feat: add model/MODEL_NAME parameters

in order to allow for better efficiency and consistency

main.py

@@ -12,10 +12,10 @@ from selenium.webdriver.common.by import By
 from bs4 import BeautifulSoup
 
 try:
-    from settings import DOCUMENT_ID, WEBHOOK_URL
+    from settings import DOCUMENT_ID, WEBHOOK_URL, MODEL_NAME
 except ImportError:
     print("settings.py nicht gefunden, verwende settings_example.py")
-    from settings_example import DOCUMENT_ID, WEBHOOK_URL
+    from settings_example import DOCUMENT_ID, WEBHOOK_URL, MODEL_NAME
 
 from transcription import transcribe_audio_with_whisper
 
@@ -146,13 +146,17 @@ def test_transcription():
     accuracy = 0
     total = 0
 
+    # Load Whisper model once for efficiency
+    import whisper
+    model = whisper.load_model(MODEL_NAME)
+
     # create or overwrite CSV file for transcription results
     with open("transcription_results.csv", "w") as csvfile:
         csvfile.write("filename,transcription\n")
 
         for mp3_path in download_dir.glob("*.mp3"):
             print(f"Verarbeite Datei: {mp3_path}")
-            transcription = transcribe_audio_with_whisper(mp3_path)
+            transcription = transcribe_audio_with_whisper(mp3_path, model=model)
 
             print(f"Transkription: {transcription}")
 

transcription.py

@@ -1,5 +1,11 @@
 import re
 
+try:
+    from settings import MODEL_NAME
+except ImportError:
+    print("settings.py nicht gefunden, verwende settings_example.py")
+    from settings_example import MODEL_NAME
+
 # Mapping von gesprochenen Buchstaben (deutsch) auf Zeichen
 SPOKEN_TO_CHAR = {
     "a": "a", "ah": "a",
@@ -60,9 +66,11 @@ def _normalize_transcription(raw_text):
             print(f"Unbekanntes Token: '{token}'")
     return ''.join(result)
 
-def transcribe_audio_with_whisper(mp3_path):
-    import whisper
-    model = whisper.load_model("small")
+def transcribe_audio_with_whisper(mp3_path, model=None):
+    if model is None:
+        import whisper
+        model = whisper.load_model(MODEL_NAME)
+
     result = model.transcribe(str(mp3_path), language='de')
     raw_text = result["text"]
     print("Raw transcription:", raw_text)