Forum - Socoder

-=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- (c) WidthPadding Industries 1987 0\|720\|0 -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=-
Socoder -> AI Threads -> Machine Made Max
Page : 1 Prev 2 3 Next 4 5 6 7 8
Fri, 10 Mar 2023, 13:53
cyangames	That's really cool spinal, nice work! Reminds me I need to finish up that Raspberry PI project sometime, err, along with everything else, ha! -=-=- Web / Game Dev, occasionally finishes off coding games also!
Fri, 10 Mar 2023, 14:16
cyangames	Slightly related, have been chatting to the M3GAN bot via DM on twitter, much fun <3 I do love robo waifus -=-=- Web / Game Dev, occasionally finishes off coding games also!
Mon, 13 Mar 2023, 05:40
spinal	I think I need to take some time an start a second thread on this max headroom topic. I managed to get the Google text to speech working to see if a different voice would sound better. But they have less options available, still stuck with female, can only convert to MP3, so other parts of the code are incompatible, so I had to introduce MP3 to wav conversation using ffmpeg. That slowed everything down, and although the voice is more natural sounding, it doesn't work as good. -=-=- Check out my excellent homepage!
Mon, 13 Mar 2023, 07:42
Jayenkai	Good grief!!
Mon, 13 Mar 2023, 11:46
spinal	Here is the google tts test... 2023/11/maxine_headspace_google_tts (.mp3) More natural, but more wrong. -=-=- Check out my excellent homepage!
Mon, 13 Mar 2023, 12:33
Jayenkai	Hmmm. I'm not really sure on the b-b-best way to go go go go about it, if I'm honest-t-t-t-t-t-t-taaa. -=-=- ''Load, Next List!''
Mon, 13 Mar 2023, 14:25
spinal	I think I need to use the same trick for the stutter that I'm using for the normal speech. I'm adding an extra word of a known length to then, then trimming it from the converted file, so the TTS engine doesn't assume it's at the end of a sentence. -=-=- Check out my excellent homepage!
Mon, 13 Mar 2023, 22:36
Jayenkai	That sounds like a good way to do it. Have you experimented with pitch-shifting, too? Obviously, you'll want to do this to some of the stutters, but I wonder how different it might sound if the voice's pitch was reduced significantly, then scaled back up by pitch-shifting the audio, later? -=-=- ''Load, Next List!''
Tue, 14 Mar 2023, 00:36
spinal	Pitch shifting might be a pain, it doesn;t look possible to adjust the pitch mid sentence, so each word would have to be adjusted the same way as the stutter. First I want a text to speech engine that is (A) not too slow (B) has a range of voices that are easy to set up and © free. -=-=- Check out my excellent homepage!
Tue, 14 Mar 2023, 01:24
Jayenkai	> Reveal 🔎 Apologies for the BBCode parser turning the ( C ) into a ©, there. .. I forgot I added that. That's kinda handy, though!! There's an ®, too! How about the Microsoft internal voices? Are those any use? -=-=- ''Load, Next List!''
Tue, 14 Mar 2023, 11:07
spinal	I've just managed to get the microsoft voices working correctly, turns out the python tts engine looks for them in a different place to where they are (now?) you have to installthem through the narrator options in control panel, then go into regedit, export the settings, change the path, import them again. Now I have all the english voices ready for testing. -=-=- Check out my excellent homepage!
Tue, 14 Mar 2023, 11:15
Jayenkai	Woohoo!
