this is a combination of speech to text and text to speech it is mainly copied from here: https://pythonspot.com/speech-recognition-using-google-speech-api/
it works offline with sphinx too, but then it is less reliable.

import speech_recognition as sr
from gtts import gTTS
import vlc
import socket
s = socket.socket()
host = socket.gethostname()
port = 3000
s.connect((host, port))

while (True == True):
  r = sr.Recognizer()
  with sr.Microphone() as source:
    r.adjust_for_ambient_noise(source, duration=1)
    print("Say something!")
    audio = r.listen(source,phrase_time_limit=5)
 
  try:
    pdMessage = r.recognize_google(audio, language="en-US") + " ;"
    message = r.recognize_google(audio, language="en-US")
    s.send(pdMessage.encode('utf-8'))
    print("PD Message: " + message)
    tts = gTTS(text = str(message), lang="en-us")
    tts.save("message.mp3")
    p = vlc.MediaPlayer("message.mp3")
    p.play()
  except sr.UnknownValueError:
    print("Google could not understand audio")
  except sr.RequestError as e:
    print("Google error; {0}".format(e))

What I would like to achieve: I have an CSV file with ~60000 quotes. The recognized word is send with [netreceive] to Pure Data and the patch chooses randomly one of the quotes where the recognized word appears. That does work.
My question: Is it possible to send the choosen quote with [netsend] back to python and transform it there into speech?
Somebody says a word and the computer answers with a quote where the word appears...
Does it make sense to use Pure Data for that task or better just use Python (but I do not know how to do that only with Python yet...)?
The Universal Sentence Encoder sounds really promising for "understanding" the meaning of a sentence and finding the most similar quote. But that is far too complicated for me to implement...
https://www.learnopencv.com/universal-sentence-encoder/

This is my latest patch:

prac.pd

sampler.pd

It needs any wav file beginning with a number such as "1.wav".

The problem I'm having is that it doesnt run efficiently and Im not sure what to do.
It plays back a sound file but when I need to load another file the whole patch freezes so much so that it stops the audio play for a moment. Its very annoying becuase I've spent a long time figuring out the patch and now It works it at the same time doesnt. Please may I have some help improving this to make it usable?

Many thanks,

J
Here are the samples I'm using with the patch
2.wav.zip 1.wav.zip

Waves observed in liquid, in a string, in a drum head etc (material waves) are functions of time and space while audio are only time-dependant. As a result you can derive audio signals from a material wave for by fixing a position and observing a property of the material wave at this dertermined position over time. For example, let's imagine a little fishing float maintained that would "freely" slide on a vertical rod maintained in a fixed position in water (in order to get rid of the space variables). The float would go up and down over time as the waves come and go, and the vertical position signal obtained would be a an audio signal (provided it's mean over time is zero), also all the derivatives of this signal accompanied by frequency multiplication and amplitude modifications, but also pressure or desity measurements or whatever you want. This is a way of thinking the link between material waves (the ones you draw, with the horizontal axis representing space, with a freezed time) and strictly time-dependent signals (very near to audio signal, just a matter of being "centered" on the x axis).
This proximity allows for "sonification" of intrinsically non-audio phenomena, like "hearing a black hole squeaking" etc.

Sorry, now that I'm reading back what I've written I feel like all this is probably obvious to you, or simply useless.

Nau

As to an audio wave not being able to curl back, it can if the wave form is looked at as the sum of 3 wave forms where the value between t(or x)=a,b,c, or d are flattened to 0. So the audio signal is not one osc~ but rather 3 truncated ones then summed [+~].

At least that would represent (visually) what a nautical wave does.

Which is why I am so intrigued. Nautical waves "do" it. But have not seen nor heard audio waves do so.

my current thinking is...

for the static version of just sound cresting use 3 wave tables assembled with the 3 shapes representated below (sin(x=0 to 3/2pi), sin(x=(pi to 3/2pi)/2offset pi), and sin(x=pi-2pi).

as to whether this will auditorially speaking result in 0, I am unconcerned, i.e. just sum them and see what it sounds like/is it discernibly different from just osc.

if someone, can tell me how to get those wave tables, that would be great. thanks, in advance.

note: also found this:

http://teachersinstitute.yale.edu/curriculum/units/2008/5/08.05.06.x.html

which may lead us to a possible Dynamic osc~ of breaking-wave form, since I am pretty sure we can do this using [osc~] plus some fancy audio arithmetic, ex. a timer to [*~ 0] where any of the 3 forms are not present, etc.

thanks, also for any and all thought put into it.

