Tone2 Rayblaster: OUT NOW! (demo version available)

[goldenanalog]

It does not totally emulate a filter but by loading a single cycle wave you get the "impulse response" of a filter.

Ingo

That makes no sense at all. You cannot possibly determine what a filter does unless you analyze 2 waveforms, one pre filter and one post filter to determine how a filter affected the original waveform.

If you were to make assumptions about the waveform that you were analysing, you could manufacture a filter based on assumed changes in the waveforms' spectra *post* filter. You're absolutely right, of course: the previous statement implies that probability is used like using Calculus to calculate the slope value at discrete points of an assumed *known* function.

My guess is that Tone2 looked at enough filter data to build filter models into a table that are recalled and fitted to the waveform being analyzed. What might be funner (If I've read it right) is have the filter response continously change over time as DSP is used to examine each zero-crossing periodic sample (as an example) and update the filter response to that sample.

I don't think there is any "table" involved. First the responses used from waveforms of different synth also sound different and second this is not limited to "analog" waveforms. If you use a complex waveform you'll also receive a complex filter. Anyway if the waveform contains a part that corresponds to a resonant waveform the result could be that you receive some kind of resonant sound too.
If you mix two waveforms within one oscillator you also get a kind of mixed filter.


Ingo

It could be similiar to what happens when uncompressed music is turned into an mp3, where intelligent judgement calls based on statistics are made wholly dependent on the trend of the waveform.

OK: Starting with reference 'parts' of filter types in a table-you examine a single-cycle waveform, and besides looking at it's spectra, you assume that:

1.) It is *post*-filter processed; and consequentially:

2.) It has an original shape determined by what template(s) the assumed-post-filtered waveform fits best. Then: looking at the waveform as a single point, your final filter shape covering the audio spectrum is a projection modelled upon statistically-based probabilities gained through intense scientific analysis of known synthesizer filter types.

Considering this method of filter-characteristic extrapolation/regeneration along with the current breed of desktops leads me to believe that it would be relatively easy to have quite an extensive table of filter 'pieces' resident to construct a complex filter from; and that the whole waveform analysis/resynthesis thing could happen practically in real time.

I may be completely wrong about applying this to Rayblaster; but it is one relatively simple way of how building a complex filter based on comparison analysis of a single waveform might work.

One thing is clear: We are getting further and further away from the original intent of modelling analog synthesizers and instruments in software; Rayblaster (and Iris) clearly leave the realm of what's possible using purely analog synthesis or digital playback of recorded instruments.

And the reason is simple: We now have enough CPU resources available that we can start to do things in real time that had to be done off-line in the past.

Edited for clarity.

[Kriminal]

Why is IMS so close to what happens within our brain?

Hilarious

This is a quote from the website:...blah blah




Ingo

My quote is from the website, i can read thanks. No need for a fanboy to repeat what i aready know.

Have you made your jupiter brass sounds for it yet?

[Shabdahbriah]

I'm trying hard not to let the vertigo inspiring demo-video sway me away from a golden opportunity to experience this "new" form of scientifically derived and applied synthesis/re-synthesis... mind-numbing hyperbole notwithstanding, it is potentially 'viable'.

Any knowledge of a price on this, yet? I suspect even an introductory price will be fairly steep, given the need to recoup some of the big-bucks spent on that video.

Further, having no personal experience/background with Tone-2, are bugs worked-out reasonably quickly?

Thanks.

[pdxindy]

It does not totally emulate a filter but by loading a single cycle wave you get the "impulse response" of a filter.

Ingo

That makes no sense at all. You cannot possibly determine what a filter does unless you analyze 2 waveforms, one pre filter and one post filter to determine how a filter affected the original waveform.

If that is the case why could i create a quite convincing recreation of a Synth Brass based on the waveform of e.g. an analog emulation or real analog synth?

