Spaces:
Runtime error
Runtime error
save
Browse files
app.py
CHANGED
@@ -5,8 +5,9 @@ import whisper
|
|
5 |
from flask import Flask, jsonify, request
|
6 |
import requests
|
7 |
import time
|
|
|
8 |
|
9 |
-
model = whisper.load_model("
|
10 |
|
11 |
app = Flask(__name__)
|
12 |
app.config['TIMEOUT'] = 60 * 10 # 10 mins
|
@@ -23,6 +24,13 @@ def runApi():
|
|
23 |
|
24 |
response = requests.get(audio_url)
|
25 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
if response.status_code == requests.codes.ok:
|
27 |
with open("audio.mp3", "wb") as f:
|
28 |
f.write(response.content)
|
@@ -35,15 +43,16 @@ def runApi():
|
|
35 |
audio = "audio.mp3"
|
36 |
|
37 |
audio = whisper.load_audio(audio)
|
38 |
-
audio = whisper.pad_or_trim(audio)
|
39 |
|
40 |
-
mel = whisper.log_mel_spectrogram(audio).to(model.device)
|
41 |
|
42 |
-
_, probs = model.detect_language(mel)
|
43 |
|
44 |
-
options = whisper.DecodingOptions(fp16 = False)
|
45 |
-
result = whisper.decode(model, mel, options)
|
46 |
|
|
|
47 |
|
48 |
end_time = time.time()
|
49 |
total_time = end_time - start_time
|
|
|
5 |
from flask import Flask, jsonify, request
|
6 |
import requests
|
7 |
import time
|
8 |
+
from transformers import pipeline
|
9 |
|
10 |
+
model = whisper.load_model("small")
|
11 |
|
12 |
app = Flask(__name__)
|
13 |
app.config['TIMEOUT'] = 60 * 10 # 10 mins
|
|
|
24 |
|
25 |
response = requests.get(audio_url)
|
26 |
|
27 |
+
pipe = pipeline(
|
28 |
+
"automatic-speech-recognition",
|
29 |
+
model="openai/whisper-large",
|
30 |
+
chunk_length_s=30,
|
31 |
+
device=model.device,
|
32 |
+
)
|
33 |
+
|
34 |
if response.status_code == requests.codes.ok:
|
35 |
with open("audio.mp3", "wb") as f:
|
36 |
f.write(response.content)
|
|
|
43 |
audio = "audio.mp3"
|
44 |
|
45 |
audio = whisper.load_audio(audio)
|
46 |
+
# audio = whisper.pad_or_trim(audio)
|
47 |
|
48 |
+
# mel = whisper.log_mel_spectrogram(audio).to(model.device)
|
49 |
|
50 |
+
# _, probs = model.detect_language(mel)
|
51 |
|
52 |
+
# options = whisper.DecodingOptions(fp16 = False)
|
53 |
+
# result = whisper.decode(model, mel, options)
|
54 |
|
55 |
+
result = pipe(audio.copy())["text"]
|
56 |
|
57 |
end_time = time.time()
|
58 |
total_time = end_time - start_time
|