The Song Fight Bulletin Board

Paco Del Stinko wrote:I would guess it's very expensive as well, no?

Well, there are different versions of Melodyne, but the first one this is going to appear in is Melodyne Plugin, which is currently going for £220 and I don't think that's unreasonably expensive (it's cheaper than Autotune, which makes me wonder why anyone in their right mind would buy Autotune).

I hope I'm not coming across as a guitar snob, because I'm not and that wasn't my intension or point. I really like what Obz does and I honestly appreciate any serious musicians art form. Just because a person doesn't play guitar, trumpet, etc, doesn't change the fact that I would collaborate with them just as I would any other musician to create music.

But, you can't deny that there are talentless people being launched to extreme fame all the time. I guess it could be argued that their talent is being a cult of personality or sex symbol, etc. But the people with the musical talent behind someone like Ashley Simpson is her band, writers, engineers and producer. Which quite often is a talented musician that doesn't want to be the in the public eye anymore.

Also, I'm not arguing, I'm just stating my opinion. Like JB said, Marylin Manson comes into the public eye with a new version of Sweet Dreams. That is exactly what I mean by music repeats itself. That just happens to be a very exact example. Every time the latest technology gets played out and boring, something new emerges from the depths of technology and new generation of musicians find ways to use it in ways that may not have been the original mindset for the inventor. Like Trent Reznor, the use of the latest technology along with his talents as a song writer, singer and composer, etc. In my original example, I was referring to Reznor as that next musician with talent to rise from the abyss. When I say "true musician", I'm not referring to only people that can play guitar or drums, etc. I'm referring to people with a true musical sense. Maybe that was missed.

BLT wrote:I hope I'm not coming across as a guitar snob, because I'm not and that wasn't my intension or point.

And I hope that you don't think that I'm pointing fingers at you, or anyone else, for that matter. Just trying to join in the convo. Who could call you a snob anyway?

Obscurity wrote:currently going for £220

Doesn't seem outrageous, for what appears to be a powerful tool. Is anyone going to pick it up? I'd be very interested in hearing some first-hand experiences, especially from you smart types.

melvin wrote:I think Obs has it right - 99.9% of music listeners don't care how the sausage was made.

I also agree with JB - it takes good musical sensibilities (i.e. talent) to make good music, regardless of the tools being used.

Still, a musician who's spent years training his mind and refining his fine motor skills (with much frustration along the way) can't help feeling slightly annoyed when some n00b can all of a sudden just push a button and achieve sonic perfection.

Agree with the first 2 points - but not so much on the third.

I mean - I am not a musician who has spent years training his mind and refining his fine motor skills, I am just a bloke who has played the same 3 chords over and over again for 20 years - so I feel no offence at being replaced by a push of a button. But I can relate in other aspects of life.

As an engineering student in the 1980s, I spent enormous amounts of time learning (amongst other things) how to do complex stuctural analysis from first principles with a pen and paper, learning how to use and do calculations with pre-laser technology theodolytes as well as learning how to program computers using now-archaic languages such as pascal or fortran. It is tempting to be annoyed at the noobs who have come along in the era of spreadsheets, structural anaysis software and total station survery gear, who can accomplish previously very demanding tasks with the press of a button - but it is entirely pointless. There is new technology available, use your existing skills to make the most of it.

Or another example I'm sure Spud may relate to - after spending years learning how to do meticulously accurate engineering drawings by hand - having that skilled devalued overnight (apparently, though not actually) by a monkey with AutoCAD. Whatever you do - this will always happen. Learn to live with it and embrace the future.

A monkey with Autocad is still a monkey.

SPUD

Hey now, I'm a CAD Monkey! But my saying goes: "When CAD is improperly used, it's just an expensive Etch-A-Sketch"

Billy's Little Trip wrote:I hope I'm not coming across as a guitar snob, because I'm not and that wasn't my intension or point.

Well, it did kinda read that way, but I already know you don't have an issue with sequenced music, so I didn't worry about it. I wasn't having a go at you early when I posted about guitar snobs, just commiserating with ujnhunter who's wife's uncle does sound like a bit of a snob.

My only disagreement with you is that I think your prediction of some kind of backlash against melodyned 'perfection' is wrong, and I don't think there's been any such backlash against the 'perfection' of midi sequencing. But this isn't an argument for me or anything, it's just a disagreement. I don't have a problem with you, I just think you're wrong

I'm rarely right.

How is this even remotely possible.

Adam! wrote:How is this even remotely possible.

flux capacitor

This software is completely incredible. What this means for music composition, performance, scholarship, etc. is indescribable.

Adam! wrote:How is this even remotely possible.

I've been mulling it over, and here's the best I can come up with:
Do a FFT transform to get a 2-dimensional array of intensity@frequencies over time. Look for where an intensity above a certain threshold exists at the same frequency band for a given period of time, and assume that's a note (you're gonna want some fuzzy matching here, I'd think). For each note thus identifed, surgically filter it to it's own audio stream (using the frequency range you've already identified the note at). Do your usual pitch-shifting on this. Mix it back with rest of the signal.

