asr

app.py

@@ -33,7 +33,11 @@ mms_transcribe = gr.Interface(
 mms_synthesize = gr.Interface(
     fn=tts.synthesize,
     inputs=[
-        gr.Text(label="Input text"),
+        gr.Text(
+            label="Input text",
+            max_length=200,  # Limit input to 200 characters
+            interactive=True,
+        ),
         gr.Dropdown(
             choices=[model for model in tts.models_info],
             label="Select a Model",

asr.py

@@ -67,13 +67,18 @@ def transcribe(audio_data, model_id) -> str:
         # file upload
         audio_input, sampling_rate = torchaudio.load(audio_data)
     else: 
-        return "<<ERROR: Invalid Audio Input Instance: {}>>".format(type(audio_data))
+        return "<<ERROR: Invalid Audio Input Instance: {}>>".format(type(audio_data)), None
 
+    print(audio_input.shape)
+    # Check audio duration
+    duration = audio_input.shape[1] / sampling_rate
+    if duration > 10:
+        return "<<ERROR: Audio duration exceeds 10 seconds. Please upload a shorter audio clip for faster processing.>>", None
+    
     model = models_info[model_id]["model"]
     processor = models_info[model_id]["processor"]
     target_sr = processor.feature_extractor.sampling_rate
     ctc_model = models_info[model_id]["ctc_model"]
-    print(target_sr)
 
     # Resample if needed
     if sampling_rate != target_sr: