Update app.py

app.py

@@ -1,214 +1,230 @@
-# app.py
-
 import streamlit as st
+import time
+import re
+import os
+import tempfile
+import pypdf
+from pydub import AudioSegment
 from utils import (
     generate_script, 
-    generate_audio_mp3,  # Updated import
+    generate_audio_mp3, 
     truncate_text, 
     extract_text_from_url, 
     transcribe_youtube_video, 
     research_topic
 )
-from prompts import SYSTEM_PROMPT  # Ensure this module exists
-import pypdf
-from pydub import AudioSegment
-import tempfile
-import os
+from prompts import SYSTEM_PROMPT
+
+def parse_user_edited_transcript(edited_text: str):
+    pattern = r"\*\*(Jane|John)\*\*:\s*(.+)"
+    matches = re.findall(pattern, edited_text)
+    if not matches:
+        # If the user completely changed the format, just treat it as "Jane"
+        return [("Jane", edited_text)]
+    return matches
+
+def regenerate_audio_from_dialogue(dialogue_items):
+    audio_segments = []
+    transcript = ""
+    crossfade_duration = 50  # ms
+
+    for speaker, line_text in dialogue_items:
+        audio_file = generate_audio_mp3(line_text, speaker)
+        seg = AudioSegment.from_file(audio_file, format="mp3")
+        audio_segments.append(seg)
+        transcript += f"**{speaker}**: {line_text}\n\n"
+        os.remove(audio_file)
+
+    if not audio_segments:
+        return None, "No audio segments were generated."
+
+    combined = audio_segments[0]
+    for seg in audio_segments[1:]:
+        combined = combined.append(seg, crossfade=crossfade_duration)
+
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as temp_audio:
+        combined.export(temp_audio.name, format="mp3")
+        final_mp3_path = temp_audio.name
+
+    return final_mp3_path, transcript
 
 def generate_podcast(file, url, video_url, research_topic_input, tone, length):
-    print("[LOG] generate_podcast called")
-    
-    # Check that only one input source is used
     sources = [bool(file), bool(url), bool(video_url), bool(research_topic_input)]
     if sum(sources) > 1:
-        print("[ERROR] Multiple input sources provided.")
-        return None, "Please provide either a PDF file, a URL, a YouTube link, or a Research topic - not multiple."
+        return None, "Provide only one input (PDF, URL, YouTube, or Research topic)."
     if not any(sources):
-        print("[ERROR] No input source provided.")
         return None, "Please provide at least one source."
 
     text = ""
     if file:
         try:
-            print("[LOG] Reading PDF file:", file.name)
             if not file.name.lower().endswith('.pdf'):
-                print("[ERROR] Uploaded file is not a PDF.")
                 return None, "Please upload a PDF file."
             reader = pypdf.PdfReader(file.name)
             text = " ".join(page.extract_text() for page in reader.pages if page.extract_text())
-            print("[LOG] PDF text extraction successful.")
         except Exception as e:
-            print("[ERROR] Error reading PDF file:", e)
-            return None, f"Error reading PDF file: {str(e)}"
+            return None, f"Error reading PDF: {str(e)}"
     elif url:
         try:
-            print("[LOG] Using URL input")
             text = extract_text_from_url(url)
             if not text:
-                print("[ERROR] Failed to extract text from URL.")
-                return None, "Failed to extract text from the provided URL."
+                return None, "Failed to extract text from URL."
         except Exception as e:
-            print("[ERROR] Error extracting text from URL:", e)
             return None, f"Error extracting text from URL: {str(e)}"
     elif video_url:
         try:
-            print("[LOG] Using YouTube video input")
             text = transcribe_youtube_video(video_url)
             if not text:
-                print("[ERROR] Failed to transcribe YouTube video.")
-                return None, "Failed to transcribe the provided YouTube video."
+                return None, "Failed to transcribe YouTube video."
         except Exception as e:
-            print("[ERROR] Error transcribing YouTube video:", e)
             return None, f"Error transcribing YouTube video: {str(e)}"
     elif research_topic_input:
         try:
-            print("[LOG] Researching topic:", research_topic_input)
             text = research_topic(research_topic_input)
             if not text:
-                print("[ERROR] No information found for the topic.")
-                return None, f"Sorry, I couldn't find recent information on '{research_topic_input}'."
+                return None, f"Sorry, no information found on '{research_topic_input}'."
         except Exception as e:
-            print("[ERROR] Error researching topic:", e)
             return None, f"Error researching topic: {str(e)}"
-    else:
-        print("[ERROR] No valid input source detected.")
-        return None, "Please provide a PDF file, URL, YouTube link, or Research topic."
 
