Commit
·
ab7253a
1
Parent(s):
d9317e3
Update app.py
Browse files
app.py
CHANGED
@@ -8,7 +8,7 @@ transcriber_hindi = pipeline("automatic-speech-recognition", model="ai4bharat/in
|
|
8 |
transcriber_bengali = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_bengali")
|
9 |
transcriber_odia = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec-odia")
|
10 |
transcriber_gujarati = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_gujarati")
|
11 |
-
transcriber_telugu = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_telugu")
|
12 |
transcriber_sinhala = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_sinhala")
|
13 |
transcriber_tamil = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_tamil")
|
14 |
transcriber_nepali = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_nepali")
|
@@ -19,16 +19,19 @@ def resample_to_16k(audio, orig_sr):
|
|
19 |
y_resampled = librosa.resample(y=audio, orig_sr=orig_sr, target_sr=16000)
|
20 |
return y_resampled
|
21 |
|
22 |
-
def transcribe(audio):
|
23 |
sr,y = audio
|
24 |
y = y.astype(np.float32)
|
25 |
y/= np.max(np.abs(y))
|
26 |
y_resampled = resample_to_16k(y,sr)
|
27 |
-
|
|
|
28 |
pipe= eval(f'transcriber_{lang}')
|
|
|
29 |
trans = pipe(y_resampled)
|
|
|
30 |
|
31 |
-
return trans["text"]
|
32 |
|
33 |
demo = gr.Interface(
|
34 |
transcribe,
|
|
|
8 |
transcriber_bengali = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_bengali")
|
9 |
transcriber_odia = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec-odia")
|
10 |
transcriber_gujarati = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_gujarati")
|
11 |
+
# transcriber_telugu = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_telugu")
|
12 |
transcriber_sinhala = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_sinhala")
|
13 |
transcriber_tamil = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_tamil")
|
14 |
transcriber_nepali = pipeline("automatic-speech-recognition", model="ai4bharat/indicwav2vec_v1_nepali")
|
|
|
19 |
y_resampled = librosa.resample(y=audio, orig_sr=orig_sr, target_sr=16000)
|
20 |
return y_resampled
|
21 |
|
22 |
+
def transcribe(audio,lang="hindi"):
|
23 |
sr,y = audio
|
24 |
y = y.astype(np.float32)
|
25 |
y/= np.max(np.abs(y))
|
26 |
y_resampled = resample_to_16k(y,sr)
|
27 |
+
if lang=="telugu":
|
28 |
+
return "No Model","So no time!"
|
29 |
pipe= eval(f'transcriber_{lang}')
|
30 |
+
start_time = time.time()
|
31 |
trans = pipe(y_resampled)
|
32 |
+
end_time = time.time()
|
33 |
|
34 |
+
return trans["text"],(end_time-start_time)
|
35 |
|
36 |
demo = gr.Interface(
|
37 |
transcribe,
|