Can Audulus create an EHX Stereo Talking Machine effect?

    I haven't really dived into Audulus, but if this effect was possible to create in Audulus with an intermediate level of work, it would be worth it to invest the time.

    My understanding is that I could create formant filters for a few different vowels by setting up multiple band pass filters in parallel, using frequencies which I've seen on the Internet. I would then sweep between the vowel sounds, using the dynamics of the played instrument, something like an envelope follower.

    I didn't see a band pass module in Audulus, but can you make one with LPF and HPF in series?

    Am I on the right track? Any advice would be appreciated. The closest I've seen to this effect on iOS is Sugarbytes WOW filter, but the degree of vowel emulation seems much less than the EHX box.

  • This is a good resource on Formant Synthesis:

    To do dynamic sweeping, you'd extract the amplitude envelope with an Envelope Follower and use that signal to fade two crossfade nodes that have their inputs swapped - check the patch attached for an illustration of that.

    There are bandpass modules that people have made here on the forum - you can download all the patches at once here:

    We are also about to implement a newer, much faster audio engine that will allow the Z-1 node to be used more efficiently, allowing for some really nice and fast analog-modelling digital filters! :)
    Formant Crossfading.audulus
  • Thank you! I'll check out your link and the attached patch.

    Would you recommend holding off on this exercise until the new audio engine comes out, or will everything I might do now just take advantage of the new algorithms?
  • @mdarby - The new audio engine won't break your patch, it will just make it run more efficiently, especially if it has Z-1 filters in it, so I'd say go ahead and start trying! :)
  • Does Audulus have a frequency analyzer (frequency on X-axis, amplitube on Y-axis)? This would help me tune the band pass filters to achieve the formant filter.

    Currently, I am running out the signal through Audiobus to a mastering app with a wave representation, but I would much rather use a built-in module. Thanks!
  • Oscilloscope module please! :-)
  • @biminiroad If you have a moment, can you take a look at my patch, which chains three BP filters for each of two frequencies? Thanks! Feedback is welcome.
  • @mdarby @DS_in_the_ATX - the oscilloscope is called the Waveform Node, in the Metering menu. Its maximum is between -1 and 1, but you can also translate incoming signals so that they fit within the view of the waveform just by using a level node in front of it. Look under the Modules Library -> Utilities -> Vias -> Waveform Via - this one also allows you to pass the signal through.

    Looked at the patch - very cool! You are, however, clipping your output very hard - if you use the Master Output module or at least put a tanh(x) function before the output (which is a soft clipping expression) you won't get such an aggressive digital clip sound. It's helpful to use waveform meters on the outputs as well, because if the wave isn't fitting in the view of the window, that means you're clipping.

    But anyway, sounds like you're well on your way to making some awesome effects! Especially with this new audio engine that's coming that will have better use of z-1 implementation, you'll get some real nice and gritty analog-sounding filters if you want!
  • @biminiroad - thank you kindly!! how cool is this..?!?!
  • @biminiroad -- thanks!

    On the Waveform node, at LFO frequencies it traces out the waveforms from the oscillators as expected. At audio frequencies, presumably because the Waveform sample rate is two slow, it starts to draw what seems like two, stereo waveforms. How do you interpret the output when this happens? What indicates clipping? At LFO rates on a sine wave, for example, when overdriven I can see the extremes of the sine wave go off screen, which presumably is clipping. I am less clear at audio rates.

  • @biminiroad - turns out I *was* using the Waveform Node (see attached), but mdarby's request put the oscilloscope idea in my mind; "(frequency on X-axis, amplitube on Y-axis)" - is it possible to control the horizontal axis of the Waveform Node with an input signal (Lissajous figures, yeah!!) ..?
    With many thanks for your patient assistance - best regards!
    env follower filter-flanger.audulus
  • @DS_in_the_ATX - not currently, but there is something in beta that will let you do all sorts of visualizations, including but not limited to Lissajous!
  • @Michael - it does not draw two stereo waveforms - it's just the peaks and valleys of the waveform like you'd see in any other kind of audio program if you're zoomed out far enough. The new visualizer we're adding will let you draw waves much faster. Scoping is complex mathematically I believe ( @macromachines would know ) and perhaps we'll have just a scope node that's more resource intensive but more accurate - dunno. That would be really good for fine tuning things though.

    Clipping is indicated by the waveform extending outside the range of the waveform node. Your best bet is to keep it low and them amplify it with your speakers instead of driving it as hot as possible. The old adage of getting as hot a signal as possible is a remnant from the analog days when you had a significant noise floor to contend with - 1/2 way down on the meter with 32 bit sampling is not perceptually different than all the way up.

    Hook up an oscillator with an amplitude of 2 to a waveform node - you'll see it peaking, like you mentioned - there's no difference in the way peaking works versus frequency, because peaking is an issue with too much amplitude (i.e., it's independent of frequency and works the same in the audio range).

    Since the waveform node doesn't draw very fast, it might not catch peaking transients, but it's generally a good indicator. I have an external meter on my Apollo interface that tells me when it's peaking, so I go by that - maybe you have something similar? Also, if you use tanh(x) expression before an output, it will basically be impossible to clip - you'll just get some nice analog-sounding overdrive.

  • @DS_in_the_ATX

    Here is a little clipping indicator I use.

    Overload Light.audulus
    2048 x 1536 - 147K
  • @JDRaoul , thank you! and @biminiroad ~ great news, thank you!
  • Would love to have something that can function as an oscilloscope.
  • @fferreres - the waveform node is an oscilloscope, though a limited one, admittedly. We have a new visualizer node that is coming that will allow you to make more complex scopes.
  • great, I think you kmow what I mean :-) great news then.