Update app.py

app.py

@@ -18,25 +18,29 @@ def process_audio_file(file):
     return inputs
 
 
-def transcribe(file):
+def transcribe(file, state=""):
     inputs = process_audio_file(file)
     with torch.no_grad():
         output_logit = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
     pred_ids = torch.argmax(output_logit, dim=-1)
-    return processor.batch_decode(pred_ids)[0]
+    text = processor.batch_decode(pred_ids)[0]
+    state += text + " "
+    return state, state
 
 
 description = "A simple interface to transcribe from spoken Japanese to Hiragana."
 article = "Author: <a href=\"https://huggingface.co/vumichien\">Vu Minh Chien</a>."
-inputs = [gr.inputs.Audio(source="microphone", type='filepath', optional=True)
-          ]
+inputs = [gr.inputs.Audio(source="microphone", type="filepath", optional=True), 
+            "state"]
+outputs = ["textbox", "state"]
+        
 examples = [["samples/BASIC5000_0001.wav"],
             ["samples/BASIC5000_0005.wav"]
         ]
 iface = gr.Interface(
     fn=transcribe,
     inputs=inputs,
-    outputs="text",
+    outputs=outputs,
     layout="horizontal",
     theme="huggingface",
     title="Transcribe Japanese audio to Hiragana",
@@ -44,5 +48,6 @@ iface = gr.Interface(
     article=article,
     allow_flagging='never',
     examples=examples,
+    live=True,
 )
 iface.launch(enable_queue=True, share=True)