Create utils/transcription.py

utils/transcription.py

@@ -0,0 +1,46 @@
+import whisper
+import torch
+from pathlib import Path
+
+class TranscriptionService:
+    def __init__(self, model_size="base"):
+        """Initialize Whisper model"""
+        self.device = "cuda" if torch.cuda.is_available() else "cpu"
+        print(f"Loading Whisper model on {self.device}...")
+        self.model = whisper.load_model(model_size, device=self.device)
+    
+    def transcribe(self, audio_path):
+        """Transcribe audio with word-level timestamps"""
+        try:
+            result = self.model.transcribe(
+                audio_path,
+                word_timestamps=True,
+                language="en"
+            )
+            
+            # Format transcript with timestamps
+            segments = []
+            for segment in result['segments']:
+                segments.append({
+                    'start': segment['start'],
+                    'end': segment['end'],
+                    'text': segment['text'].strip(),
+                    'words': segment.get('words', [])
+                })
+            
+            return {
+                'text': result['text'],
+                'segments': segments,
+                'language': result['language']
+            }
+            
+        except Exception as e:
+            raise Exception(f"Transcription failed: {str(e)}")
+    
+    def get_text_at_time(self, transcript, start_time, end_time):
+        """Get transcript text for a specific time range"""
+        text_parts = []
+        for segment in transcript['segments']:
+            if segment['start'] >= start_time and segment['end'] <= end_time:
+                text_parts.append(segment['text'])
+        return ' '.join(text_parts)