song

rag.py

@@ -213,4 +213,4 @@ def retrieve_context(vectorstore, query: str, k: int = 8, mode: str = None) -> s
         )
 
     context = "\n\n---\n\n".join([r.page_content for r in all_results])
-    return context
+    return context

tts.py

@@ -38,6 +38,9 @@ VOICE_MAP = {
 
     # Lecture professor
     "PROFESSOR": ("en-GB-RyanNeural",              "-5%", "-3Hz"),   # British Male, measured/academic
+
+    # Poem — warm, measured narration (same as storytelling but slightly slower)
+    # NARRATOR voice is already defined above and is reused for poem mode
 }
 
 DEFAULT_VOICE = ("en-US-EmmaMultilingualNeural", "-5%", "+0Hz")
@@ -74,7 +77,7 @@ def parse_script(script: str) -> list:
 
     if not cleaned:
         print("[VoiceVerse TTS] Warning: No speaker labels found. Using NARRATOR fallback.")
-        full_text = re.sub(r"^[A-Z]+:\s*", "", script, flags=re.MULTILINE)
+        full_text = re.sub(r"^[A-Z]+:\s*", "", script, flags=re.MULTILINE)  # strip any label
         full_text = " ".join(full_text.strip().split())
         if full_text:
             cleaned = [("NARRATOR", full_text)]
@@ -192,4 +195,4 @@ def script_to_audio(script: str, output_path: str) -> str:
         except Exception:
             pass
 
-    return output_path
+    return output_path