Language param + posture every 2nd frame

live_processor.py

@@ -268,8 +268,8 @@ class LiveSessionProcessor:
             except Exception as e:
                 print(f"[LiveProcessor] Face detect error: {e}")
 
-        # Posture detection every 5th frame
-        if self._video_frame_count % 5 == 0 and self._posture_det is not None:
+        # Posture detection every 2nd frame
+        if self._video_frame_count % 2 == 0 and self._posture_det is not None:
             try:
                 self._posture_frame_count += 1
                 p_result = self._posture_det.detect(image_bytes)
@@ -281,7 +281,7 @@ class LiveSessionProcessor:
 
         self._run_fusion()
 
-    def process_audio_bytes(self, audio_bytes: bytes):
+    def process_audio_bytes(self, audio_bytes: bytes, language: str = None):
         """Process raw audio bytes for voice emotion + STT (no WebRTC)."""
         if not self._session_active:
             return
@@ -309,7 +309,7 @@ class LiveSessionProcessor:
                 tmp.close()
 
                 segments, _info = self._whisper_model.transcribe(
-                    tmp_path,  beam_size=1
+                    tmp_path, beam_size=1, language=language
                 )
                 for seg in segments:
                     text = seg.text.strip()