally left the work unfinished as a commentary on the AIDS crisis.

Now, a newly “completed” version of the work—made with the help of artificial intelligence—is generating controversy: A social media user employed an A.I. image generator to expand Haring’s designs across the blank sections of the canvas, ultimately posting the altered image on X, formerly known as Twitter.

I'm sorry, there's just not a way to draw a particularly good non-AI analogy to using AI. And really, with visual art, it's even easier to say a hard "no" to it. A lot of that has to do with not being able to trust the people using the software, and that's not even really an aspect of the vocal AI stuff that I've even delved into.

There are ways to "restore" "Thank Him" without generating any new artificial voices or instruments. Indeed, Peter Jackson is rumored to be considering or possibly already working on using his machine-learning extraction/separation technology to clean up the Beatles' Star Club tapes. I think there's a limit to how much improvement that will offer, but that is a *perfect* use of the technology as I've often said.

Much like there are people who develop the "skill set" to finagle and massage the AI prompt process to get less messy results, there is no doubt potentially skill involved in making a fake AI Brian Wilson vocal. I've never said otherwise. But it's still also a computer algorithm doing a TON of the work, and it's that algorithm that is making that fake voice. No human being has a port where thousands of examples of Brian Wilson's recordings can be loaded, processed, and then a "Brian Wilson" voice can come out of your mouth.

Yes, some people who sing to guide their AI vocals are better at it. I've clicked on a number of "Paul McCartney" AI vocals that sound like Borat run through a McCartney filter, because the person singing it can't sing the right way, and/or they're not massaging the AI processing enough to make it sound less like themselves and more like whomever they're trying to sound like.

The cream of the crop Brian/Beach Boys examples we've been talking about are far, far less problematic than most of the stuff out there, no doubt both because it's made by someone who knows the proper context to start from, and no doubt because they have much more experience using the software than others. I'm *guessing* they probably also have a relatively unfettered "American" accent singing voice.

And yet, what I'm hearing is still an artificial program generating it, and as I've mentioned (and I'll go into in more detail as promised after my recent re-listen session to the AI stuff) I think even the best stuff still just doesn't sound *just like* Brian, or whomever it is in question. Some of them sound closer than others. With some, it's a uncanny valley thing where a little run of syllables sounds pretty accurate and then some sort of inflection is completely off, and in other cases it's just not even close across the board.

[Dae Lims]

One thing I like about Dae Lims work is that there is such a human presence to it. He isn't typing a prompt into ChatGPT asking for a song to be created out of thin air, he is meticulously working at these tracks, and it shows. His work is one of the few things I actually like from this new wave of modern AI tech.

Exactly. There's some other cat out there dabbling in this with Beach Boys music and I couldn't get 30 seconds into the first track. When it comes to most things AI-- from writing to art-- I'm solidly in HeyJude's camp. It's Dae Lims' love for, and deep grasp of, the music and the band AND his musical abilities that make his stuff worthwhile to me.

Very kind of you to say, I'm glad you enjoy the stuff. It's most definitely a labor of love. Appreciate all the kind comments.

[rab2591]

I'm not like a painting/art aficionado, but from everything I've seen, painters/illustrators and other visual artists are emphatically against using AI *in any way*, including restoring or "completing" pieces. Like, exponentially more than even musicians/singers from what I've seen. Probably because, right now, it's easier for some schlub to type stuff into a prompt and spit a "finished" piece out.

Even if that is the case, it doesn't detract from the idea that people (moreso your average human) can find enjoyment and beauty in a work of art that was restored using AI. Which was my point. My point isn't that some people are adamantly opposed to AI - I myself work in the design field (graphic design), and can see how AI will be negatively affecting my job in the not-too-distant future - and I loathe how companies like Adobe are basically casting graphic designers aside (while using their work) for AI profit - and yet I can enjoy a painting that was restored using AI (especially if most of the original creator's work was already present in the piece). I can also see why actual painters would be pissed off at this technology, but again, that doesn't detract from the fact that some people can still look upon an AI restored painting and enjoy it.

If you want like an A+, pristine example of "MISSING THE POINT", check this out:

https://www.smithsonianmag.com/smart-news/keith-haring-painting-artificial-intelligence-180983563/

This is someone who took a work of art that had a clear message, and they disrespected that clear message by "completing" it. It was not a restoration. And it seemingly wasn't done by someone with a clear love or knowledge of the original artist. Whereas stuff created by Dae Lims is made with the knowledge/love of this band.

I could bring up the painting Ecce Homo as an example of why humans shouldn't restore art, but clearly there are good and bad ways to restore art with the human hand. Likewise, because one person, in poor taste, "finished" a painting via AI that was meant to be left unfinished, it does not mean we should cast away the entire AI restoration process.

