Spaces:
Running
Running
Changed to a single input (audio file)
Browse files
app.py
CHANGED
@@ -9,11 +9,9 @@ whisper = pipeline(model='jlvdoorn/whisper-large-v2-atco2-asr-atcosim', use_auth
|
|
9 |
# bert_atco_ner = pipeline(model='Jzuluaga/bert-base-ner-atc-en-atco2-1h')
|
10 |
|
11 |
#%%
|
12 |
-
def transcribe(
|
13 |
if audio_file is not None:
|
14 |
return whisper(audio_file)['text']
|
15 |
-
if audio_mic is not None:
|
16 |
-
return whisper(audio_mic)['text']
|
17 |
else:
|
18 |
return 'There was no audio to transcribe...'
|
19 |
|
@@ -48,7 +46,7 @@ def transcribe(audio_mic, audio_file):
|
|
48 |
#%%
|
49 |
iface = gr.Interface(
|
50 |
fn=transcribe,
|
51 |
-
inputs=
|
52 |
outputs=gr.Text(label='Transcription'),
|
53 |
title='Whisper Large v2 - ATCO2-ASR-ATCOSIM',
|
54 |
description='This demo will transcribe ATC audio files by using the Whisper Large v2 model fine-tuned on the ATCO2 and ATCOSIM datasets. Further it uses a Named Entity Recognition model to extract callsigns, commands and values from the transcription. This model is based on Google\'s BERT model and fine-tuned on the ATCO2 dataset.',
|
|
|
9 |
# bert_atco_ner = pipeline(model='Jzuluaga/bert-base-ner-atc-en-atco2-1h')
|
10 |
|
11 |
#%%
|
12 |
+
def transcribe(audio_file):
|
13 |
if audio_file is not None:
|
14 |
return whisper(audio_file)['text']
|
|
|
|
|
15 |
else:
|
16 |
return 'There was no audio to transcribe...'
|
17 |
|
|
|
46 |
#%%
|
47 |
iface = gr.Interface(
|
48 |
fn=transcribe,
|
49 |
+
inputs=gr.Audio(source='upload', type='filepath'),
|
50 |
outputs=gr.Text(label='Transcription'),
|
51 |
title='Whisper Large v2 - ATCO2-ASR-ATCOSIM',
|
52 |
description='This demo will transcribe ATC audio files by using the Whisper Large v2 model fine-tuned on the ATCO2 and ATCOSIM datasets. Further it uses a Named Entity Recognition model to extract callsigns, commands and values from the transcription. This model is based on Google\'s BERT model and fine-tuned on the ATCO2 dataset.',
|