Update app.py

app.py

@@ -6,7 +6,7 @@ text = "[Github]"
 paper_link = "https://arxiv.org/pdf/2307.03183.pdf"
 paper_text = "[Paper]"
 
-model = whisper.load_model("tiny")
+model = whisper.load_model('large-v1')
 print('model loaded')
 
 def round_time_resolution(time_resolution):
@@ -16,9 +16,11 @@ def round_time_resolution(time_resolution):
     return rounded_time_resolution
 
 def predict(audio_path_m, audio_path_t, time_resolution):
-    return asr_output, at_output
-    if (audio_path_m is None) != (audio_path_t is None):
-        return "Please only upload one recording, either upload it or record using microphone.", "Please only upload one recording, either upload it or record using microphone."
+    # print(audio_path_m, audio_path_t)
+    # print(type(audio_path_m), type(audio_path_t))
+    #return audio_path_m, audio_path_t
+    if ((audio_path_m is None) != (audio_path_t is None)) == False:
+        return "Please upload and only upload one recording, either upload the audio file or record using microphone.", "Please upload and only upload one recording, either upload the audio file or record using microphone."
     else:
         audio_path = audio_path_m or audio_path_t
         audio_tagging_time_resolution = round_time_resolution(time_resolution)
@@ -26,19 +28,19 @@ def predict(audio_path_m, audio_path_t, time_resolution):
         audio_tag_result = whisper.parse_at_label(result, language='follow_asr', top_k=5, p_threshold=-1, include_class_list=list(range(527)))
         asr_output = ""
         for segment in result['segments']:
-          asr_output = asr_output + str(segment['start']) + 's-' + str(segment['end']) + 's: ' + segment['text'] + '\n'
+          asr_output = asr_output + format(segment['start'], ".1f") + 's-' + format(segment['end'], ".1f") + 's: ' + segment['text'] + '\n'
         at_output = ""
         for segment in audio_tag_result:
             print(segment)
-            at_output = at_output + str(segment['time']['start']) + 's-' + str(segment['time']['end']) + 's: ' + ','.join([x[0] for x in segment['audio tags']]) + '\n'
+            at_output = at_output + format(segment['time']['start'], ".1f") + 's-' + format(segment['time']['end'], ".1f") + 's: ' + ','.join([x[0] for x in segment['audio tags']]) + '\n'
             print(at_output)
         return asr_output, at_output
 
 iface = gr.Interface(fn=predict,
-                    inputs=[gr.Audio(type="filepath", source='microphone'), gr.Audio(type="filepath"), gr.Textbox(value='10', label='Time Resolution in Seconds (Must be must be an integer multiple of 0.4, e.g., 0.4, 2, 10)')],
+                    inputs=[gr.Audio(type="filepath", source='microphone', label='Please either upload an audio file or record using the microphone.', show_label=True), gr.Audio(type="filepath"), gr.Textbox(value='10', label='Time Resolution in Seconds (Must be must be an integer multiple of 0.4, e.g., 0.4, 2, 10)')],
                     outputs=[gr.Textbox(label="Speech Output"), gr.Textbox(label="Audio Tag Output")],
                     cache_examples=True,
                     title="Quick Demo of Whisper-AT",
                     description="We are glad to introduce Whisper-AT - A new joint audio tagging and speech recognition model. It outputs background sound labels in addition to text." + f"<a href='{paper_link}'>{paper_text}</a> " + f"<a href='{link}'>{text}</a> <br>" +
-                    "Whisper-AT is authored by Yuan Gong, Sameer Khurana, Leonid Karlinsky, and James Glass (MIT & MIT-IBM Watson AI Lab).")
+                    "Whisper-AT is authored by Yuan Gong, Sameer Khurana, Leonid Karlinsky, and James Glass (MIT & MIT-IBM Watson AI Lab). It is an Interspeech 2023 paper.")
 iface.launch(debug=True)