where next with software instruments?

[Music Engineer]

such higher level meta parameters sound excellent. I would rather not have to make them myself as meaningful structures are not easy (as we can see with physical modeled instruments). However, a system where a synth can have such parameters and the ability for the user to 'unfold' the parameter and edit it would be fascinating.

in my straightliner synth, it is possible to route midi controllers to multiple target parameters and define an upper and lower limit - the values of the target parameters corresponding to 0 and 127 respectively. when choosing the 'upper' limit, in fact, below the 'lower' limit, one can also reverse the direction of the modulation. with this system, one can essentially use midi controllers for meta parameters - like routing a controller to the tunings of the individual oscillators with appropriate ranges, one creates 'detune' parameter. meta parameters are actually not that new: the well known 'accent' of bassline synth is a meta-parameter. it usually controls filter cutoff, envelope depth, resonance, volume and perhaps even more parameters at the same time. generally i tend to think that meta parameters should be defined in terms of perceptual measures (such as 'brightness', 'inharmonicity', etc.) rather than in physical ones (such as string-tension). ...more a proponent of physically informed synthesis rather than downright physical modeling

[Ben [Camel Audio]]

Interesting dicussion

How about something like this - take a synth with a fairly large library of good patches like Discovery or impOSCar. Then, pre-compute the statistical relationships between the various parameters in those "good" patches and provide some kind of guided randomization interface that uses these weights to cut down on the number of dead ends. Ideally this randomization interface could also be the performance interface.

I think it would be really interesting to do something based on statistical analysis of a large database of presets - ideally with some form of attribute tagging. I think the approach could be a great way to generate new sounds. Hopefully I'll find some time to spend on this in the future. I think it would be a cool thing to do as research MSc or PhD these (if by some chance someone reading this is interested doing something along these lines, please get in touch as I'd be happy to help.)

such higher level meta parameters sound excellent. I would rather not have to make them myself as meaningful structures are not easy (as we can see with physical modeled instruments). However, a system where a synth can have such parameters and the ability for the user to 'unfold' the parameter and edit it would be fascinating.

I agree that this is a cool thing to be able to do - and agree with what accidose said about being able to do somewhat physical modelling like things by mapping multiple destinations each with its own custom curve to a single knob. Alchemy has a similar approach to Kore, in that you have a bunch of meta (or performance) knobs which can be applied to whatever you like. We have 8 general purpose knobs, 2 xy squares and ADSR performance controls. I'm not quite sure how far Kore takes it, but in Alchemy, you can make a single performance knob control every single other knob on the interface (including all modulation depths and modulator parameters), each with a specific amount of depth and its own custom 'modmap' curve (for doing anything from specific curves to switching etc). This information is stored as part of the preset, so sound designers have a lot of freedom to make patches that do interesting things - from physical modelling type stuff, to switching between different melodic or drum patterns driven from the arpeggiator to something Junkie XL did which is emulate the sound effects from an old school computer game complete with parameters like number of players etc.

a lot of this stuff is already possible if you assign multiple parameters to 1 knob, which i think is a major thing missing in most synths today. to be able to assign many different things to 1 controller in varying degrees of strength/direction.

reason thor can do this, but its pretty sad that there are only 2 knobs, and some things arent able to be assinged.

for example, i think we could emulate different synths and stuff by creating modulations that act like: "scale filter resonance to cutoff 30%, begin at filter level 50, transition exponentialy"

that way you could say like, all the time when the filter is low, the resonance will be higher and vice versa, so then you could modulate the filter how ever and the resonacne woudl automatically follow.

or you could even say, "scale filter cutoff to osc mix by -35%, starting at mix 70, transition lineraly"

Interesting ideas - you could do all those things with Alchemy.

generally i tend to think that meta parameters should be defined in terms of perceptual measures (such as 'brightness', 'inharmonicity', etc.) rather than in physical ones (such as string-tension). ...more a proponent of physically informed synthesis rather than downright physical modeling

I agree - I think its more user friendly.

[BONES]

How about something like this - take a synth with a fairly large library of good patches like Discovery or impOSCar. Then, pre-compute the statistical relationships between the various parameters in those "good" patches and provide some kind of guided randomization interface that uses these weights to cut down on the number of dead ends. Ideally this randomization interface could also be the performance interface.

