fix

Stephen A-transcript.json

@@ -1 +0,0 @@
-[{"speaker": "Stepen A", "start": 1.0101867572, "end": 17.1731748727, "duration": 16.1629881154, "transcript": "You can call it day two but it's really day one because the full two-hour show with me and the man Shannon Sharpless. I know it was Hiccup last year, I know it was Hiccup yesterday Shannon, I know it was Hiccup yesterday. Alright, I mean, cause you called me something other than Stephen A, but I got a solution to the problem. ", "speakers_labeled": "Stepen A"}, {"speaker": "Shannon Sharpe", "start": 12.5891341256, "end": 13.6078098472, "duration": 1.0186757216, "transcript": "I don't think I'm going to stay forever.", "speakers_labeled": "Shannon Sharpe"}, {"speaker": "Stepen A", "start": 18.3276740238, "end": 19.4142614601, "duration": 1.0865874363, "transcript": "Ah, there we go!", "speakers_labeled": "Stepen A"}, {"speaker": "Stephen A and Shannon", "start": 19.4142614601, "end": 19.9235993209, "duration": 0.5093378608, "transcript": "Yo!", "speakers_labeled": "Stephen A and Shannon"}, {"speaker": "Stepen A", "start": 19.9235993209, "end": 20.0424448217, "duration": 0.1188455008, "transcript": "You", "speakers_labeled": "Stepen A"}, {"speaker": "Stephen A and Shannon", "start": 20.4499151104, "end": 25.6791171477, "duration": 5.2292020374, "transcript": "What's up, y'all? My name is Steven A. What's up, baby? And you are Molly Carol.", "speakers_labeled": "Stephen A and Shannon"}, {"speaker": "Molly", "start": 25.6791171477, "end": 28.5993208829, "duration": 2.9202037351, "transcript": "Yes, you said it right, day two.", "speakers_labeled": "Molly"}, {"speaker": "Stephen A and Shannon", "start": 26.3582342954, "end": 26.5619694397, "duration": 0.2037351443, "transcript": "Thank you.", "speakers_labeled": "Stephen A and Shannon"}, {"speaker": "Stephen A and Shannon", "start": 28.3616298812, "end": 30.1103565365, "duration": 1.7487266553, "transcript": "It only took me seven years. You know what I...", "speakers_labeled": "Stephen A and Shannon"}, {"speaker": "Molly", "start": 29.3803056027, "end": 34.0322580645, "duration": 4.6519524618, "transcript": "Stephen A is still learning. Karam, good job. It rhymes with harem. Karam. ", "speakers_labeled": "Molly"}, {"speaker": "Shannon Sharpe", "start": 31.8760611205, "end": 32.8268251273, "duration": 0.9507640068, "transcript": "Good job. It rhymes with.", "speakers_labeled": "Shannon Sharpe"}, {"speaker": "Shannon Sharpe", "start": 33.0984719864, "end": 40.3650254669, "duration": 7.2665534805, "transcript": "She said it rhymes with harem. We just called you cute. She had to tell the world it rhymes with harem? I'm just wondering.", "speakers_labeled": "Shannon Sharpe"}, {"speaker": "Molly", "start": 37.6825127334, "end": 38.1578947368, "duration": 0.4753820034, "transcript": "because she had.", "speakers_labeled": "Molly"}, {"speaker": "Molly", "start": 40.3650254669, "end": 41.3497453311, "duration": 0.9847198642, "transcript": "That's what it rhymes with.", "speakers_labeled": "Molly"}, {"speaker": "Stepen A", "start": 42.7419354839, "end": 49.2784380306, "duration": 6.5365025467, "transcript": "Thanks for watching ESPN on YouTube. For live streaming sports and premium content, subscribe to ESPN Plus.", "speakers_labeled": "Stepen A"}]

app.py

@@ -312,9 +312,8 @@ if "audio" in locals():
             diarization = load_rttm_file(st.session_state.rttm )
         else:
             # with ProgressHook() as hook:
-            audio_ = create_audio_stream(audio)
             # diarization = pipeline(audio_, hook=hook)
-            diarization = pipeline(audio_)
+            diarization = pipeline(create_audio_stream(audio))
             # dump the diarization output to disk using RTTM format
             with open(f'{audio_name.split(".")[0]}.rttm', "w") as f:
                 diarization.write_rttm(f)