I'm sure that wouldn't give you the same results that melodyne does, but it's all I can come up with.

When you put in the disc to install it, a whole bunch of teeny tiny monkeys with perfect pitch come out, and notate what they hear. Fast fourier is so yesterday.

*scary little girl voice* "It's the meers"

obscurity wrote:

Adam! wrote:How is this even remotely possible.

Do a FFT transform to get a 2-dimensional array of intensity@frequencies over time. Look for where an intensity above a certain threshold exists at the same frequency band for a given period of time, and assume that's a note (you're gonna want some fuzzy matching here, I'd think). For each note thus identifed, surgically filter it to it's own audio stream (using the frequency range you've already identified the note at). Do your usual pitch-shifting on this. Mix it back with rest of the signal.

That's the only thing I can think of too, but there are SO many problems with that. As I'm sure you know, on guitars each note in a chord isn't just the fundamental, but a whole complicated series of harmonics and overtones, as well as non-harmonically related content. Well, like the programmer himself says, it shouldn't really work on paper.

Adam! wrote:

obscurity wrote:

Adam! wrote:How is this even remotely possible.

Do a FFT transform to get a 2-dimensional array of intensity@frequencies over time. Look for where an intensity above a certain threshold exists at the same frequency band for a given period of time, and assume that's a note (you're gonna want some fuzzy matching here, I'd think). For each note thus identifed, surgically filter it to it's own audio stream (using the frequency range you've already identified the note at). Do your usual pitch-shifting on this. Mix it back with rest of the signal.

That's the only thing I can think of too, but there are SO many problems with that. As I'm sure you know, on guitars each note in a chord isn't just the fundamental, but a whole complicated series of harmonics and overtones, as well as non-harmonically related content. Well, like the programmer himself says, it shouldn't really work on paper.

It's got to be a synthesized copy of every pitch within a note with a roll up and roll down to the next frequency, but chopped into millions of little fragments.
....does that sound convincing?

Billy's Little Trip wrote:

Adam! wrote:

obscurity wrote: Do a FFT transform to get a 2-dimensional array of intensity@frequencies over time. Look for where an intensity above a certain threshold exists at the same frequency band for a given period of time, and assume that's a note (you're gonna want some fuzzy matching here, I'd think). For each note thus identifed, surgically filter it to it's own audio stream (using the frequency range you've already identified the note at). Do your usual pitch-shifting on this. Mix it back with rest of the signal.

That's the only thing I can think of too, but there are SO many problems with that. As I'm sure you know, on guitars each note in a chord isn't just the fundamental, but a whole complicated series of harmonics and overtones, as well as non-harmonically related content. Well, like the programmer himself says, it shouldn't really work on paper.

It's got to be a synthesized copy of every pitch within a note with a roll up and roll down to the next frequency, but chopped into millions of little fragments.
....does that sound convincing?

That seems to support the monkey theory.

Adam! wrote:That's the only thing I can think of too, but there are SO many problems with that.

Yeah, I didn't say it was the answer, just the best I could come up with I guess as a further thought, once you've got something that you think is a note, you could assume it's a fundamental and go play 'hunt the harmonics' based on frequency multiples, and strip them out too. But yeah, I imagine the code's gonna be full of fudging and special-casing and general kruftiness.

maybe it works by magic





-hand twisty thing-

roymond wrote:

Billy's Little Trip wrote:It's got to be a synthesized copy of every pitch within a note with a roll up and roll down to the next frequency, but chopped into millions of little fragments.
....does that sound convincing?

That seems to support the monkey theory.

I don't know nuttin bout no monkeys, but I started typing at random non stop one time, and wouldn't you know it, buried in the text was Shakespeare's Romeo and Juliet. Go figger.

Billy's Little Trip wrote: I don't know nuttin bout no monkeys, but I started typing at random non stop one time, and wouldn't you know it, buried in the text was Shakespeare's Romeo and Juliet. Go figger.

I guess you are personally worth at least 10 000 monkeys

Adam! wrote:

obscurity wrote:

Adam! wrote:How is this even remotely possible.

Do a FFT transform to get a 2-dimensional array of intensity@frequencies over time. Look for where an intensity above a certain threshold exists at the same frequency band for a given period of time, and assume that's a note (you're gonna want some fuzzy matching here, I'd think). For each note thus identifed, surgically filter it to it's own audio stream (using the frequency range you've already identified the note at). Do your usual pitch-shifting on this. Mix it back with rest of the signal.

That's the only thing I can think of too, but there are SO many problems with that. As I'm sure you know, on guitars each note in a chord isn't just the fundamental, but a whole complicated series of harmonics and overtones, as well as non-harmonically related content. Well, like the programmer himself says, it shouldn't really work on paper.

The harmonic/overtone signature would then probably help identify each note on each string... yes, it would make Fourier transforms a lot more complicated, but I imagine if you had a set of transforms for each signature (one signature for one string) then you'd have a significantly easier job.