The problem with that is what is good for you is likely to be useless to me. Beyond that, what you describe is pretty ,much ORION's Genetics, which we've had for around 5 years now. You select two presets and press the BREED button. Every time you do, it randomly selects parameters from each. You just keep pressing it until you get something that works. About 75% of the presets in my synths are derived from that process, especially by letting Genetics choose two presets at random. Mostly the results need some finessing but its a good way to get completely different sounds from a certain set of constraints.

in my straightliner synth, it is possible to route midi controllers to multiple target parameters and define an upper and lower limit - the values of the target parameters corresponding to 0 and 127 respectively. when choosing the 'upper' limit, in fact, below the 'lower' limit, one can also reverse the direction of the modulation. with this system, one can essentially use midi controllers for meta parameters

Yamaha CS1x did this 10 years ago.

So, you can see that really there is nothing new under the sun here. No revolution, not even an evolution. I think the area where the real revolution can happen is in people's approach to the task. Reading this and other forum, it seems to me that very few people have adapted successfully to the software environment. e.g. We still see demand for multi-timbral instruments, which are completely and utterly redundant in the software world.
People need to look at things differently. Maybe some workflows encourage this but few people take it to the next level. e.g. Some people will add an instance of EnergyXT as a VSTi and build a layered patch in it, like a Combination in a KORG workstation, but few would intuit that eXT is really redundant in most hosts and it should be more efficient to add all that complexity directly in the main host; cutting out the middle-man, if you like. If people saw their host/sequencer/DAW as a single, unified instrument, rather than the place to bring instruments together, it would encourage them to work in different ways and achieve greater results.
The hardware-alike UI paradigm dominates because people's thinking and workflow is still shaped by that world. Maybe a new UI would encourage people to think differently but personally, I would hate to have an extra layer of obfuscation to deal with, as suggested by this whole meta-parameter concept. Its doable right now, within current paradigms, and I'm happy for it to remain that way.

[gnu23]

2) Get rid of Steinberg & Midi.

MIDI existed long before Steinberg VST. It has been the foundation of every home studio sequencer from about 1985 onward and every computer music product since about the same time. (a check of Wikipedia shows MIDI was introduced in 1983) I can't think of a better way for computers and instruments to communicate, control, and synch up with each other. If there were, MIDI would have already gone by the wayside.

I don't doubt that VST has its problems, but it has wide adoption making it hard to replace.

As with a lot of things in this sphere, it's not the standard that is so much at fault as it is the implementations of a given standard.

[Hewitt Huntwork]

I like the drop-down programming menus in Firebird. All synths should have a programming mode with drop down menus.

[ew]

I'm not quite sure how far Kore takes it, but in Alchemy, you can make a single performance knob control every single other knob on the interface (including all modulation depths and modulator parameters), each with a specific amount of depth and its own custom 'modmap' curve (for doing anything from specific curves to switching etc).

I haven't hit any limits with assigning multiple controls to the same knob with Kore, but then again, I think the most I've tried to assign is 16. You can set the depth individually, but there's no curve per se (nice idea, Ben... )

However, an advantage to Kore is that the same knob can control multiple parameters in different instruments at once if your user page is at the performance level

ew

[aaastronomer]

Collapse of modern civilization ........ back to acoustic instruments only, yay!

u.s.a going now...all hail prez. trig!

[aaastronomer]

If you honestly think software can just keep getting more efficient on the same hardware, beyond a certain point, you're somewhat mistaken.

That misses the point completely. The hardware is exponentially more powerful every year, yet is still being taxed by the new software....

...written in the same old taxing programming languages (C/C++/Java)...need new more precise and formal languages (maybe based on category theory).

[Uncle E]

If you really want to explore new avenues, try SynthEdit, SynthMaker or Reaktor.

Exactly, I don't understand how this thread has persisted so long when Reaktor (in the form of Generator) has been around for over a decade. Consider Oasys PCI, Scope DP, and Kyma, as well, and it's clear we've been beyond interface restrictions for a long time.

[Music Engineer]

(a check of Wikipedia shows MIDI was introduced in 1983) I can't think of a better way for computers and instruments to communicate, control, and synch up with each other. If there were, MIDI would have already gone by the wayside.

ohh, midi has certainly limits which do not seem up to todays technical standards anymore. the foremost flaw in my opinion is the low resolution for controllers. maybe OSC can replace it in the mid-term future...

http://www.kvraudio.com/forum/viewtopic.php?t=228085

[Music Engineer]

in my straightliner synth, it is possible to route midi controllers to multiple target parameters and define an upper and lower limit - the values of the target parameters corresponding to 0 and 127 respectively. when choosing the 'upper' limit, in fact, below the 'lower' limit, one can also reverse the direction of the modulation. with this system, one can essentially use midi controllers for meta parameters

Yamaha CS1x did this 10 years ago.