Tue, 14 Mar 2023, 15:52
spinal	I think I've got the speaking much closer to what I'm aiming for... 2023/11/max_test_14_3_23 (.mp3) ''' To install more voices (they WILL be needed) Open Regedit and navigate to Computer\HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech_OneCore\ Right click 'Voices' and export the file. Open the file in Notepad++ or other text editor, then find+replace Speech_OneCore with Speech. Save the file, then double-click it to inport the keys back into the registry. This will import all of your installed voiced into the location that pyttsx3 is looking for. ''' import random import time import wave import tempfile import os import numpy as np import pyttsx3 import pyaudio import re import pygame import sys import array from datetime import datetime random.seed(datetime.now().timestamp()) screen = pygame.display.set_mode((320, 240)) clock = pygame.time.Clock() s_num = 0 def draw_circle(screen, x, y, radius): screen.fill((0,0,0)) pygame.draw.circle(screen, (255, 0, 0), (x, y), radius) def speak(text): # Initialize the text-to-speech engine engine = pyttsx3.init('sapi5') # Set the text-to-speech parameters engine.setProperty('rate', 200) engine.setProperty('volume', 1.0) engine.setProperty('pitch', random.uniform(0.9, 1.1)) # Set pitch to random value voices = engine.getProperty('voices') engine.setProperty('voice', voices[10].id) # Generate a temporary WAV file to save the audio data with tempfile.NamedTemporaryFile(delete=False) as f: file_path = f.name # Speak the text and save the audio data to the temporary WAV file engine.save_to_file(text, file_path) # Start the speech engine in a separate thread engine.startLoop(False) # Wait for the engine to finish speaking while engine.iterate(): time.sleep(0.1) Print('Working...') # Stop the speech engine engine.stop() # Read the audio data from the temporary WAV file with wave.open(file_path, 'rb') as wav_file: audio_data = np.frombuffer(wav_file.readframes(-1), dtype=np.int16) audio_rate = wav_file.getframerate() audio_format = wav_file.getsampwidth() start_trim = 0 duration = len(audio_data)-1 end_trim = duration threshold = 10 while audio_data[start_trim] < threshold and start_trim < duration : start_trim = start_trim + 1 while audio_data[end_trim] < threshold and end_trim > 0 : end_trim = end_trim - 1 #string = str(start_trim)+','+str(duration)+','+str(end_trim) #print(string) # Remove the temporary WAV file os.remove(file_path) return audio_data[start_trim:end_trim], audio_rate, audio_format def stutter(word, end_trim): # Speak the word and get the audio data, rate, and format audio_data, audio_rate, audio_format = speak(word) # Convert the audio data to a numpy array audio_data = np.frombuffer(audio_data, dtype=np.int16) # Choose a random position to stutter from #position = random.uniform(0.0, 0.5) # Get the duration of the word's audio duration = (len(audio_data)-end_trim) / audio_rate # Calculate the start and end times of the stuttered audio start_time = 0 end_time = duration * random.uniform(0.4, 0.6) if len(word) <= 5: start_time = 0 end_time = duration # Calculate the start and end frames of the stuttered audio start_frame = int(start_time * audio_rate) end_frame = int(end_time * audio_rate) # Extract the stuttered audio and repeat it stuttered_audio = audio_data[start_frame:end_frame] #repeated_audio = np.concatenate([stuttered_audio] * 2) repeated_audio = array.array('h') stutter_type = random.randint(0, 1) if stutter_type == 0: spd = 105 sz = int(((len(stuttered_audio)-1)100)/spd) for x in range(sz): pos = int((xspd)/100) num = stuttered_audio[pos] repeated_audio.append(num) spd = 110 sz = int(((len(stuttered_audio)-1)100)/spd) for x in range(sz): pos = int((xspd)/100) num = stuttered_audio[pos] repeated_audio.append(num) spd = 115 sz = int(((len(stuttered_audio)-1)100)/spd) for x in range(sz): pos = int((xspd)/100) num = stuttered_audio[pos] repeated_audio.append(num) if stutter_type == 1: spd = 100 sz = int(((len(stuttered_audio)-1)100)/spd) for x in range(sz): pos = int((xspd)/100) num = stuttered_audio[pos] repeated_audio.append(num) for x in range(sz): pos = int((xspd)/100) num = stuttered_audio[pos] repeated_audio.append(num) return repeated_audio, audio_rate, audio_format def stutter_text(text): # Split the text into sentences sentences = re.split('[.?!