     try:
         text = truncate_text(text)
         script = generate_script(SYSTEM_PROMPT, text, tone, length)
     except Exception as e:
-        print("[ERROR] Error generating script:", e)
         return None, f"Error generating script: {str(e)}"
 
     audio_segments = []
     transcript = ""
+    crossfade_duration = 50  # ms
 
     try:
-        print("[LOG] Generating audio segments...")
-        # Define crossfade duration in milliseconds
-        crossfade_duration = 50  # 50ms crossfade for smooth transitions
-
-        for i, item in enumerate(script.dialogue):
-            try:
-                audio_file = generate_audio_mp3(item.text, item.speaker)  # Updated function call
-                line_audio = AudioSegment.from_file(audio_file, format="mp3")  # Changed format to mp3
-                audio_segments.append(line_audio)
-                transcript += f"**{item.speaker}**: {item.text}\n\n"
-                os.remove(audio_file)
-            except Exception as e:
-                print(f"[ERROR] Error generating audio for dialogue item {i+1}: {e}")
-                continue
+        for item in script.dialogue:
+            audio_file = generate_audio_mp3(item.text, item.speaker)
+            seg = AudioSegment.from_file(audio_file, format="mp3")
+            audio_segments.append(seg)
+            transcript += f"**{item.speaker}**: {item.text}\n\n"
+            os.remove(audio_file)
 
         if not audio_segments:
-            print("[ERROR] No audio segments were generated.")
-            return None, "No audio segments were generated."
+            return None, "No audio segments generated."
 
-        print("[LOG] Combining audio segments with crossfades...")
-        # Initialize combined audio with the first segment
         combined = audio_segments[0]
-
-        # Append remaining segments with crossfade
         for seg in audio_segments[1:]:
             combined = combined.append(seg, crossfade=crossfade_duration)
 
-        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as temp_audio:  # Changed suffix to mp3
-            combined.export(temp_audio.name, format="mp3")  # Changed format to mp3
-            print("[LOG] Podcast generated:", temp_audio.name)
-            return temp_audio.name, transcript
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as temp_audio:
+            combined.export(temp_audio.name, format="mp3")
+            final_mp3_path = temp_audio.name
+            return final_mp3_path, transcript
 
     except Exception as e:
-        print("[ERROR] Error generating audio:", e)
         return None, f"Error generating audio: {str(e)}"
 
 def main():
-    # Set Streamlit page config
-    st.set_page_config(
-        page_title="MyPod - AI based Podcast Generator",
-        layout="centered"
-    )
+    st.set_page_config(page_title="MyPod - AI-based Podcast Generator", layout="centered")
 
     st.title("🎙 MyPod - AI-based Podcast Generator")
-    st.markdown(
-        """
-        <style>
-        .main {
-            background-color: #f9f9f9;
-        }
-        .block-container {
-            padding: 2rem;
-            border-radius: 10px;
-            background-color: #ffffff;
-            box-shadow: 0 0 10px rgba(0,0,0,0.1);
-        }
-        </style>
-        """,
-        unsafe_allow_html=True
-    )
-
-    st.markdown(
-        "Welcome to **MyPod**, your go-to AI-powered podcast generator! 🎉\n\n"
-        "MyPod transforms your documents, webpages, YouTube videos, or research topics into a more human-sounding, conversational podcast.\n"
-        "Select a tone and a duration range. The script will be on-topic, concise, and respect your chosen length.\n\n"
-        "### How to use:\n"
-        "1. **Provide one source:** PDF, URL, YouTube link (Requires User Auth - Work in Progress), or a Topic to Research.\n"
-        "2. **Choose the tone and the target duration.**\n"
-        "3. **Click 'Generate Podcast'** to produce your podcast.\n\n"
-        "**Research a Topic:** Please be as detailed as possible in your topic statement. If it's too niche or specific, "
-        "you might not get the desired outcome. We'll fetch information from Wikipedia and RSS feeds (BBC, CNN, Associated Press, "
-        "NDTV, Times of India, The Hindu, Economic Times, Google News) or the LLM knowledge base to get recent info about the topic.\n\n"
-        "**Token Limit:** Up to ~2,048 tokens are supported. Long inputs may be truncated.\n"
-        "**Note:** YouTube transcription uses Whisper on CPU and may take longer for very long videos.\n\n"
-        "⏳**Please be patient while your podcast is being generated.** This process involves content analysis, script creation, "
-        "and high-quality audio synthesis, which may take a few minutes.\n\n"
-        "🔥 **Ready to create your personalized podcast?** Give it a try now and let the magic happen! 🔥"
-    )
-
-    st.write("---")
-
-    # Create 2 columns for inputs
-    col1, col2 = st.columns(2)
 
