Switch whisper base->small for much better multilingual STT

live_processor.py

@@ -193,13 +193,13 @@ class LiveSessionProcessor:
 
         self._fusion = FuzzyFusionEngine()
 
-        # Load faster-whisper for STT (base multilingual — much better for non-English)
+        # Load faster-whisper for STT (small multilingual — best accuracy for non-English on CPU)
         try:
             from faster_whisper import WhisperModel
             self._whisper_model = WhisperModel(
-                "base", device="cpu", compute_type="int8"
+                "small", device="cpu", compute_type="int8"
             )
-            print("[LiveProcessor] faster-whisper base (multilingual) loaded")
+            print("[LiveProcessor] faster-whisper small (multilingual) loaded")
         except ImportError:
             print("[LiveProcessor] faster-whisper not available, STT disabled")
             self._whisper_model = None