So take an additive synth, create a complex waveform without any use of a filter and bring it into RayBlaster... Or use anything from a DX7... Does RB tell you that no filter was used? Or does it still create a filter profile based on that sound?

It is impossible to tell that a filter was even used cause any filtered waveform could also be created without any filter... If you are analyzing one waveform, it must have a set of assumptions and make its best guess as to a filter model based on those assumptions.


Regardless, it sounds like an interesting synthesis method...

[steff3]

I mean , has to test it when it got released.

the video they have on their website is a laugh. Asynthesis that allows for sounds that are not possible with other synthesizer -but why throughout that video I have the impression of having heard all that before? solely deja-vu?
And yes, the level of skills of that poor guy had not necessarily been emphasised, totally audible.

will give it a try when it got released.

best

[Ingonator]

It does not totally emulate a filter but by loading a single cycle wave you get the "impulse response" of a filter.

Ingo

That makes no sense at all. You cannot possibly determine what a filter does unless you analyze 2 waveforms, one pre filter and one post filter to determine how a filter affected the original waveform.

If that is the case why could i create a quite convincing recreation of a Synth Brass based on the waveform of e.g. an analog emulation or real analog synth?

So take an additive synth, create a complex waveform without any use of a filter and bring it into RayBlaster... Or use anything from a DX7... Does RB tell you that no filter was used? Or does it still create a filter profile based on that sound?

It is impossible to tell that a filter was even used cause any filtered waveform could also be created without any filter... If you are analyzing one waveform, it must have a set of assumptions and make its best guess as to a filter model based on those assumptions.


Regardless, it sounds like an interesting synthesis method...

I created several waveforms based on a mix of different partials using Camel Audio Cameleon 500o without any additional filters in Cameleon.
Even with few partials you could get quite complex results which lead to several different timbres while turning the Formant knob.

Like i already mentioned Rayblaster will try to create a filter response from any waveform, no matter which origin and shape it has.


Ingo

[goldenanalog]

Rayblaster will try to create a filter response from any waveform, no matter which origin and shape it has.


Ingo

If this includes an undistorted (pure) sinewave; then I highly suspect that Rayblaster is working from some form of a table, Ingo.

A pure sinewave has nothing to analyse, other then a frequency and an amplitude.

[Shabdahbriah]

Rayblaster will try to create a filter response from any waveform, no matter which origin and shape it has.


Ingo

If this includes an undistorted (pure) sinewave; then I highly suspect that Rayblaster is working from some form of a table...

Indeed.

[TwoToneshuzz]

I think the video is not a turn off. It's an interesting comcept on the whole. I like the scientific basis. The impuls wave speed the physiology of the ear, pyscho acoustics, why shouldn't this advanced tech become available in our sound generators.

If we are able to land curiosity on Mars, this is solid proof that advances are being made at a breakneck pace in all areas of science. There are more scientists, and more computing craft available today then ever before in the history of man.

It's going to be an interesting ride. So eat your vegetables and keep a good health.

[Ingonator]

Rayblaster will try to create a filter response from any waveform, no matter which origin and shape it has.


Ingo

If this includes an undistorted (pure) sinewave; then I highly suspect that Rayblaster is working from some form of a table, Ingo.

A pure sinewave has nothing to analyse, other then a frequency and an amplitude.

With a pure Sine not much happens but in the higher Fromant value the Sine turns into a Pulse as usually in very high values this turns into a BP filter. For "normal" use you usually use the region between -200 and 0 (= middle position).

Anyway i have just used a waveform created with Cameleon 5000 that uses the odd harmonics 1 up to 15 and no filter. i then moved the Formant knob in Rayblaster and took some screenshots of a Signal Analyzer.
Here is a PDF with the screenshot and a screenshot of the waveform (display in Rayblaster) at the beginning:

https://dl.dropbox.com/u/53230726/Raybl ... es%201.pdf

Some of the interesting timbres could be between the Formant values used for the test. At lower harmonics (mean you "zoom" into the waveform" this could be closer to a LPF or a BPF and in higher harmonics it could get more complex.
The result also depends on the start phase, especially when lower harmonic values (1 or negative) are used. The "harmonic" value has a range of -128 to +16.
The result also changes drastically when you use a random selection of a few additive partials which could lead to complex waveforms.

