Update app.py

app.py

@@ -38,12 +38,6 @@ OUTPUT_DIR = "outputs"
 os.makedirs(OUTPUT_DIR, exist_ok=True)
 os.environ["COQUI_TOS_AGREED"] = "1"
 
-# Initialize TTS model
-device = "cuda" if torch.cuda.is_available() else "cpu"
-
-tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
-logger.info("TTS model initialized on %s", device)
-
 # Define Pydantic model for slide data
 class Slide(BaseModel):
     title: str
@@ -180,7 +174,7 @@ async def validate_and_convert_speaker_audio(speaker_audio):
         return None
 
 # Helper function to generate audio using Coqui TTS API
-def generate_xtts_audio(text, speaker_wav, output_path):
+def generate_xtts_audio(tts, text, speaker_wav, output_path):
     if not tts:
         logger.error("TTS model not initialized")
         return False
@@ -275,22 +269,13 @@ def extract_json_from_message(message):
     logger.warning("Unsupported message type for JSON extraction: %s", type(message))
     return None
 
-# Function to generate Markdown and convert to PDF (landscape, centered)
+# Function to generate Markdown and convert to PDF (portrait, centered)
 def generate_slides_pdf(slides):
     pdf = MarkdownPdf()
-    # Add LaTeX preamble for landscape orientation
-    preamble = r"""
-    \usepackage{pdflscape}
-    \newcommand{\blandscape}{\begin{landscape}}
-    \newcommand{\elandscape}{\end{landscape}}
-    """
-    pdf.set_preamble(preamble)
     
     for slide in slides:
         content_lines = slide['content'].replace('\n', '\n\n')
         markdown_content = f"""
-\\blandscape
-
 <div style="display: flex; flex-direction: column; justify-content: center; align-items: center; height: 100%; text-align: center; padding: 20px;">
 # {slide['title']}
 
@@ -300,8 +285,6 @@ def generate_slides_pdf(slides):
 {content_lines}
 </div>
 
-\\elandscape
-
 ---
 """
         pdf.add_section(Section(markdown_content, toc=False))
@@ -309,7 +292,7 @@ def generate_slides_pdf(slides):
     pdf_file = os.path.join(OUTPUT_DIR, "slides.pdf")
     pdf.save(pdf_file)
     
-    logger.info("Generated PDF slides (landscape): %s", pdf_file)
+    logger.info("Generated PDF slides (portrait): %s", pdf_file)
     return pdf_file
 
 # Async function to update audio preview
@@ -330,11 +313,19 @@ async def on_generate(api_service, api_key, serpapi_key, title, topic, instructi
         """
         return
 
-    if not tts:
+    # Initialize TTS model
+    tts = None
+    try:
+        device = "cuda" if torch.cuda.is_available() else "cpu"
+        tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to(device)
+        logger.info("TTS model initialized on %s", device)
+    except Exception as e:
+        logger.error("Failed to initialize TTS model: %s", str(e))
         yield f"""
         <div style="display: flex; flex-direction: column; justify-content: center; align-items: center; height: 100%; min-height: 700px; padding: 20px; text-align: center; border: 1px solid #ddd; border-radius: 8px;">
-            <h2 style="color: #d9534f;">TTS model not initialized</h2>
-            <p style="margin-top: 20px;">Please ensure the Coqui TTS model is properly installed and try again.</p>
+            <h2 style="color: #d9534f;">TTS model initialization failed</h2>
+            <p style="margin-top: 20px;">Error: {str(e)}</p>
+            <p>Please ensure the Coqui TTS model is properly installed and try again.</p>
         </div>
         """
         return
@@ -619,7 +610,7 @@ Example for 1 content slide:
                         current_text = ". ".join(sentences) + "."
                         logger.info("Retry %d for slide %d with simplified text: %s", attempt, i + 1, current_text)
                     
-                    success = generate_xtts_audio(current_text, validated_speaker_wav, audio_file)
+                    success = generate_xtts_audio(tts, current_text, validated_speaker_wav, audio_file)
                     if not success:
                         raise RuntimeError("TTS generation failed")