Update asr.py

asr.py

@@ -66,10 +66,19 @@ model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
 # )
 
 
-def transcribe(
-    audio_data, lang="eng (English)"
-):    
-    sr, audio_samples = audio_data
+def transcribe(audio_data, lang="eng (English)"):    
+
+    if isinstance(audio_data, tuple):
+        # microphone
+        sr, audio_samples = audio_data
+        print("case1", audio_samples[:5])
+        assert sr == ASR_SAMPLING_RATE, "Invalid sampling rate"
+    else:
+        # file upload
+        isinstance(audio_data, str)
+        print("case2 1", audio_data)
+        audio_samples = librosa.load(audio_fp, sr=ASR_SAMPLING_RATE, mono=True)[0]
+        print("case2", audio_samples[:5])
 
     lang_code = lang.split()[0]
     processor.tokenizer.set_target_lang(lang_code)