\n]', text) #print(text) # put a word after every text, to make the tts sound like it's continueing a sentence, then remove that word from the audio normal_audio_data, audio_rate, audio_format = speak("test") test_size = int(len(normal_audio_data)) # Iterate over each sentence and stutter some of the words audio_data_list = [] for sentence in sentences: sentence = sentence + '.' print(sentence) # Split the sentence into words words = sentence.strip().split() # Check if there are any words in the sentence if len(words) == 0: continue # Process each word in the sentence i = 0 while i < len(words): # Choose a random number of words to speak normally num_normal_words = random.randint(5, 10) if i + num_normal_words > len(words): num_normal_words = len(words) - i # Speak the first few words normally normal_text = ' '.join(words[i:i+num_normal_words]) normal_text = normal_text + ' test ' # the word 'test' is removed from the audio afterwards. normal_audio_data, audio_rate, audio_format = speak(normal_text) duration = len(normal_audio_data)-test_size audio_data_list.append(normal_audio_data[0:duration]) # Check if there are any words left to stutter if i + num_normal_words >= len(words): break # Stutter the next word temp_audio_data, audio_rate, audio_format = speak(" test") test_size = int(len(temp_audio_data)) stuttered_word = words[i+num_normal_words] stuttered_word = stuttered_word + ' test ' stuttered_audio_data, audio_rate, audio_format = stutter(stuttered_word, test_size) # Repeat the stuttered word repeat_audio_data, audio_rate, audio_format = speak(stuttered_word) # Get the duration of the word's audio duration = (len(normal_audio_data)-test_size) / audio_rate # Calculate the start and end times of the stuttered audio start_time = 0 end_time = duration # Calculate the start and end frames of the stuttered audio start_frame = int(start_time audio_rate) end_frame = int(end_time * audio_rate) # Concatenate the audio data and append it to the list audio_data_list.append(stuttered_audio_data[start_frame:end_frame]) # Get the duration of the word's audio duration = (len(repeat_audio_data)-test_size) / audio_rate # Calculate the start and end times of the stuttered audio start_time = 0 end_time = duration # Calculate the start and end frames of the stuttered audio start_frame = int(start_time * audio_rate) end_frame = int(end_time * audio_rate) audio_data_list.append(repeat_audio_data[start_frame:end_frame]) # Move to the next word i += num_normal_words + 1 # Concatenate all the audio data and write it to a stream audio_data = np.concatenate(audio_data_list) CHUNK = 1024 FORMAT = pyaudio.paInt16 CHANNELS = 1 RATE = audio_rate player = pyaudio.PyAudio() stream = player.open(format=FORMAT, channels=CHANNELS, rate=RATE, input=True, frames_per_buffer=CHUNK, output=True) stream.write(audio_data.tobytes()) # Write entire chunk of audio data to stream stream.stop_stream() stream.close() player.terminate() # Save the audio as a raw pcm file global s_num s_num += 1 file_name = 'sentence_{}_{}.pcm'.format(s_num, audio_rate) with open(file_name, 'ab') as f: f.write(audio_data) def stutter_text_from_file(file_path): # Read the input text from the file with open(file_path, 'r') as f: text = f.read() # Split the text into sentences sentences = re.split('[.?!\n]', text) # Process each sentence for i, sentence in enumerate(sentences): # Check if this is the last sentence and it doesn't end with punctuation if i == len(sentences) - 1 and sentence.strip() != '': # Add a period to the end of the sentence sentence = sentence.strip() + '.' # Check if the sentence is not empty if len(sentence.strip()) > 0: stutter_text(sentence.strip()) if __name__ == '__main__': # Get the path to the input file file_path = 'readme.txt' # Stutter the text from the file stutter_text_from_file(file_path) --v -=-=- Check out my excellent homepage!
Tue, 14 Mar 2023, 15:58
Jayenkai	Cool That's working really nicely!
Wed, 15 Mar 2023, 09:58
spinal	2023/11/max_test_15_3_23 (.mp3) ðŸ‘ -=-=- Check out my excellent homepage!
Wed, 15 Mar 2023, 10:14
Jayenkai	Yup!
