Made for: https://youtu.be/E3gf88rSzqo

viztoaudio.py

## NOTE
## I'm only releasing this code for educational purposes
## Do not abuse it please
## And this is terrible code, so don't use this as a reference on how to code properly
## Made for: https://youtu.be/E3gf88rSzqo
from PIL import Image
from PIL import ImageFilter
import math
import wave
import struct
import time
import os

video_framerate = 30 #Framerate of the input image sequence, you can and should use floating point numbers here, if required
seq_folder_name = "sequence" #Name of the folder where your image sequence is in
wav_folder_name = "wavs" #Name of the folder where the wavs are going to go to
max_freq = 4500 #Max frequency your audio will go up to (if you set this too low the program will crash)

#Some weird audio stuff I don't understand
nchannels = 1  #Amount of channels in audio. Keep this at 1
sampwidth = 2
framerate = 44100 #Sample rate, you'll probably want to keep this as is
nframes = int(framerate/video_framerate)
comptype = "NONE"
compname = "not compressed"
audiodata = {}

def calculateHz(w):
	out = 0
	outplus = 0
	for x in [(84,0,100)]: #insert segments in the following format (in pixels from start) -> (end of the current segment,start of the current segment,how many hz this segment represents)
		out = convertToHz(x,w) #Have this thing give us the Hz
		if out > -1: break #That thing figured out the Hz, break outta loop
		outplus += x[2] #Add the Hz from this segment and move over to the next one
	return out + outplus
	
def convertToHz(x,w): #Do quick maffs to figure out the Hz
	a,b,c = x #To be fair, even I don't understand this part of the code anymore
	d = a-b #But it should be self-explanatory
	out = -1 #So I'm not going to comment on
	if w < a: #What this part actually
		out = (w-b)/d*c #does...
	return round(out) #Round the Hz value so it's nice and smooth

def calculatedB(h):
	db = ((h - 146) / 934) * -120 #146 is how many pixels from the top the audio visualization starts, 934 is how many pixels are there after that and -120 is the decibel range (in this case, from 0dB to -120dB)
	return round(db) #Round the dB value so it's nice and smooth

for x in range(1, max_freq):
	audiodata[x] = [] #Fill our dict with pretty much nothing

i = 0
for img_file in os.listdir(seq_folder_name): #Go through all the image files
	i+=1
	print(os.path.splitext(img_file)[0])
	im		= Image.open(os.path.join(seq_folder_name,img_file))
	pix		= im.load()
	width, height = im.size
	for w in range(width): #Go trough every pixel horizontally
		for h in range(height): #Go trough every pixel vertically
			if not pix[w,h] == (0,0,0): #Detect all pixels that aren't 100% black
				#Uncomment the following line to see the Hz/dB output
				#print(str(calculateHz(w)) + "hz : " + str(calculatedB(h)) + "db")
				audiodata[calculateHz(w)].append(calculatedB(h)) #Calculate the data and add it to our dict
				break #Don't look down vertically anymore, we aready have the data we need 
	for x in range(1,max_freq): #This part just fixes stuff but it might be broken, I'm not sure
		if len(audiodata[x]) < i: audiodata[x].append(-120)
for sound,sounds in audiodata.items(): #Go through all the values stored in the dict
	print(sound)
	frequency = sound
	wav_file = wave.open(wav_folder_name + "\\" + str(frequency) + 'hz.wav', 'w')
	wav_file.setparams((nchannels, sampwidth, framerate, nframes, comptype, compname))
	for volume in sounds:
		amplitude = 10**((100+volume)/20)# 100000 is max value; convert volume from dB to raw value
		try:
			for i in range(nframes):
				#Honestly don't know what this for loop does, I just copy-pasted it from stackoverflow
				#I mean, it creates sinewaves at the frequency and amplitude we chose, but I dunno how it works
				#You don't actually need to know how this part works as the program still works regardless of if you understand how it works or not
				sample = math.sin(2*math.pi*frequency*(float(i)/framerate))*amplitude/2
				wav_file.writeframes(struct.pack('h', int(sample))) 
		except:
			print("Error at: " + sound)
	wav_file.close()