I'm sorry, there's just not a way to draw a particularly good non-AI analogy to using AI. And really, with visual art, it's even easier to say a hard "no" to it. A lot of that has to do with not being able to trust the people using the software, and that's not even really an aspect of the vocal AI stuff that I've even delved into.

My point is that someone created software to look at the brush strokes and textures of specific artists, and they used those traits to restore specific paintings, and thus your average human can look upon those works and potentially see something beautiful. If people find it beautiful, who are we to call them (or their feeling of beauty) absurd?

There are ways to "restore" "Thank Him" without generating any new artificial voices or instruments. Indeed, Peter Jackson is rumored to be considering or possibly already working on using his machine-learning extraction/separation technology to clean up the Beatles' Star Club tapes. I think there's a limit to how much improvement that will offer, but that is a *perfect* use of the technology as I've often said.

Yes, there are other ways to restore. But again, my point is that Dae Lims created a "restoration" of sorts of 'Thank Him' and hundreds of people love it (per the likes on Youtube). People are finding this work beautiful - and thus their feelings shouldn't be cast aside as absurd.

Much like there are people who develop the "skill set" to finagle and massage the AI prompt process to get less messy results, there is no doubt potentially skill involved in making a fake AI Brian Wilson vocal. I've never said otherwise. But it's still also a computer algorithm doing a TON of the work, and it's that algorithm that is making that fake voice. No human being has a port where thousands of examples of Brian Wilson's recordings can be loaded, processed, and then a "Brian Wilson" voice can come out of your mouth.

However the voice is created, it is still a sound that is pleasing to many listeners.

Yes, some people who sing to guide their AI vocals are better at it. I've clicked on a number of "Paul McCartney" AI vocals that sound like Borat run through a McCartney filter, because the person singing it can't sing the right way, and/or they're not massaging the AI processing enough to make it sound less like themselves and more like whomever they're trying to sound like.

The cream of the crop Brian/Beach Boys examples we've been talking about are far, far less problematic than most of the stuff out there, no doubt both because it's made by someone who knows the proper context to start from, and no doubt because they have much more experience using the software than others. I'm *guessing* they probably also have a relatively unfettered "American" accent singing voice.

And yet, what I'm hearing is still an artificial program generating it, and as I've mentioned (and I'll go into in more detail as promised after my recent re-listen session to the AI stuff) I think even the best stuff still just doesn't sound *just like* Brian, or whomever it is in question. Some of them sound closer than others. With some, it's a uncanny valley thing where a little run of syllables sounds pretty accurate and then some sort of inflection is completely off, and in other cases it's just not even close across the board.

Mostly agree with all of this. As I've said before, I hear the vocal flubs, I hear the digital processing in all of these tracks. But, using 'Thank Him' as an example again, I can more easily tune out any vocal abnormalities in the AI version than tune out the static/hiss/poor-quality of the original real demo. Thus I greatly enjoy the AI version.

That being said, if I were to be cast away on a desert island and I was given the choice to take either AI 'Thank Him' or the lousy sounding demo track, I would take the demo track every time. But that doesn't detract from my thorough enjoyment of an AI track, either. I do see the necessity/logic behind focusing on the real beach boys. In the last couple days of talking about this subject, I have listened to Dae Lims Smile and a couple other AI tracks, but have mostly been listening to Holland, Song Cycle, BWPS, etc. I'd think most people who listen to this stuff are like that: we find these AI tracks interesting/beautiful, but clearly the real band and the real music are what gravitates us toward this band/music (both real and fake).

[Don Malcolm]

So based on that, and all of that too is what Sheryl Crow has used on her own music I'm sure, again the question is how is AI as a new tech any different? We're already hearing a majority of pop music that has been altered and sterilized by existing technology to make it sound different than when it was recorded. Some including me would say the same radical shift in the creation of music happened when audio engineers learned how to edit tape seamlessly. It created a product that was not a pure or true reproduction of a musical performance, and also led to some of the greatest popular music ever recorded.

And don't get me wrong, of course I share similar concerns about the AI technology and where or how far it could go, but ultimately it's a tool just like digital recording, sequencing, editing, and quantizing is a tool that has fundamentally altered the process of making and recording music, and shows no signs of disappearing.

I really liked her line "the thing that creates art is the human experience, not a computer's experience" but then I, too, immediately thought about how literally everything in modern music is done with the aid of computers...And most egregiously the use of fiddling with someone's vocal to make them "pitch perfect". So are we actually hearing a 100% real voice anymore? Most likely not, when it comes to any top 40 hit. We are hearing someone's voice being filtered through a computer with pitch correction. It is 95% their voice, but it's not a 100% real vocal anymore - and it's because of a computer.

