Create transcrriber.py

transcrriber.py

@@ -0,0 +1,41 @@
+import speech_recognition as sr
+import numpy as np
+import collections
+import config
+
+class SpeechTranscriber:
+    def __init__(self):
+        self.recognizer = sr.Recognizer()
+        self.recognizer.energy_threshold = config.ENERGY_THRESHOLD
+        self.recognizer.dynamic_energy_threshold = config.DYNAMIC_ENERGY_THRESHOLD
+        self.recognizer.pause_threshold = config.PAUSE_THRESHOLD
+        self.audio_buffer = collections.deque(maxlen=config.BUFFER_DURATION * 10)
+        self.last_processed = 0
+        
+    def add_audio_chunk(self, audio_chunk):
+        self.audio_buffer.extend(audio_chunk)
+        
+    def get_transcript_chunk(self):
+        # Only process if we have enough audio
+        if len(self.audio_buffer) < config.SAMPLE_RATE * config.MIN_PROCESSING_DURATION:
+            return None
+            
+        # Convert to AudioData format
+        audio_data = sr.AudioData(
+            np.array(self.audio_buffer).tobytes(), 
+            config.SAMPLE_RATE, 
+            2  # Sample width in bytes
+        )
+        
+        try:
+            # Use Google Web Speech API for best accuracy
+            text = self.recognizer.recognize_google(audio_data)
+            return text
+        except sr.UnknownValueError:
+            return None
+        except sr.RequestError as e:
+            print(f"Speech recognition error: {str(e)}")
+            return None
+        finally:
+            # Clear buffer after processing
+            self.audio_buffer.clear()