OpenAI_TTS_New / app.py
shurikenji's picture
Update app.py
9873378 verified
import gradio as gr
import os
import tempfile
import re
from pydub import AudioSegment # Library to combine audio files
from openai import OpenAI
# Max character limit per API request
MAX_CHAR_LIMIT = 4096
def clean_text(text):
# Replace newlines with spaces and multiple spaces with a single space
cleaned_text = re.sub(r'\s+', ' ', text.strip()) # Replace multiple spaces and newlines with a single space
return cleaned_text
def split_text(text, limit=MAX_CHAR_LIMIT):
# Split text into chunks of <= MAX_CHAR_LIMIT characters
words = text.split(' ')
chunks = []
current_chunk = ""
for word in words:
# Add words to the current chunk without exceeding the character limit
if len(current_chunk) + len(word) + 1 <= limit: # +1 for space
current_chunk += word + " "
else:
chunks.append(current_chunk.strip()) # Append the current chunk
current_chunk = word + " " # Start a new chunk
if current_chunk:
chunks.append(current_chunk.strip()) # Add the last chunk
return chunks
def tts(text, model, voice, speed, api_key, base_url):
if api_key == '':
raise gr.Error('Please enter your Key')
cleaned_text = clean_text(text)
chunks = split_text(cleaned_text)
audio_segments = []
try:
client = OpenAI(api_key=api_key, base_url=base_url+'/v1') # Use selected base_url
# Process each chunk of text
for chunk in chunks:
response = client.audio.speech.create(
model=model, # "tts-1", "tts-1-hd"
voice=voice, # 'alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'
input=chunk,
speed=speed
)
# Create a temp file to save the audio for each chunk
with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as temp_file:
temp_file.write(response.content)
temp_file_path = temp_file.name
audio_segments.append(AudioSegment.from_mp3(temp_file_path))
except Exception as error:
raise gr.Error("An error occurred while generating speech. Please check your API key and try again.")
# Concatenate all audio chunks into one final audio file
final_audio = sum(audio_segments)
# Save the concatenated audio to a final file
with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as final_temp_file:
final_audio.export(final_temp_file.name, format="mp3")
final_audio_path = final_temp_file.name
return final_audio_path
with gr.Blocks() as demo:
gr.Markdown("# <center> OpenAI TTS Unlimited Character </center>")
with gr.Row(variant='panel'):
api_key = gr.Textbox(type='password', label='OpenAI API Key', placeholder='Enter your API key to access the TTS demo')
model = gr.Dropdown(choices=['tts-1', 'tts-1-hd', 'tts-1-1106', 'tts-1-hd-1106'], label='Model', value='tts-1')
voice = gr.Dropdown(choices=['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'], label='Voice Options', value='alloy')
speed = gr.Slider(minimum=0.5, maximum=2.0, step=0.1, label="Speed", value=1.0)
# Add dropdown for URL selection
base_url = gr.Dropdown(choices=['https://gpt1.shupremium.com', 'https://gpt1.shupremium.com','https://gpt2.shupremium.com','https://gpt3.shupremium.com' ,'https://gpt4.shupremium.com', 'https://gpt5.shupremium.com'], label="API Endpoint", value='https://gpt5.shupremium.com')
text = gr.Textbox(label="Input text", placeholder="Enter your text and then click on the 'Text-To-Speech' button, or simply press the Enter key.")
char_counter = gr.Markdown("Character count: 0")
btn = gr.Button("Text-To-Speech")
output_audio = gr.Audio(label="Speech Output")
def update_char_counter(text):
cleaned_text = clean_text(text) # Clean the text by removing extra spaces and newlines
return f"Character count: {len(cleaned_text)}"
text.change(fn=update_char_counter, inputs=text, outputs=char_counter)
text.submit(fn=tts, inputs=[text, model, voice, speed, api_key, base_url], outputs=output_audio, api_name="tts_enter_key", concurrency_limit=None)
btn.click(fn=tts, inputs=[text, model, voice, speed, api_key, base_url], outputs=output_audio, api_name="tts_button", concurrency_limit=None)
demo.launch()