liuyang commited on
Commit
e79159f
Β·
1 Parent(s): e3d9c9e

Add audio diarization task to Gradio interface: Introduced a new button and function for audio diarization, allowing users to process audio with speaker separation. Updated existing button labels for clarity.

Browse files
Files changed (1) hide show
  1. app.py +25 -3
app.py CHANGED
@@ -1316,8 +1316,20 @@ def format_segments_for_display(result):
1316
 
1317
  return output
1318
 
 
 
 
 
 
 
 
 
 
 
 
 
1319
  @spaces.GPU
1320
- def process_audio_gradio(task_json, num_speakers, language, translate, prompt, group_segments, use_diarization, batch_size, model_name):
1321
  """Gradio interface function"""
1322
 
1323
  result = transcriber.process_audio_transcribe(
@@ -1421,7 +1433,8 @@ with demo:
1421
  value=True
1422
  )
1423
 
1424
- process_btn = gr.Button("πŸš€ Transcribe Audio", variant="primary")
 
1425
 
1426
  with gr.Column():
1427
  output_text = gr.Markdown(
@@ -1443,7 +1456,7 @@ with demo:
1443
 
1444
  # Event handlers
1445
  process_btn.click(
1446
- fn=process_audio_gradio,
1447
  inputs=[
1448
  task_json_input,
1449
  num_speakers,
@@ -1457,6 +1470,15 @@ with demo:
1457
  ],
1458
  outputs=[output_text, output_json]
1459
  )
 
 
 
 
 
 
 
 
 
1460
 
1461
  # Examples
1462
  gr.Markdown("### πŸ“‹ Usage Tips:")
 
1316
 
1317
  return output
1318
 
1319
+
1320
+ @spaces.GPU
1321
+ def audio_diarization_task(task_json, num_speakers):
1322
+ """Gradio interface function"""
1323
+
1324
+ result = transcriber.process_audio_diarization(
1325
+ task_json=task_json,
1326
+ num_speakers=num_speakers if num_speakers > 0 else None,
1327
+ )
1328
+ #formatted_output = format_segments_for_display(result)
1329
+ return "OK", result
1330
+
1331
  @spaces.GPU
1332
+ def audio_transcribe_task(task_json, num_speakers, language, translate, prompt, group_segments, use_diarization, batch_size, model_name):
1333
  """Gradio interface function"""
1334
 
1335
  result = transcriber.process_audio_transcribe(
 
1433
  value=True
1434
  )
1435
 
1436
+ process_btn = gr.Button("πŸš€ Audio Transcribe Task", variant="primary")
1437
+ process_btn1 = gr.Button("πŸš€ Audio Diarization Task", variant="primary")
1438
 
1439
  with gr.Column():
1440
  output_text = gr.Markdown(
 
1456
 
1457
  # Event handlers
1458
  process_btn.click(
1459
+ fn=audio_transcribe_task,
1460
  inputs=[
1461
  task_json_input,
1462
  num_speakers,
 
1470
  ],
1471
  outputs=[output_text, output_json]
1472
  )
1473
+
1474
+ process_btn1.click(
1475
+ fn=audio_diarization_task,
1476
+ inputs=[
1477
+ task_json_input,
1478
+ num_speakers
1479
+ ],
1480
+ outputs=[output_text, output_json]
1481
+ )
1482
 
1483
  # Examples
1484
  gr.Markdown("### πŸ“‹ Usage Tips:")