Thu, 16 Mar 2023, 03:26
spinal	Got gfx doing something! Don't fall off your chair at the complexity of the CGI... View on YouTube -=-=- Check out my excellent homepage!
Thu, 16 Mar 2023, 03:30
Jayenkai	... Needs to be a taller head with some hair, I think. Looks more like Zippy!! At least the lip-sync kinda works. Keep going -=-=- ''Load, Next List!''
Thu, 16 Mar 2023, 06:07
Jayenkai	moves thread to it's own... .. thread If you spot any irregularities in the selection of posts that I've moved over into this thread, please do let me know. -=-=- ''Load, Next List!''
Fri, 17 Mar 2023, 04:57
spinal	Next update will be a little better, I just added openAI. FYI, it's bloody useless, unless you just want it to spout random nonsense. I believe it's possible to access Google assistant from python, but I bet it's difficult with lots of hoops to jump through and a fee of some sort. -=-=- Check out my excellent homepage!
Fri, 17 Mar 2023, 05:43
Jayenkai	Sounds like you're almost there, I wonder why OpenAI haven't yet added speech and faces? -=-=- ''Load, Next List!''
Fri, 17 Mar 2023, 14:04
spinal	View on YouTube
Fri, 17 Mar 2023, 15:31
Jayenkai	Daaamn... And you only started this last Sunday in a language you hadn't previously used??! See, now THIS is the power of AI It's about using it, and learning with it. Fantastic!!! -=-=- ''Load, Next List!''
Fri, 17 Mar 2023, 15:54
spinal	I had used micropython a couple of years ago, but all I can really remember is that python is basic-ish and uses indents for loops and IFS instead of curly brackets. As for me learning, that isn't really happening. It's a case of tell the AI "I want to do xyz", the AI then spits out some code, I run it, it fails, I tell it, it gives me another idea. It's just a faster way to Google something. [Edit] oh, wow, pyhton has a Wikipedia API..... -=-=- Check out my excellent homepage!
Sun, 19 Mar 2023, 07:39
spinal	OK, now that I have speach working quite well, I need to work on the face some more. So after a few attempts to get chatGPT to draw me some simple concentric circles that are layered and move in parallax according to the mouse x location, with a few "no that's wrong" and rewording the question a few times, then eventually manually moving a couple of lines of code around, I got this... import pygame # Initialize Pygame pygame.init() # Set up the screen screen_width = 640 screen_height = 480 screen = pygame.display.set_mode((screen_width, screen_height)) pygame.display.set_caption("Circle Example") # Set up colors WHITE = (255, 255, 255) RED = (255, 0, 0) BLUE = (0, 0, 255) # Set up circle parameters circle_radius = 20 num_circles = 7 # Set up the clock clock = pygame.time.Clock() # Main game loop done = False while not done: # Handle events for event in pygame.event.get(): if event.type == pygame.QUIT: done = True # Fill the screen with white screen.fill(WHITE) # Get the center and mouse positions center_x = screen_width / 2 center_y = screen_height / 2 mouse_x, mouse_y = pygame.mouse.get_pos() # Calculate the spacing between the center and mouse circles spacing = abs(mouse_x - center_x) / (num_circles - 1) # Draw the circles pygame.draw.circle(screen, RED, (int(center_x), int(center_y)), circle_radius2) for i in range(num_circles - 2): if mouse_x > center_x: x = center_x + spacing (i + 1) else: x = center_x - spacing * (i + 1) pygame.draw.circle(screen, (255,i25,0), (int(x), int(center_y)), (circle_radiusnum_circles)/(i+1)) pygame.draw.circle(screen, BLUE, (int(mouse_x), int(center_y)), circle_radius) # Update the screen pygame.display.flip() # Wait for the next frame clock.tick(60) # Quit Pygame pygame.quit() --v This is going to be the basis for the face movement animation, essentially 2d layers that can move around to give the effect of a well drawn face... -=-=- Check out my excellent homepage!
Page : 1 Prev 2 3 Next 4 5 6 7 8