Update main.py
Browse files
main.py
CHANGED
@@ -13,9 +13,9 @@ def hello():
|
|
13 |
return "Semabox, listens to you!"
|
14 |
|
15 |
# Load the Whisper model
|
16 |
-
print("Loading Whisper model
|
17 |
model = whisper.load_model("tiny")
|
18 |
-
print("Whisper model loaded
|
19 |
|
20 |
# Get time of request
|
21 |
def get_time():
|
@@ -31,36 +31,36 @@ def get_time():
|
|
31 |
|
32 |
|
33 |
def transcribe(audio_path):
|
34 |
-
print(f"Transcribing audio from: {audio_path}", flush=True)
|
35 |
|
36 |
# Load audio and pad/trim it to fit 30 seconds
|
37 |
-
print("Loading and processing audio
|
38 |
audio = whisper.load_audio(audio_path)
|
39 |
audio = whisper.pad_or_trim(audio)
|
40 |
|
41 |
# Make log-Mel spectrogram and move to the same device as the model
|
42 |
-
print("Creating log-Mel spectrogram
|
43 |
mel = whisper.log_mel_spectrogram(audio).to(model.device)
|
44 |
|
45 |
# Detect the spoken language
|
46 |
print("Detecting language...", flush=True)
|
47 |
_, probs = model.detect_language(mel)
|
48 |
language = max(probs, key=probs.get)
|
49 |
-
print(f"Detected language: {language}", flush=True)
|
50 |
|
51 |
# Decode the audio
|
52 |
print("Decoding audio...", flush=True)
|
53 |
options = whisper.DecodingOptions(fp16=False)
|
54 |
result = whisper.decode(model, mel, options)
|
55 |
|
56 |
-
print("Transcription complete
|
57 |
return result.text, language
|
58 |
|
59 |
@app.route('/transcribe', methods=['POST'])
|
60 |
def transcribe_audio():
|
61 |
# Record the time when the request was received
|
62 |
request_received_time, _ = get_time()
|
63 |
-
print(f"Received request at /transcribe at {request_received_time}", flush=True)
|
64 |
|
65 |
if 'audio' not in request.files:
|
66 |
print("Error: No audio file provided", flush=True)
|
@@ -74,7 +74,7 @@ def transcribe_audio():
|
|
74 |
audio_path = os.path.join("temp_audio", audio_file.filename)
|
75 |
os.makedirs("temp_audio", exist_ok=True)
|
76 |
audio_file.save(audio_path)
|
77 |
-
print(f"Audio file saved to: {audio_path} (Size: {audio_file_size} bytes)", flush=True)
|
78 |
|
79 |
# Record the time before starting transcription
|
80 |
transcription_start_time = time.time()
|
@@ -92,13 +92,13 @@ def transcribe_audio():
|
|
92 |
|
93 |
# Clean up the saved file
|
94 |
os.remove(audio_path)
|
95 |
-
print(f"Audio file removed from: {audio_path}", flush=True)
|
96 |
|
97 |
# Record the time when the response is being sent
|
98 |
response_sent_time, _ = get_time()
|
99 |
|
100 |
# Return the transcription, detected language, and timing information
|
101 |
-
print(f"Transcription: {transcription}, Language: {language}", flush=True)
|
102 |
return jsonify({
|
103 |
"transcription": transcription,
|
104 |
"language": language,
|
@@ -110,9 +110,9 @@ def transcribe_audio():
|
|
110 |
|
111 |
@app.route('/healthcheck', methods=['GET'])
|
112 |
def healthcheck():
|
113 |
-
print("Received request at /healthcheck", flush=True)
|
114 |
return jsonify({"status": "API is running"}), 200
|
115 |
|
116 |
if __name__ == '__main__':
|
117 |
-
print("Starting Flask app
|
118 |
app.run(host="0.0.0.0", port=5000)
|
|
|
13 |
return "Semabox, listens to you!"
|
14 |
|
15 |
# Load the Whisper model
|
16 |
+
print("Loading Whisper model...\n", flush=True)
|
17 |
model = whisper.load_model("tiny")
|
18 |
+
print("Whisper model loaded.\n", flush=True)
|
19 |
|
20 |
# Get time of request
|
21 |
def get_time():
|
|
|
31 |
|
32 |
|
33 |
def transcribe(audio_path):
|
34 |
+
print(f"Transcribing audio from: {audio_path}\n", flush=True)
|
35 |
|
36 |
# Load audio and pad/trim it to fit 30 seconds
|
37 |
+
print("Loading and processing audio...\n", flush=True)
|
38 |
audio = whisper.load_audio(audio_path)
|
39 |
audio = whisper.pad_or_trim(audio)
|
40 |
|
41 |
# Make log-Mel spectrogram and move to the same device as the model
|
42 |
+
print("Creating log-Mel spectrogram...\n", flush=True)
|
43 |
mel = whisper.log_mel_spectrogram(audio).to(model.device)
|
44 |
|
45 |
# Detect the spoken language
|
46 |
print("Detecting language...", flush=True)
|
47 |
_, probs = model.detect_language(mel)
|
48 |
language = max(probs, key=probs.get)
|
49 |
+
print(f"Detected language: {language}\n", flush=True)
|
50 |
|
51 |
# Decode the audio
|
52 |
print("Decoding audio...", flush=True)
|
53 |
options = whisper.DecodingOptions(fp16=False)
|
54 |
result = whisper.decode(model, mel, options)
|
55 |
|
56 |
+
print("Transcription complete.\n", flush=True)
|
57 |
return result.text, language
|
58 |
|
59 |
@app.route('/transcribe', methods=['POST'])
|
60 |
def transcribe_audio():
|
61 |
# Record the time when the request was received
|
62 |
request_received_time, _ = get_time()
|
63 |
+
print(f"Received request at /transcribe at {request_received_time}\n", flush=True)
|
64 |
|
65 |
if 'audio' not in request.files:
|
66 |
print("Error: No audio file provided", flush=True)
|
|
|
74 |
audio_path = os.path.join("temp_audio", audio_file.filename)
|
75 |
os.makedirs("temp_audio", exist_ok=True)
|
76 |
audio_file.save(audio_path)
|
77 |
+
print(f"Audio file saved to: {audio_path} (Size: {audio_file_size} bytes)\n", flush=True)
|
78 |
|
79 |
# Record the time before starting transcription
|
80 |
transcription_start_time = time.time()
|
|
|
92 |
|
93 |
# Clean up the saved file
|
94 |
os.remove(audio_path)
|
95 |
+
print(f"Audio file removed from: {audio_path}\n", flush=True)
|
96 |
|
97 |
# Record the time when the response is being sent
|
98 |
response_sent_time, _ = get_time()
|
99 |
|
100 |
# Return the transcription, detected language, and timing information
|
101 |
+
print(f"Transcription: {transcription}, Language: {language}\n", flush=True)
|
102 |
return jsonify({
|
103 |
"transcription": transcription,
|
104 |
"language": language,
|
|
|
110 |
|
111 |
@app.route('/healthcheck', methods=['GET'])
|
112 |
def healthcheck():
|
113 |
+
print("Received request at /healthcheck\n", flush=True)
|
114 |
return jsonify({"status": "API is running"}), 200
|
115 |
|
116 |
if __name__ == '__main__':
|
117 |
+
print("Starting Flask app...\n", flush=True)
|
118 |
app.run(host="0.0.0.0", port=5000)
|