Create app.py

app.py

@@ -0,0 +1,108 @@
+# app.py
+
+import gradio as gr
+import torch
+import torchaudio
+import google.generativeai as genai
+from transformers import AutoProcessor, AutoModel
+import numpy as np
+import os
+import json
+
+# Initialize Gemini AI
+genai.configure(api_key='YOUR_GEMINI_API_KEY')
+model = genai.GenerativeModel('gemini-pro')
+
+# Initialize F5-TTS model
+processor = AutoProcessor.from_pretrained("SWivid/F5-TTS")
+f5_model = AutoModel.from_pretrained("SWivid/F5-TTS")
+
+def clone_voice(audio_file):
+    waveform, sample_rate = torchaudio.load(audio_file)
+    # Implement voice cloning logic here
+    # For demonstration, we'll just return a placeholder
+    return "Voice cloned successfully"
+
+def generate_podcast_script(content, duration):
+    prompt = f"""
+    Create a podcast script for two people discussing the following content:
+    {content}
+    
+    The podcast should last approximately {duration}. Include natural speech patterns,
+    humor, and occasional off-topic chit-chat. Use speech fillers like "um", "ah",
+    "yes", "I see", "Ok now". Vary the emotional tone (e.g., regular, happy, sad, surprised)
+    and indicate these in [square brackets]. Format the script as follows:
+
+    Host 1: [emotion] Dialog
+    Host 2: [emotion] Dialog
+    
+    Ensure the conversation flows naturally and stays relevant to the topic.
+    """
+    response = model.generate_content(prompt)
+    return response.text
+
+def text_to_speech(text, speaker_id):
+    inputs = processor(text=text, return_tensors="pt")
+    speech = f5_model.generate_speech(inputs["input_ids"], speaker_id=speaker_id, vocoder_output=True)
+    return speech.cpu().numpy()
+
+def create_podcast(content, duration, voice1, voice2):
+    script = generate_podcast_script(content, duration)
+    lines = script.split('\n')
+    audio_segments = []
+    
+    for line in lines:
+        if line.startswith("Host 1:"):
+            audio = text_to_speech(line[7:], speaker_id=0)  # Assuming speaker_id 0 for Host 1
+            audio_segments.append(audio)
+        elif line.startswith("Host 2:"):
+            audio = text_to_speech(line[7:], speaker_id=1)  # Assuming speaker_id 1 for Host 2
+            audio_segments.append(audio)
+    
+    # Concatenate audio segments
+    podcast_audio = np.concatenate(audio_segments)
+    return (22050, podcast_audio)  # Assuming 22050 Hz sample rate
+
+def gradio_interface(content, duration, voice1, voice2):
+    script = generate_podcast_script(content, duration)
+    return script
+
+def render_podcast(script, voice1, voice2):
+    lines = script.split('\n')
+    audio_segments = []
+    
+    for line in lines:
+        if line.startswith("Host 1:"):
+            audio = text_to_speech(line[7:], speaker_id=0)
+            audio_segments.append(audio)
+        elif line.startswith("Host 2:"):
+            audio = text_to_speech(line[7:], speaker_id=1)
+            audio_segments.append(audio)
+    
+    podcast_audio = np.concatenate(audio_segments)
+    return (22050, podcast_audio)
+
+# Gradio Interface
+with gr.Blocks() as demo:
+    gr.Markdown("# AI Podcast Generator")
+    
+    with gr.Row():
+        content_input = gr.Textbox(label="Paste your content or upload a document")
+        document_upload = gr.File(label="Upload Document")
+    
+    duration = gr.Radio(["1-5 min", "5-10 min", "10-15 min"], label="Estimated podcast duration")
+    
+    with gr.Row():
+        voice1_upload = gr.Audio(label="Upload Voice 1", type="filepath")
+        voice2_upload = gr.Audio(label="Upload Voice 2", type="filepath")
+    
+    generate_btn = gr.Button("Generate Script")
+    script_output = gr.Textbox(label="Generated Script", lines=10)
+    
+    render_btn = gr.Button("Render Podcast")
+    audio_output = gr.Audio(label="Generated Podcast")
+    
+    generate_btn.click(gradio_interface, inputs=[content_input, duration, voice1_upload, voice2_upload], outputs=script_output)
+    render_btn.click(render_podcast, inputs=[script_output, voice1_upload, voice2_upload], outputs=audio_output)
+
+demo.launch()