+    # Standard input UI
+    col1, col2 = st.columns(2)
     with col1:
-        file = st.file_uploader("Upload PDF (Only .pdf)", type=["pdf"])
+        file = st.file_uploader("Upload PDF (.pdf only)", type=["pdf"])
         url = st.text_input("Or Enter URL")
-        video_url = st.text_input("Or Enter YouTube Link (Requires User Auth - Work in Progress)")
-
+        video_url = st.text_input("Or Enter YouTube Link")
     with col2:
         research_topic_input = st.text_input("Or Research a Topic")
-        tone = st.radio(
-            "Tone",
-            ["Humorous", "Formal", "Casual", "Youthful"],
-            index=2
-        )
-        length = st.radio(
-            "Length",
-            ["1-3 Mins", "3-5 Mins", "5-10 Mins", "10-20 Mins"],
-            index=0
-        )
+        tone = st.radio("Tone", ["Humorous", "Formal", "Casual", "Youthful"], index=2)
+        length = st.radio("Length", ["1-3 Mins", "3-5 Mins", "5-10 Mins", "10-20 Mins"], index=0)
 
-    st.write("")
     generate_button = st.button("Generate Podcast")
 
     if generate_button:
-        # Run the generate_podcast function
-        with st.spinner("Generating your podcast, please wait..."):
-            podcast_file, transcript = generate_podcast(
-                file, url, video_url, research_topic_input, tone, length
-            )
-        
+        # Show a custom progress bar & messages
+        progress_bar = st.progress(0)
+        progress_text = st.empty()
+
+        # We don't have real-time progress from TTS, so we'll do a pseudo countdown:
+        # Step 1: Start
+        progress_text.write("Alright, let's get started...")
+        progress_bar.progress(10)
+        time.sleep(1.5)
+
+        # Step 2: Some cheeky text
+        progress_text.write("Working on the magic. Hang tight!")
+        progress_bar.progress(40)
+        time.sleep(1.5)
+
+        # Step 3: Almost there
+        progress_text.write("Almost done. Adding a dash of awesomeness...")
+        progress_bar.progress(70)
+        time.sleep(1.5)
+
+        # Step 4: Actually generate
+        podcast_file, transcript = generate_podcast(
+            file, url, video_url, research_topic_input, tone, length
+        )
+        time.sleep(1.5)
+
         if podcast_file is None:
+            # Reset progress to 0 if error
+            progress_bar.progress(100)
             st.error(transcript)
+            return
         else:
+            progress_bar.progress(100)
+            progress_text.write("Done!")
+
             st.success("Podcast generated successfully!")
             audio_file = open(podcast_file, 'rb')
             audio_bytes = audio_file.read()
             audio_file.close()
-
             st.audio(audio_bytes, format='audio/mp3')
-            st.markdown(transcript)
 
-            # Clean up the temp file
+            # Show transcript in an editable box
+            st.markdown("### Generated Transcript (Editable)")
+            edited_text = st.text_area(
+                "Feel free to tweak lines, fix errors, or reword anything.",
+                value=transcript,
+                height=300
+            )
+
+            if st.button("Regenerate Audio From Edited Text"):
+                # Another pseudo progress bar for regeneration
+                regen_bar = st.progress(0)
+                regen_text = st.empty()
+
+                regen_text.write("Let's do this revision!")
+                regen_bar.progress(25)
+                time.sleep(1.0)
+
+                regen_text.write("Cooking up fresh audio...")
+                regen_bar.progress(60)
+                time.sleep(1.0)
+
+                # Parse and regenerate
+                dialogue_items = parse_user_edited_transcript(edited_text)
+                new_audio_path, new_transcript = regenerate_audio_from_dialogue(dialogue_items)
+
+                regen_bar.progress(90)
+                time.sleep(1.0)
+
+                if new_audio_path is None:
+                    regen_bar.progress(100)
+                    st.error(new_transcript)
+                else:
+                    regen_bar.progress(100)
+                    regen_text.write("All set!")
+                    st.success("Regenerated audio below:")
+
+                    new_audio_file = open(new_audio_path, 'rb')
+                    new_audio_bytes = new_audio_file.read()
+                    new_audio_file.close()
+                    st.audio(new_audio_bytes, format='audio/mp3')
+                    st.markdown(new_transcript)
+
+                    # Clean up
+                    os.remove(new_audio_path)
+
+            # Clean up the original file
             os.remove(podcast_file)
 
 if __name__ == "__main__":