import json
import asyncio
import edge_tts
from pydub import AudioSegment
import os
import gradio as gr
from gradio_client import Client
import shutil
import uuid
from dotenv import load_dotenv
import re
load_dotenv()
def sanitize_filename(filename):
"""Convert a string to a safe filename by removing special characters and spaces"""
safe_filename = re.sub(r'[^a-zA-Z0-9_-]', '', filename.replace(' ', '_'))
return safe_filename.lower()[:50]
async def get_voices():
"""Get all available English voices from edge-tts"""
voices = await edge_tts.list_voices()
english_voices = [
voice for voice in voices
if voice["Locale"].startswith(("en-US", "en-GB", "en-AU", "en-CA", "en-IN"))
]
formatted_voices = [
f"{voice['ShortName']} ({voice['Gender']}, {voice['Locale']})"
for voice in english_voices
]
return formatted_voices
def extract_voice_name(voice_string):
"""Extract the voice short name from the formatted string"""
return voice_string.split(" (")[0]
async def generate_audio(text, voice, filename):
communicate = edge_tts.Communicate(text, extract_voice_name(voice))
await communicate.save(filename)
async def create_podcast_version(data, speaker_name, speaker_voice, title):
session_id = str(uuid.uuid4())
temp_dir = f'temp_{session_id}'
safe_title = sanitize_filename(title)
if not os.path.exists(temp_dir):
os.makedirs(temp_dir)
try:
speaker_version = AudioSegment.empty()
for i, entry in enumerate(data['conversation']):
if 'speakertext' in entry:
temp_file = f'{temp_dir}/speaker_{i}.mp3'
await generate_audio(entry['speakertext'], speaker_voice, temp_file)
audio = AudioSegment.from_file(temp_file)
speaker_version += audio
os.remove(temp_file)
speaker_path = f"{safe_title}_{speaker_name.lower()}_only.mp3"
speaker_version.export(speaker_path, format="mp3")
return speaker_path, temp_dir
except Exception as e:
if os.path.exists(temp_dir):
shutil.rmtree(temp_dir)
raise e
def generate_podcast(title, channel_name, speaker_name, speaker_voice):
try:
if not all([title, channel_name, speaker_name, speaker_voice]):
raise ValueError("All fields must be filled out")
client = Client(os.getenv('API_URL'))
result = client.predict(
message=f"""{os.getenv('API_MESSAGE')} {{
"title": "{title}",
"channel": "{channel_name}",
"speaker": "{speaker_name}",
"conversation": [
{{
"speakertext": ""
}}
]
}}
give 36 sentences for the speaker.
""",
request=os.getenv('API_REQUEST'),
param_3=0.5,
param_4=8100,
param_5=0.5,
param_6=0,
api_name="/chat"
)
try:
podcast_data = json.loads(result)
except json.JSONDecodeError:
json_start = result.find('```') + 3
json_end = result.rfind('```')
if json_start > 2 and json_end > json_start:
if result[json_start:json_start+4] == 'json':
json_start = result.find('\n', json_start) + 1
json_str = result[json_start:json_end].strip()
podcast_data = json.loads(json_str)
else:
raise ValueError("Could not parse JSON from response")
speaker_path, temp_dir = asyncio.run(
create_podcast_version(
podcast_data,
speaker_name,
speaker_voice,
title
)
)
if os.path.exists(temp_dir):
shutil.rmtree(temp_dir)
return [speaker_path, podcast_data]
except Exception as e:
return [None, f"Error: {str(e)}"]
with gr.Blocks(theme=gr.themes.Soft()) as interface:
available_voices = asyncio.run(get_voices())
gr.Markdown("# Easy Podcast (Single Speaker)")
gr.Markdown("Generate a podcast with one speaker. Choose a voice and customize details to create your perfect audio.
To use elevelabs voices or cloned voices, or to automate the podcast video creation with avatar contact me at aheedsajid@gmail.com
Support me USDT (TRC-20) (TAe7hsSVWtMEYz3G5V1UiUdYPQVqm28bKx)")
with gr.Row():
with gr.Column():
title = gr.Textbox(
label="Podcast Topic",
placeholder="e.g., The Future of AI",
show_label=True
)
channel_name = gr.Textbox(
label="Channel Name",
placeholder="e.g., TechTalks",
value="WeePakistan",
show_label=True
)
with gr.Column():
speaker_name = gr.Textbox(
label="Speaker Name",
placeholder="e.g., John",
value="Andrew",
show_label=True
)
with gr.Row():
speaker_voice = gr.Dropdown(
choices=available_voices,
value=next((v for v in available_voices if "Andrew" in v), available_voices[0]),
label="Speaker Voice",
info="Select voice for the speaker"
)
generate_btn = gr.Button("Generate Podcast", variant="primary")
with gr.Row():
speaker_audio = gr.Audio(label="Speaker Audio")
conversation_json = gr.JSON(label="Generated Conversation")
generate_btn.click(
fn=generate_podcast,
inputs=[title, channel_name, speaker_name, speaker_voice],
outputs=[speaker_audio, conversation_json]
)
if __name__ == "__main__":
interface.launch()