diff --git a/step4_translate.py b/step4_translate.py
index 09f04a2..7b1e228 100644
--- a/step4_translate.py
+++ b/step4_translate.py
@@ -238,10 +238,19 @@ def translate_file(input_path: Path, client: OpenAI, model: str) -> Path:
     
     # Check cache for each line
     cached_count = 0
+    skipped_count = 0
     to_translate = []
     
     for line_data in parsed_lines:
         english = line_data["english"]
+        speaker = line_data["speaker"]
+        
+        # Skip translation for opening song (speaker is "Song")
+        if speaker == "Song":
+            line_data["chinese"] = english  # Keep English for song lyrics
+            skipped_count += 1
+            continue
+        
         cached = get_cached_translation(english)
         if cached:
             line_data["chinese"] = cached
@@ -250,6 +259,7 @@ def translate_file(input_path: Path, client: OpenAI, model: str) -> Path:
             to_translate.append(line_data)
     
     print(f"  Cached translations: {cached_count}")
+    print(f"  Skipped (Song): {skipped_count}")
     print(f"  To translate: {len(to_translate)}")
     
     # Translate in batches
@@ -274,10 +284,10 @@ def translate_file(input_path: Path, client: OpenAI, model: str) -> Path:
             if translation and translation != "[Translation failed]":
                 save_translation_cache(line_data["english"], translation)
     
-    # Combine results (cached + translated)
+    # Combine results (cached + translated + skipped)
     result = parsed_lines
     
-    print(f"  Translation complete: {cached_count} from cache, {translated_count} new")
+    print(f"  Translation complete: {cached_count} from cache, {translated_count} new, {skipped_count} skipped (Song)")
     
     # Save JSON output
     output_filename = input_path.stem.replace("_speakers", "") + "_translated.json"