Hacker News Reader

What do you think the AI could do that you couldn't by searching the web?

And is the difference between the hypothetical bedroom-produced version of the same Radiohead album and the actual album actually that noticeable?

I work in film/video, and you definitely can produce something that would have cost millions of dollars 10 years ago by yourself, with minimal costs. I struggle to compare the scale of engineering needed to make albums versus movies.

> What do you think the AI could do that you couldn't by searching the web?

Create enough hours in the day to do the work.

No amount of skill or knowledge will get me over the hump that it takes hundreds, maybe thousands, of person hours to produce a Radiohead album.

I'm sorry - hours of doing what exactly?

Ok computer took two months of mixing, and some time for the mastering. 80% of it were live takes. What's the AI supposed to do?

The 2 months of mixing

I don't know what to tell you brother, two months of mixing is very doable for humans beings... As it was done by two people to make OK Computer? I don't think having an AI will change much for your music production if 2 months of mixing is too much work. Don't even know what you'd ask it to do - you'd probably have to readjust it until it's the way you want it...it would probably take the same time don't you think?

So we’re talking about democratizing production yeah?

Imagine I’m a schmuck who can croon into a microphone. It’s just me. I have a dayjob because this is a hobby.

“Just a quick 2 months of work” is a total nonstarter. I’ll just keep crooning in the shower

“Just 2 months, it’s not a lot” is like telling someone “Oh cleaning the house is easy, just ask your maid”

Well we're talking about making a record like OK Computer. It's a work of art, it takes time and intent and attention to details.

If you want to record a couple jazz songs for kicks you can very easily and satisfactorily do this today, for very little money (just higher up this thread people have recommended great tools) and it will be higher quality that what was playing on the radio in the 90s. And I'd argue that if you were as good a crooner as Radiohead are -eh- experimental musicians, I'm sure you'd find the time to make an album that would make yourself rich.

Sorry I'm a little ranty - gonna get a drink I guess

You don't need two months of mixing for your crooner album. You can mix it yourself on the weekend and be 80% there.

Like the guy that did Tubular Bells keep re-recording it because it's not good enough for him - but I'm fine with the version he first produced. AI isn't going to change anything here.

I get it, but I see so many possibilities here. Fitter Happier and Climbing up the Walls have string arrangements performed by an orchestra. Can you imagine writing a score and spinning up a virtual orchestra to play it? There are so many subtleties in phrasing, dynamics, and intonation that make the difference between just Ok and excellent. It doesn’t seem impossible to imagine giving your score to your virtual musicians and communicating with them in natural language and hummed phrases in an iterative process until they get things just right. You’d have to agree this would be democratizing, right?

I guess it is. Kind of like being able to talk to an AI and have it make a movie out of it. Thanks for the chat!

A lot of time and attention can still be spent with 2 months if AI can do a lot of it.

OKC is one of the best Albums of all time imo. But it’s also one of the most popular ones in my own list. So not like all the other works of art for music I come across requiring such intense amt of human labor put in. The capital wouldn’t be there. Capital means capitalism. Which isn’t democratic.

Reading on the recording of OK Computer, "about 80%" of it was live, no audio separation, it took two months to mix, I can't find the time it took to master. I don't know where the thousands of man-hours were spent?

OK Computer ended up with 12 songs, although OK Computer OKNOTOK 1997 2017, the 2017 reissue of OK Computer included 3 extra previously unreleased tracks (plus a bunch more extras). If we assume those 3 tracks were mixed and mastered for OK Computer's original release, but just didn't make the cut for whatever reason, we have 15 tracks.

Thousands of man-hours is the five members of the band, plus their producer, plus an audio-engineer working 40 hrs/wk.

There are 43 business days in 2 months, or just under 3 days per track. 3 days to do a bunch of technical work to balance the recordings juuust right, and then balance those recordings between each other to make it worthy of being called a Radiohead album. Personally, 3 days per track actually doesn't sound like enough time for that since you mix it, and then have to listen to that particular mix until you get bored, and then come back to it after, listen to it again some more, figure out how you feel about that mix, and then repeat.

