fix: whisper model params

app.py

@@ -84,7 +84,17 @@ def transcribe(image, audio):
 
     y = y.astype(np.float32)
     y /= np.max(np.abs(y))
-    return generate_answer(image, transcriber({"sampling_rate": sr, "raw": y})["text"])
+
+    input_features = transcriber.feature_extractor(
+        y, sampling_rate=sr, return_tensors="pt"
+    ).input_features
+
+    transcription = transcriber.model.generate(input_features)
+    transcription_text = transcriber.tokenizer.decode(
+        transcription[0], skip_special_tokens=True
+    )
+
+    return generate_answer(image, transcription_text)
 
 
 qa_interface = gr.Interface(