When you find that everything has been done already?

[soundmodel]

When you find that everything has been done already?

I find that a large portion of audio processing is not scientifically particularly novel, but merely some small reiteration on well-known principles. This saddens me.

But this also sparked the thought that "does this mean that everything doable has been done already"?

[Michael L]

There is still a lot to explore in physical modeling - timbre, harmonics, and how a sound changes over time.

[jupiter8]

It is a matter of perspective i think. I can easily think of several ideas that hasn't been done yet. And it depends on what you mean with new or novel. As a sidenote i just discovered that something i thought was very new and fresh was invented in the 60ies! People we're smarter back then than we give them credit for.

I believe there's plenty to be done regarding useability. Not doing anything new in particular but in a different way that gets you faster to the target. Like Soothe and Gullfoss for example. I hate to say it because it's been so damn hyped but there's plenty to do in the field of AI or machinelearning.

Ans there's rarely any gigantic leapd in technology just many small ones so it might seem nothing is happening but there's plenty to be enthusiastic about.

[soundmodel]

There is still a lot to explore in physical modeling - timbre, harmonics, and how a sound changes over time.

This is what I've been looking to as well, but one problem is that "good sound" is not determined by realism necessarily. So even when one could study how to recreate "real-world" expression curves or something, it's not necessarily useful for musical purposes, because some people prefer the sound of "simple filter openings".

So the measure of "progress" in art is not particularly trivial.

Sometimes I find that 90s music software was "just enough already for electronic music". Evidence: well a lot of good music was produced already in the 90s. 2000s brings A LOT of plug-ins, but I'm not sure if they add musically in such large proportion.

[Benutzername]

If you look at it from the final outcome then we have reached the end many years ago. No new algorithm of any shape or form will create a sound that hasn't been done yet.

There is still a lot of room for improvement in the UI. But the audio world is extremely conservative in this regard. A virtual Moog filter will only sound as good as a hardware Moog filter when the user is able to turn a black knurled 3D knob on the screen. The same filter controlled by a material design slider on a web page will never sound as good.

[Michael L]

This is what I've been looking to as well, but one problem is that "good sound" is not determined by realism necessarily.

Compare a $1000 cello and a $1M cello.
Both are 'real' but one 'sounds gooder'
Same with voices
But accurately modeling the details of a high quality cello or voice has not been done

What Would Jimi Do?

[soundmodel]

This is what I've been looking to as well, but one problem is that "good sound" is not determined by realism necessarily.

Compare a $1000 cello and a $1M cello.

A lot of excellent 90s electronic music was produced on budget gear of that age.

And this comparison is not fair in digital instruments, where it's pretty much possible to produce any algorithm one wants with very low cost. Heck, it could even be possible to produce a $1M cello simulation that sounds 95% like it.

[Michael L]

Like everything, it 'could even be possible' but it hasn't been done.
QED

[CinningBao]

I am not a line-coder, more a modular coder, but I still try to think as low-level as I can. And create my own solutions with bits of maths which I can add to my personal library.

I've been looking at, reading up on, and watching people who know what they're talking about, talk about EQs. Most EQs are designed upon the feedback/inverse-feedback principle, from the old analogue days. (you guys in this forum already probably know this, so I'll keep it swift). Other kinds of EQs have been developed purely in the digital realm.. FIIR (?).. convolution.. (and others, I'm sure) but they all suffer from the same problem of messing with the phase, unless linear phase algos are used, buffer requirements... Linear phase, _half_ solves the phase problem, but adds its own potential artifacts.

At some point, some crazy coder guy/gal somewhere will 'realise' another way to solve the frequency adjustment issue without affecting the phase or introducing pre-ringing/smearing.

What I'm saying is, we haven't found all the maths yet which could be used for, or repurposed into, tools for adjusting frequency content of audio material. We've got old feedback solutions reinterpreted into 'real circuitry' and fairly new digital solutions which have their own problems.

And since there's no way it's going to be me who comes up with anything as useful as that, I'll just carry on being creative with the low-level things I fudge together.

But that's just a problem with EQ, that's not even being creative with the concept of oscillators.. Physical modelling will roll on forever, getting better, weirder.. as we try to more accurately model the space and universe around us.

[soundmodel]

But that's just a problem with EQ, that's not even being creative with the concept of oscillators.. Physical modelling will roll on forever, getting better, weirder.. as we try to more accurately model the space and universe around us.

Physical modelling has been interesting, but I find that something like Chromaphone seems to pretty much do everything interesting. There was recently a new synth in this category:

https://www.kvraudio.com/news/expressiv ... 761/0/5168

But it seems as if the theory is limited in the possible elements. There are still pipes and plates and such. Possibly due to the feasibility of modelling with simple shapes.

But will these replace samples? No I don't think so. Resampling is one of the most powerful algos and it doesn't require any new algo to use. It just takes many iterations on existing algorithms.

old dist box at settings A
->
old dist box at settings B
->
old dist box at settings C
->
...
=
A quite new and possibly non-replicable distortion sound while still using just old dist box.

There has been some advancement in modelling when applied to other acoustical domains, but I'm not sure if the findings in these are usable for aesthetic sound.

[Richard_Synapse]

Physical/Analog modeling, user interface design and machine learning are a couple of areas with plenty of room left for innovation.

If none of that is interesting to you, you can always take existing concepts and combine them in novel ways

Richard

[melomood]

Is that a hint? A tease? A clue??

[yofiel]

code library so people can make their own stuff: 3000 downloads

Painstakingly optimized 32-voice synth with 100 waveform types, modulation matrix, and carefully designed saturating filter: 50 downloads.

Todays sharing of EQ library functions: 100 views in 4 hours.

Painstakingly written article on how the Arctic is expected to completely melt away in 2035: one view by a friend, and 50 comments by republicans telling me to stuff it without bothering to open the link.

[Tj Shredder]

When I was mainly in Max/MSP, I never had the problem to do things in new ways. When I look at the “market” there is close to 100% reiteration of old stuff…
Maybe a matter of perspective.
Read the Computer Music Journal to get an idea what has not been commercialized yet.
For example, have you ever heard about tropical additive synthesis? It exists, but I had to make Grid patch myself to hear it…

[xoxos]

not new, but hasn't been done for public access

in my youth, computers used to have orthogonal draw commands eg. "r12u4r16m(15,9)m(3,7)" which seems like an easy format for customisable quadrature oscillators, eg. "the way EEs draw pictures on oscilloscopes". an interface could probably include first order bezier without involving too much overhead; sort of a novelty, sort of a timbrally funky thing to do, and with lots of appeal for persons wishing to wow others.

a few transform options (offset, scale, rotate) would allow for the production of little films and make it decidedly more interesting sounding.

similarly, what about fourier synthesis using paths = probably no aural entrants.

i think there are lots of things that haven't been done, but people enthusiastic about synthesis tend to be too focused, and people enthusiastic about some things are too fucused so there you are. spending a couple of years not thinking about synthesis gave me significant perspective on the things i wasn't doing. how many even percussion synths using multidimensional oscillators? there are a fair number of comments on the internet where engineers and mathematicians fail to understand why musicians want to do something differently. one of my favourite vst companies has a volume of material appropriated from persons who didn't want to instrumentalise a routine.