Update app.py

app.py

@@ -57,6 +57,7 @@ def recognize(audio: tuple[int, np.ndarray], models, language):
             time = timer() - start
             logger.debug("recognized by %s: result '%s', time %.3f s.", name, result, time)
             results.append([name, result, f"{time:.3f} s."])
+
     except Exception as e:
         raise gr.Error(f"{e} Audio: sample_rate: {sample_rate}, waveform.shape: {waveform.shape}.") from e
     else:
@@ -82,16 +83,15 @@ def recognize_with_vad(audio: tuple[int, np.ndarray], name: str):
         if waveform.ndim == 2:
             waveform = waveform.mean(axis=1)
 
-        model = models_vad[name].with_vad(vad)
-        results = []
+        model = models_vad[name].with_vad(vad, batch_size=1)
+        results = ""
         for res in model.recognize(waveform, sample_rate=sample_rate):
             logger.debug("recognized by %s: result '%s'.", name, res)
-            results.append([res.start, res.end, res.text])
+            results += f"[{res.start:5.1f}, {res.end:5.1f}]: {res.text}\n"
+            yield results
 
     except Exception as e:
         raise gr.Error(f"{e} Audio: sample_rate: {sample_rate}, waveform.shape: {waveform.shape}.") from e
-    else:
-        return results
 
 
 with gr.Blocks() as recognize_short:
@@ -110,7 +110,7 @@ with gr.Blocks() as recognize_long:
     with gr.Row():
         gr.ClearButton(audio)
         btn = gr.Button("Recognize", variant="primary")
-    output = gr.Dataframe(headers=["start", "end", "result"], wrap=True)
+    output = gr.TextArea(label="result")  # headers=["start", "end", "result"], wrap=True, every=0.1)
     btn.click(fn=recognize_with_vad, inputs=[audio, name], outputs=output)
 
 with gr.Blocks() as demo: