Temp

src/stt/listener.py

@@ -41,6 +41,9 @@ class AudioListener:
             target_samples = int(self.sample_rate * self.chunk_duration)
             chunk = chunk[:target_samples]
 
+            # Flatten to 1D array (samples,) as expected by faster-whisper
+            chunk = chunk.flatten()
+
             # Use call_soon_threadsafe to put the chunk into the asyncio queue from the callback thread
             self.loop.call_soon_threadsafe(self.audio_queue.put_nowait, chunk)
             self._buffer = []

src/stt/transcriber.py

@@ -46,8 +46,8 @@ class Transcriber:
             return ""
 
         try:
-            # faster-whisper expects audio in float32
-            audio_data = audio_chunk.astype("float32")
+            # faster-whisper expects audio in float32 and 1D array
+            audio_data = audio_chunk.astype("float32").flatten()
 
             # Transcribe the audio
             segments, info = self.model.transcribe(audio_data, beam_size=5)