stephen-a-smith-has-jokes-with-shannon-sharpe-first-take-transcript.json

@@ -1 +0,0 @@
-[{"speaker": "SPEAKER_00", "start": 0.9422750424448219, "end": 17.10526315789474, "duration": 16.162988115449917, "transcript": "You can call it day two but it's really day one because the full two-hour show with me and the man Shannon Sharpless. I know it was hiccup last year. I know it was hiccup yesterday Shannon. All right I mean because you called me something other than Stephen A. But I got a solution to the problem."}, {"speaker": "SPEAKER_01", "start": 12.521222410865875, "end": 13.539898132427846, "duration": 1.018675721561971, "transcript": "I don't think it makes sense."}, {"speaker": "SPEAKER_01", "start": 18.2427843803056, "end": 18.853989813242784, "duration": 0.611205432937183, "transcript": "Ah!"}, {"speaker": "SPEAKER_01", "start": 18.870967741935488, "end": 18.98981324278438, "duration": 0.11884550084889156, "transcript": "Oh"}, {"speaker": "SPEAKER_00", "start": 19.193548387096776, "end": 19.906621392190154, "duration": 0.7130730050933778, "transcript": "Here we go!"}, {"speaker": "SPEAKER_01", "start": 20.31409168081494, "end": 25.611205432937183, "duration": 5.297113752122243, "transcript": "What's up, y'all? My name is Stephen A. What's up, baby? And you are Molly Carol."}, {"speaker": "SPEAKER_02", "start": 25.611205432937183, "end": 28.565365025466896, "duration": 2.954159592529713, "transcript": "Yes, you said it right. Day 2."}, {"speaker": "SPEAKER_01", "start": 26.273344651952463, "end": 26.544991511035654, "duration": 0.27164685908319086, "transcript": "Thank you."}, {"speaker": "SPEAKER_01", "start": 28.293718166383705, "end": 29.006791171477083, "duration": 0.7130730050933778, "transcript": "That one took me seven years."}, {"speaker": "SPEAKER_00", "start": 29.006791171477083, "end": 30.246179966044146, "duration": 1.2393887945670627, "transcript": "You know what I mean?"}, {"speaker": "SPEAKER_02", "start": 29.278438030560274, "end": 34.01528013582343, "duration": 4.7368421052631575, "transcript": "You know what I mean. Stephen is still learning. Karam. Good job. It rhymes with harem. Karam. Yes. "}, {"speaker": "SPEAKER_00", "start": 31.09507640067912, "end": 31.536502546689306, "duration": 0.4414261460101869, "transcript": "Tara."}, {"speaker": "SPEAKER_00", "start": 31.723259762309, "end": 40.602716468590835, "duration": 8.879456706281836, "transcript": "Good job. It rhymes with harem. Yes. Harem. Camera. Camera. She said it rhymes with harem. Cube. Go ahead. We just called you Cube. OK. Come on. Did she have to tell the world it rhymed with harem? I'm just wondering. That's what it rhymes with."}, {"speaker": "SPEAKER_02", "start": 35.39049235993209, "end": 35.54329371816638, "duration": 0.1528013582342922, "transcript": "Thank you."}, {"speaker": "SPEAKER_01", "start": 36.307300509337864, "end": 36.64685908319185, "duration": 0.339558573853985, "transcript": "\u0443\u0439\u0434\u0435\u0442"}, {"speaker": "SPEAKER_02", "start": 37.597623089983024, "end": 38.08998302207131, "duration": 0.4923599320882843, "transcript": "Did she have to do it?"}, {"speaker": "SPEAKER_02", "start": 40.12733446519525, "end": 40.50084889643463, "duration": 0.37351443123938566, "transcript": "Whatever."}, {"speaker": "SPEAKER_02", "start": 40.602716468590835, "end": 41.28183361629881, "duration": 0.6791171477079772, "transcript": "You're lying to us."}, {"speaker": "SPEAKER_00", "start": 42.69100169779287, "end": 49.24448217317487, "duration": 6.553480475382003, "transcript": "Thanks for watching ESPN on YouTube. For live streaming sports and premium content, subscribe to ESPN+."}]

stephen-a-smith-has-jokes-with-shannon-sharpe-first-take.rttm

@@ -1,23 +0,0 @@
-SPEAKER stream 1 0.942 16.163 <NA> <NA> SPEAKER_00 <NA> <NA>
-SPEAKER stream 1 12.521 1.019 <NA> <NA> SPEAKER_01 <NA> <NA>
-SPEAKER stream 1 18.243 0.611 <NA> <NA> SPEAKER_01 <NA> <NA>
-SPEAKER stream 1 18.854 0.017 <NA> <NA> SPEAKER_00 <NA> <NA>
-SPEAKER stream 1 18.871 0.119 <NA> <NA> SPEAKER_01 <NA> <NA>
-SPEAKER stream 1 18.990 0.017 <NA> <NA> SPEAKER_00 <NA> <NA>
-SPEAKER stream 1 19.007 0.187 <NA> <NA> SPEAKER_01 <NA> <NA>
-SPEAKER stream 1 19.194 0.713 <NA> <NA> SPEAKER_00 <NA> <NA>
-SPEAKER stream 1 20.314 5.297 <NA> <NA> SPEAKER_01 <NA> <NA>
-SPEAKER stream 1 25.611 2.954 <NA> <NA> SPEAKER_02 <NA> <NA>
-SPEAKER stream 1 26.273 0.272 <NA> <NA> SPEAKER_01 <NA> <NA>
-SPEAKER stream 1 28.294 0.713 <NA> <NA> SPEAKER_01 <NA> <NA>
-SPEAKER stream 1 29.007 1.239 <NA> <NA> SPEAKER_00 <NA> <NA>
-SPEAKER stream 1 29.278 4.737 <NA> <NA> SPEAKER_02 <NA> <NA>
-SPEAKER stream 1 31.095 0.441 <NA> <NA> SPEAKER_00 <NA> <NA>
-SPEAKER stream 1 31.723 8.879 <NA> <NA> SPEAKER_00 <NA> <NA>
-SPEAKER stream 1 35.390 0.153 <NA> <NA> SPEAKER_02 <NA> <NA>
-SPEAKER stream 1 36.307 0.340 <NA> <NA> SPEAKER_01 <NA> <NA>
-SPEAKER stream 1 37.598 0.492 <NA> <NA> SPEAKER_02 <NA> <NA>
-SPEAKER stream 1 39.924 0.051 <NA> <NA> SPEAKER_02 <NA> <NA>
-SPEAKER stream 1 40.127 0.374 <NA> <NA> SPEAKER_02 <NA> <NA>
-SPEAKER stream 1 40.603 0.679 <NA> <NA> SPEAKER_02 <NA> <NA>
-SPEAKER stream 1 42.691 6.553 <NA> <NA> SPEAKER_00 <NA> <NA>