CALL ME PEAK!

as a little fun project i tried to educate my peak to "speak".

the first word he should learn is his own name "peak".

for this i used all of the three envelopes and both pseudo envelopes by the LFO, 15 mod matrix routes, wavetable sweeps and little FM / VSYNC.

as there is only one filter possible per patch, i decided for a 12b/oct low pass with modulated resonance. speech synthesis usually requires a set of (at least) two bandpass filters, which are not possible on peak, so i had to find other ways to shape the energy spectrum / harmonics, i.e. using the slew of a rect LFO and some filter mod by osc3 

The hardest thing seems to be the last char "k" (which are difficult to synthesize even for classic speech synths). it seems it is much easier to let him speak "deep", because the "p" is easier to implement with a limited "envelope" form by LFO and "d" at the beginning relatively easy with "full features" envelopes (attach/decay phase) ß)

To form the "p" i used standard envelopes (filter and amp) while pushing energy info with some noise.

the slight i-e wavetable shape and exponential pitch movement of osc2 gives the "ea" part which get damped by lfo1 and 2 envelopes to let the consonants trough...

this was just a fun project out of my hand / just by hearing - without any deeper planning and analysis tools and i die not further optimized the resource usage - i.e. saving some mod slots by using existing hard wired mod pathes.

so it may possible to get even better results if further experimenting with. try it - the patch could be dowloaded.

have fun with the patch...

due to the limits of the synth engine the patch sounds at best around C3 (anywhere between G2 and G3) to my ears.

Top