Also, in the context of AI, when I first heard the AI 'Thank Him' or the coda on the AI 'Holy Man' it felt like when I first heard 'Wouldn't It Be Nice' in stereo for the first time. And I'm not exaggerating that - it was that same endorphin rush when I hear a gem from the real Beach Boys for the very first time. And this is likely because I know that 'Thank Him' and 'Holy Man' are real songs and Dae Lims is just cleaning them up (by replicating vocals) and adding some Beach Boys-like harmonies. But the fact that a computer helped make that and my brain enjoyed it, that experience cannot be discounted. So as much as I want to agree with Sheryl (because I think at heart her quote above is very true), I also can't ignore that my brain felt a meaningful connection with a fake Brian Wilson vocal.

What a crazy time to be alive.

Let me second that last thought, Rab! But the key thing here, I think, is that human emotion that Dae Lims is trying to honor and replicate...because for 90+% of us here, there is something in the BBs music that has a talismanic power over those of us who "get it." And it's a kind of conditioned aural response that is, as Adam Marsland discusses in his third installment about PET SOUNDS, linked to our disparate (but highly related) connections with (and yearning for) the spiritual power that can be manifested in music.

When we hear Dae Lims capture a large dollop of that sound in so many myriad reconstructions and re-imaginings of the BBs music, a large portion of us are conditioned to respond to that viscerally; it reaches us despite any intellectual resistance that might apply (and does clearly exist in others). There is a certain style of singing that is uniquely the BBs; other artists have approached it, and we sometimes get a kind of muted variant of such a response from that work. But when Dae Lims hits a certain vein of that singing style, it's closer to the uncut "sonic heroin" that floods into our ears and into our hearts than the human imitators with their watered-down hommages. I've had the same response as you, Rab, and while I can't totally dismiss Jude's skepticism, all I know is that Dae Lims has managed to tap into something special that produces that same emotional response.

And given that the band is never going to be able to do anything again--no new tracks, no reworkings of stillborn or abandoned efforts--and even in light of the fact that there is plenty of real music from the band to listen to, the fact is that the ongoing desire or quest for such a feeling is not "absurd"... it's part of a need for something life-fulfilling, something soothing, particularly in "a crazy time to be alive" when the band is not capable of providing that to us. And if AI can be wielded by someone in order to make that happen, then it's a (Paul Revere & the Raiders reference here!) "good thing."

And I think most folks can tell the difference--I don't get that feeling from everything Dae Lims does, but I get much, much less of it from anyone else attempting to use AI technology to create music. Thus I think we should accede to Jude's basic point, but also agree with GF that as a tool specifically applied to music, AI has potential to become a valuable artistic application in the right hands. And it will be practitioners with the approach exhibited by Dae Lims who represent the best hope for it to be applied with respect and a sense of responsibility so that there will be a clear line of demarcation between "artistic enhancement" and "absurdity."

[rab2591]

Don, thanks much for that reply - you summed up perfectly what I can't really articulate well myself.

[HeyJude]

I don't think anybody much has like overt, purposeful ill intent when making AI vocals. And certainly, I'm very willing to take at their word both "Dae Lims" and those who like the work that it's not done with any ill intent.

That's sometimes part of the problem with some AI stuff, though. Like, how many people using AI to create art out there right now are doing it just to f**k with people, or to ruin anything?

Even the person that completed that poignant painting mentioned in my previous post, it doesn't appear as though they were trying to troll anyone or ruin anything. But they did. As with many cases, it's often unclear even *after* some AI stuff is lambasted, whether the person making it even *understands* the criticisms.

While some might be feigning ignorance, I think some really are just like "I'm made an art too! I'm an artist. I don't understand the problem."

I think the story of *why* people use AI to create art is still being written of course. I think, especially when it comes to visual art, there are a lot of lazy and/or untalented people who really get off on doing something that makes them *feel* like they're creating, like they're talented, like they're artistic. There is also a contingent of tech bros, adjacent to all the NFT stuff and whatnot, that are using AI to run various grifts. There are also some people just tooling around with it.

It's a very messy situation, and as I've often said, we haven't even delved into the broad moral/ethical implications of the stuff on this thread, up to and including a "Skynet/Terminator" sort of scenario. There are also smaller, more practical problems that will surely arise, including people NOT marking this stuff as fake/AI and letting it proliferate.

In the case with this BB stuff we're talking about, it is clearly done with a lot of workshopping and intent, and plenty of self-awareness as to the nature of what they're doing.

So, as I've mentioned in the more distant past, I really don't like bursting in like the Kool Aid man and questioning the work of one of the only people making this stuff who are being as thoughtful as possible about it, labeling it appropriately as AI, etc. But it's some of the only work in the realm of BB stuff that is getting attention, so it's what ends up being the topic of discussion.

[Joel Goldenberg]

From Dae Lims, a superb Good Vibrations AI stereo remix.
https://www.youtube.com/watch?v=kW83O-xC8bQ

[pixletwin]

From Dae Lims, a superb Good Vibrations AI stereo remix.
https://www.youtube.com/watch?v=kW83O-xC8bQ

Tasteful remix. Well done daelims.