Of course, you can half-ass it and throw the tracks into Ableton, spend 20 minutes and call it a day, but you didn't used to get to be Radiohead with that kind of effort. ML algos on the horizon seem like they'll be able to do just that.

If production gets easier, I’m guessing the competitive bar will raise to a point where you still can’t do this with little effort. There may be an adjustment period where you can, but I think it will be temporary. An arms race will kick off with the new tools.

You don't need AI for this, you can do it now. Any analog music production gear in existence has a software emulation that's 99.99% accurate and available for a few hundred bucks at most (and of course can be easily acquired for free, as with all software). High quality audio interfaces are available for cheap. Most expensive thing would be renting out a studio to record vocals with a good mic, and even that is arguably unnecessary. Plenty of in-the-box musicians and producers are making top quality songs with not much more than their Macbook.

I’ve recorded at home with a similar setup. Getting things to sound just right takes a lot of effort and tinkering though which I just don’t have time for due to the demands of work and family. I might have a couple hours per week to devote to music. If I could go from writing a song to making a professional sounding recording in a couple hours, that would be amazing. This isn’t a hypothetical either. I’m an ML researcher and plan to experiment with stuff like this when I have time. Making it stupid easy to produce professional sounding recordings would absolutely be democratizing in my opinion.

Already lots of magic tooling for exactly this, some of which (allegedly) uses AI. Look into Gullfoss for AI assistance with mixing and iZotope Ozone which has an "AI mastering assistant." And there's even more basic "one knob sound good" plugins of varying effectiveness, like OTT, Sausage Fattener and Soundgoodizer.

Thanks for the recommendations. All of my music equipment is in storage now, but I hope to set up a small studio later this year. Will definitely check this stuff out.

The gear has indeed gotten cheaper, but you need to

- record one or more sounds

- have extensive knowledge about filters, mixing, mastering

Without the latter, even if you're amazing musician/singer and you have decent gear and a nice interace and good ideas, you still need to sit down and mix the thing so that it sounds good in the car/airpods/phone/laptop.

> Plenty of in-the-box musicians and producers are making top quality songs with not much more than their Macbook.

I'll bet you money that if one of their songs has repeat status in your music player, that song is likely not the first song they've ever recorded.

They had to build mechanical skills first. AI-generated music skips that.

Raydev appears to have no interest in learning anything about the music production process nor putting in any work. I suggest that they are using "expensive gear" as a proxy for "effort, work, sweat, time"

mechanical skills are EARNED, and are a lot of fun to exercise when functioning.

*before anyone suggests "ableism" I refer you to Django Reinhardt, the drummer from Def Leppard, and anyone who has overcome physical disabilities to express themselves upon the musical instrument of their choice.

One of the highest selling songs of the decade, Old Town Road, was made by some kid on a laptop in his bedroom. He has outsold Radiohead, the song went 16x platinum. Production value means nothing in the success of music, even some of the most successful major artists with major label backing put out badly mixed garbage and still outsell Radiohead. Music is like code: consumers don’t care about the input.

https://en.m.wikipedia.org/wiki/Old_Town_Road

"democratization"!? It wouldn't be YOU or another PERSON gaining these capacities, in such a scenario, it would be the tool, and you would be the toolmakers customer.

ATM, LLM's of reasonable sophistication for such tasks are not running on your laptop, and at the point they do, something else will be getting all the limelight, and it won't be within your reach.

As a lifelong musician and composer and producer, it's just automating things away that I don't want automated away. Producing music is about making decisions.

You act as if Radiohead just inserted money and out came pro music, but really you appear to be wishing to skip any hard work and appear to be asking for instant gratification. Well, I can tell you, as someone who could readily train AI in music, that you only cheat yourself by cutting corners. Knowledge is power to do what you actually really deeply want to. Handing off the minute decisions that add up to making a masterpiece = cheating oneself of being a creator.

Somebody's about to become like illustrators and copywriters — mostly irrelevant unless you are the very best at what you do.