So, you can see that really there is nothing new under the sun here. No revolution, not even an evolution.

O.K. i didn't mean to suggest, that this is some outstanding innovation. nice feature, though

[aaastronomer]

It seems that software instruments are powerful enough now that you can achieve just about any sound you can imagine in very high quality with them. However, the interfaces still mostly mimic traditional synths and samplers and are fairly slow and tedious means of exploring the sound universes of these tools. Isn't the next step completely rethinking the UI of these tools? Why not try something radical and put all that CPU under our hoods to better creative use? The evolutionary patch approach in the Nord G2 is a step in the right direction but there's so much more that could be done.

i'd like trainable 'style generators' with my next gen synths. just load samples of reference music into a training set and the synth: (1) analyzes the sonic content of the training set to generate presets that fit the style, (2) analyzes the beat structure to generate appropriate midi, (3) detects and analyzes any effecting (probable) in sample set to generate appropriate presets, (4) lightly randomizes generated contents, (5) plays back new contents for review.

all of this is slowly coming together as shown in each 'computer music journal' issue (how bout that 'dynamic score' in the current issue?).

[mcnoone]

Don't forget that in another 7 or so years, the first quantum computers may be available to the public (only big government has them now). With quantum computers, things like your eye movement to computer control will be possible to the average person. So your eyes can do the work. Robots will also be more intelligent, so they might be able to write your music for you, while you lay on the couch with your chips and beer, watching TV and doing nothing...wow technology really sucks. I for one would like to get a female robot to replace my wife.

[BONES]

i'd like trainable 'style generators' with my next gen synths. just load samples of reference music into a training set and the synth: (1) analyzes the sonic content of the training set to generate presets that fit the style, (2) analyzes the beat structure to generate appropriate midi, (3) detects and analyzes any effecting (probable) in sample set to generate appropriate presets, (4) lightly randomizes generated contents, (5) plays back new contents for review.

Buy yourself a radio, you sad f**k. Is that really all music means to you? A random collection of styles? Its pathetic.
But it brings up an interesting aspect to all of this, which is that it seems most people want the equipment to do more of the work for them. As much as there are parts of the process that shit me to tears, it makes achieving the result all the more satisfying. Take out the hard work and what would be the point? You may as well go and join a band, FFS.

[pdxindy]

such higher level meta parameters sound excellent. I would rather not have to make them myself as meaningful structures are not easy (as we can see with physical modeled instruments). However, a system where a synth can have such parameters and the ability for the user to 'unfold' the parameter and edit it would be fascinating.

I agree that this is a cool thing to be able to do - and agree with what accidose said about being able to do somewhat physical modelling like things by mapping multiple destinations each with its own custom curve to a single knob. Alchemy has a similar approach to Kore, in that you have a bunch of meta (or performance) knobs which can be applied to whatever you like. We have 8 general purpose knobs, 2 xy squares and ADSR performance controls. I'm not quite sure how far Kore takes it, but in Alchemy, you can make a single performance knob control every single other knob on the interface (including all modulation depths and modulator parameters), each with a specific amount of depth and its own custom 'modmap' curve (for doing anything from specific curves to switching etc). This information is stored as part of the preset, so sound designers have a lot of freedom to make patches that do interesting things - from physical modelling type stuff, to switching between different melodic or drum patterns driven from the arpeggiator to something Junkie XL did which is emulate the sound effects from an old school computer game complete with parameters like number of players etc.

Alchemy is interesting because it is an additive synth (well, more than just that).

Additive synthesis is vastly powerful, but hard to work with. Still, imagine trying to resynthesize a speaking voice with a subtractive synth! For the sort of meta parameters we are talking about, I think that the sort of additive/spectral hybrid like Alchemy lends itself to such controls more than the standard subtractive model.

In general, I think it is hard for the user to set up huge complex modulation schemes that are musically useful/interesting.

The resynthesis is kinda like a meta parameter itself.

So it could be interesting to take the resynthesis engine and analyze its output. So you are using a bunch of partials and to replicate a voice, you are creating a kind of formula where each partial has its own modulation over time. In analyzing a bunch of wind instruments, there might be common formulas that can then be the basis for a kind of preset resynthesis which can be applied on top of existing sounds.

I think the potential for modeling instruments in Alchemy is quite exciting.