Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from flask import Flask, request, jsonify
|
2 |
+
import whisper
|
3 |
+
import os
|
4 |
+
|
5 |
+
# Initialize Flask app
|
6 |
+
app = Flask(__name__)
|
7 |
+
|
8 |
+
print("\nHello welcome to SemaBox\n", flush=True)
|
9 |
+
|
10 |
+
# Load Whisper model
|
11 |
+
model = whisper.load_model("small")
|
12 |
+
|
13 |
+
def transcribe(audio_path):
|
14 |
+
# Load audio and pad/trim it to fit 30 seconds
|
15 |
+
audio = whisper.load_audio(audio_path)
|
16 |
+
audio = whisper.pad_or_trim(audio)
|
17 |
+
|
18 |
+
# Make log-Mel spectrogram and move to the same device as the model
|
19 |
+
mel = whisper.log_mel_spectrogram(audio).to(model.device)
|
20 |
+
|
21 |
+
# Detect the spoken language
|
22 |
+
_, probs = model.detect_language(mel)
|
23 |
+
detected_language = max(probs, key=probs.get)
|
24 |
+
print(f"Detected language: {detected_language}")
|
25 |
+
|
26 |
+
# Decode the audio
|
27 |
+
options = whisper.DecodingOptions(fp16=False)
|
28 |
+
result = whisper.decode(model, mel, options)
|
29 |
+
return result.text, detected_language
|
30 |
+
|
31 |
+
# Define the route for transcription
|
32 |
+
@app.route('/transcribe', methods=['POST'])
|
33 |
+
def transcribe_audio():
|
34 |
+
# Check if an audio file is included in the request
|
35 |
+
if 'audio' not in request.files:
|
36 |
+
return jsonify({"error": "No audio file provided"}), 400
|
37 |
+
|
38 |
+
audio_file = request.files['audio']
|
39 |
+
|
40 |
+
# Save the uploaded audio file
|
41 |
+
audio_path = os.path.join("temp_audio", audio_file.filename)
|
42 |
+
audio_file.save(audio_path)
|
43 |
+
|
44 |
+
# Transcribe the audio
|
45 |
+
transcription, language = transcribe(audio_path)
|
46 |
+
|
47 |
+
# Clean up the saved file
|
48 |
+
os.remove(audio_path)
|
49 |
+
|
50 |
+
# Return the transcription and detected language
|
51 |
+
return jsonify({"transcription": transcription, "language": language}), 200
|
52 |
+
|
53 |
+
# Healthcheck endpoint
|
54 |
+
@app.route('/healthcheck', methods=['GET'])
|
55 |
+
def healthcheck():
|
56 |
+
return jsonify({"status": "API is running"}), 200
|
57 |
+
|
58 |
+
# Run the Flask app
|
59 |
+
if __name__ == '__main__':
|
60 |
+
# Ensure the temp_audio directory exists
|
61 |
+
if not os.path.exists("temp_audio"):
|
62 |
+
os.makedirs("temp_audio")
|
63 |
+
|
64 |
+
app.run(host="0.0.0.0", port=5000)
|