The frequency spectrum and corresponding waveform could drastically change when different "Osc windows" (e.g. "Cosine", "Soft Saw", "Comb 4X", "FM1", "Exponential 1", "Sine 8X",...) and/or "PW sequences (e.g. "Saw<PW>DbSaw", "Square<PW>Peek",...) are applied.



Ingo

[goldenanalog]

With a pure Sine not much happens but in the higher Fromant value the Sine turns into a Pulse

This would happen at the Nyquist limit; but in real-life a pure sinewave is a single fundamental harmonic w/o any overtones, no matter what the frequency.

A pulse is a square wave with a duty cycle of less then 50% (usually much less); it is not a sinewave.

If there is a filter that is extrapolated from a pure tone; this then might be considered 'default' for Rayblaster since Rayblaster is being given no actual information to construct a filter from.

What would be interesting to see is that if you fed Rayblaster the same pure sinewave (f=c1;A=c2) twice, would the resulting filter signatures be exactly the same, or does Rayblaster use random number seeding to skew filter generation?

[Ingonator]

If there is a filter that is extrapolated from a pure tone; this then might be considered 'default' for Rayblaster since Rayblaster is being given no actual information to construct a filter from.

You could be right with that. I think the default behavior is that a "Formant" value of -200 to 0 corresponds to a LPF while a value of 0 to +200 corresponds to a BPF. This changes more or less based on which filter response could be detected by Rayblaster. Anyway this does not prove that there is any kind of "table" involved. If that would be the case you would not get different responses for different LPFs of different synths which seems to be the case, especially for resonant waveforms (i created the waveforms for more than 10 different analog synths/emulations).
It does also not seem to be necessary that the original source/synth of a waveform used a filter at all. The only important thing for the analysis seems to be the shape of the waveform which of course is quite obvious.

At very high Formant values (above e.g. +100) the result is almost always a BPF while there are variations based on the imported filter response.


Ingo

[Ingonator]

At the Rayblaster Beta forum Markus announced that both the video and the product text on the website will be improved ASAP.


Ingo

[ariston]

Any knowledge of a price on this, yet? I suspect even an introductory price will be fairly steep, given the need to recoup some of the big-bucks spent on that video.

Good one.

[Ingonator]

I just created another PDF file based on using a a Resonant waveform of the 12db LPF of Saurus inside Rayblaster. At the beginning this includes a screenshot of the waveform (from Rayblaster) and later screenshots of a Signal Analyzer at different "Formant" values (from -200 to +150):

https://dl.dropbox.com/u/53230726/Raybl ... es%201.pdf

The name "Soft Saw" below "LoadWave1" is not the name of the wave but the name of the "Osc window". That "Osc window" in the waveform display is the grey part in front of the blue background. This could have a huge impact on the resulting sound. all over there are 23 different "Osc windows" where some could emulate a certain filter (e.g. "Comb 4X").

For a comparison here is a screenshot of the Signal Analyzer using Saurus with the 12dB LPF and Resonance (should be close to the waveform used in Rayblaster):
https://dl.dropbox.com/u/53230726/Sauru ... er%201.png

At Formant values above +75 (0 is the middle position) the filter starts to morph into a BP filter. This is a normal behavior in Rayblaster. The "normal" range is between -200 and 0 or slightly above 0.

As you could see at the PDF the "resonant peak" moves with increasing the Formant knob, like it would do with the Cutoff in the original synth.
If you load the same waveform into a "normal" sampler you could just cut off the "resonant peak" but not shift it this way.


Ingo