Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# importer gradio
|
2 |
+
import gradio as gr
|
3 |
+
from transformers import pipeline
|
4 |
+
# Importer nemo.collections.asr
|
5 |
+
import nemo.collections.asr as nemo_asr
|
6 |
+
# Instancier le modèle
|
7 |
+
asr_canary = nemo_asr.models.ASRModel.from_pretrained("nvidia/canary-1b-flash")
|
8 |
+
# Instanstier le modèle
|
9 |
+
asr_whisper = pipeline("automatic-speech-recognition", model="openai/whisper-small")
|
10 |
+
|
11 |
+
# Fonction de transcription whisper
|
12 |
+
def transcrire1(fpath):
|
13 |
+
output = asr_whisper(fpath)
|
14 |
+
return output["text"]
|
15 |
+
|
16 |
+
# Fonction de transcription canary-1b-flash
|
17 |
+
def transcrire2(fpath, source_lang, target_lang):
|
18 |
+
transcriptions = asr_canary.transcribe([fpath],
|
19 |
+
source_lang = source_lang, target_lang = target_lang)
|
20 |
+
text = transcriptions[0].text
|
21 |
+
|
22 |
+
return text
|
23 |
+
|
24 |
+
# Créer les blocs
|
25 |
+
demo = gr.Blocks(theme='JohnSmith9982/small_and_pretty')
|
26 |
+
# Créer un interface ASR whisper avec un microphone
|
27 |
+
mic_transcrire = gr.Interface(
|
28 |
+
fn=transcrire1,
|
29 |
+
inputs=gr.Audio(sources="microphone",
|
30 |
+
type="filepath"),
|
31 |
+
cache_examples=True,
|
32 |
+
outputs=gr.Textbox(label="Transcription",
|
33 |
+
lines=3),
|
34 |
+
title = 'Transcrire par microphone - Whisper')
|
35 |
+
|
36 |
+
# Créer un interface ASR whisper par audio
|
37 |
+
fich_transcrire = gr.Interface(
|
38 |
+
fn=transcrire1,
|
39 |
+
inputs=gr.Audio(sources="upload",
|
40 |
+
type="filepath"),
|
41 |
+
outputs=gr.Textbox(label="Transcription",
|
42 |
+
lines=3),
|
43 |
+
title = 'Transcrire un fichier audio - Whisper'
|
44 |
+
)
|
45 |
+
|
46 |
+
|
47 |
+
# Créer un interface ASR canary avec un microphone
|
48 |
+
mic_transcrire1 = gr.Interface(
|
49 |
+
fn=transcrire2,
|
50 |
+
inputs=[gr.Audio(sources="microphone",type="filepath"),
|
51 |
+
gr.Dropdown(choices = ['fr', 'en'], label ='Source languge'),
|
52 |
+
gr.Dropdown(choices = ['fr', 'en'], label = 'Target language')],
|
53 |
+
cache_examples=True,
|
54 |
+
outputs=gr.Textbox(label="Transcription",
|
55 |
+
lines=3),
|
56 |
+
title = 'Transcrire par microphone - Canary')
|
57 |
+
|
58 |
+
# Créer un interface ASR canary par audio
|
59 |
+
fich_transcrire1 = gr.Interface(
|
60 |
+
fn=transcrire2,
|
61 |
+
inputs=[gr.Audio(sources="upload",type="filepath"),
|
62 |
+
gr.Dropdown(choices = ['fr', 'en'], label ='Source languge'),
|
63 |
+
gr.Dropdown(choices = ['fr', 'en'], label ='Target language')],
|
64 |
+
outputs=gr.Textbox(label="Transcription",
|
65 |
+
lines=3),
|
66 |
+
title= 'Transcrire un fichier audio - Canary'
|
67 |
+
)
|
68 |
+
|
69 |
+
# Faire un tabbed des interfaces sur demo
|
70 |
+
with demo:
|
71 |
+
gr.TabbedInterface(
|
72 |
+
[mic_transcrire,
|
73 |
+
fich_transcrire,
|
74 |
+
mic_transcrire1,
|
75 |
+
fich_transcrire1],
|
76 |
+
["Transcrire Microphone",
|
77 |
+
"Transcrire Audio",
|
78 |
+
"Transcrire Microphone",
|
79 |
+
"Transcrire Audio"],
|
80 |
+
)
|
81 |
+
|
82 |
+
demo.launch()
|
83 |
+
|