just stumps me, why on one hand Nature makes perfect osc sound but does not do the same thing for liquids (matter?).

peace and love. may your day be blessed.
-s

@whale-av I love it how the post is named "16 parameters for one voice" and ended up in the extreme...hahah, 1_vod_12_2560 is pretty neat, sounds a lot like an organ, and yes the end is pretty interesting, hearing the different patterns of the sine waves. it would be nice to play with their intensities while there are no curves left, to see how different timbres are formed, i dont think exclusion is the way but rather integrating everything, a time for curves then those curves could sit still, the glissandi, then microtonal, its the best.
but what do you mean a curve for each sine? that would be a lot of curves, wouldnt it be better for one curve to control all their intensities? like the way you had done: each sine with a graph and only one curve controlling their position on the graph(maybe a graph generator to generate for each sine a graph, or for example: 2000 graphs to be followed by all the sine waves). or like the way we did way back when forming a list to pass through combinations of the sine´s volumes.
BtW, vod_13_20480 and vod_14_50960 didnt work, i just heard sometimes like glitchy noise, like a "popping".

@H.H.-Alejandro Hello Ale.
Not finished....... just for testing..... Ale4.zip

All effects and audio meters etc. removed...... and only the first 2560 sines have micro adjustment.
In 1_vod_12_2560 it takes curves 3 to 18.
When running iannix I seem to get some distortion...... but if I stop sending the curves from iannix it is better.
This suggests that the messages are interfering with the audio..... but I am not sure...... in fact I think not....... something else.... maybe high frequency beating.
When you stop iannix....... or it gets to the end of the curves.... you can hear the beating at different rates of the sine waves....... it is interesting..... but starting the curves again makes it less audible.

Maybe massive sines controlled by curves are less interesting than massive sines on their own?
I will build one with only one curve for every sine..... to hear what Pieter suurmond expected..... so with a fixed note.

vod_13_20480 and vod_14_50960 are there for you to try..... (20480 sines and 50960 sines).
They take a while to load...... and for me they just distort..... but they really are a massive "ask" of the computer.
As I said above they are not finished. A lot of work needs to be done for all the curves to be used....... and as I think they will not work I have not done the work yet.

I really like 1_vod_12_2560 when iannix stops...... it is endlessly changing and interesting......
David.

(First, thanks, @seb-harmonik.ar for the tip. Really appreciate it.)

Originally I thought of doing this just as sort of challenge, to see if I could get it worked out and thought given the current pantheon of wave generators perhaps we could use another one.

What it does is invert the curvature of the osc~ wave form so in a quadrant of the wavelength where it normally curves down, here it curves up, and vice versa for the other two quadrants of the wave.

(more pics below)

oscinv~-help.pd
oscinv~.pd

It's accomplished by aligning the original osc~ with two square (phasor~) waves: the first to get the shape right (one wavelength for each quadrant of the original wave (so 4x the freq) and a second square wave to shift the wave up and down in the quadrants that need it.

The help file is setup to allow you to hear the difference between the osc~ and the oscinv~. (Note: if you change on the fly, the table does not write properly (i think (unless someone can tell me otherwise) because of the discrepancy/lag between changing the frequency value and the writing of the table).

The resulting tone is Very different than the original pure osc~ tone and it seems to me it has some sort of resonance(?).

If not usable, I hope you enjoy the thinking that went into it.

(It was a helluva of a lot of fun to work on. *Once I got the sucker! ).

Peace, happy pd-ing to one and all.

Sincerely,
Scott

p.s. oh, it has one inlet (frequency) and one creation argument (also frequency) and sends out a ~ signal.

p.p.s my thinking is those who make and/or use synths (esp.) might have some use for it.

band pass filters are kind of like exponentially dampened sine waves. keeping this in mind, try recording 4 sine waves of equal amplitudes added at those exact frequencies and look at the waveform, it's very similar. It's just a result of adding those frequencies together

in this case it is a phase issue because the band pass filters all have the same kind of phase response, and when the frequencies are that close together they will combine to create a gibbs-like effect in that first part

imagine adding a bunch of sine waves together: if they are positive, then the first bump will be tall because the first bump of each sine wave adds together. However, as they progress they interfere destructively after that more

as for how to change it, maybe you could try just inverting the polarity ([*~ -1]) of every other filter output in order of frequency?

@whale-av Did you get the part about the number of sine waves needed? EACH morphing timbre will use 5 sine waves (the volume of the 5 sine waves will fuse and make ONE sound, these 5 sine waves are for each timbre curve, there are 31 curves for timbre so 5 times 31 is 155 sine waves...)