Spaces:
Running
Running
Update main.py
Browse files
main.py
CHANGED
@@ -1,32 +1,192 @@
|
|
1 |
-
from
|
2 |
-
|
3 |
-
import whisperx
|
4 |
-
import torch
|
5 |
import tempfile
|
6 |
-
import shutil
|
7 |
import os
|
|
|
|
|
|
|
8 |
|
9 |
-
|
|
|
|
|
10 |
|
11 |
-
|
12 |
-
device = "cuda" if torch.cuda.is_available() else "cpu"
|
13 |
-
model = whisperx.load_model("medium", device)
|
14 |
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
|
23 |
-
|
24 |
-
|
25 |
-
|
|
|
|
|
|
|
26 |
|
27 |
-
|
28 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
|
30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
except Exception as e:
|
32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from flask import Flask, request, jsonify
|
2 |
+
import whisper
|
|
|
|
|
3 |
import tempfile
|
|
|
4 |
import os
|
5 |
+
from werkzeug.utils import secure_filename
|
6 |
+
import logging
|
7 |
+
from datetime import datetime
|
8 |
|
9 |
+
# Configure logging
|
10 |
+
logging.basicConfig(level=logging.INFO)
|
11 |
+
logger = logging.getLogger(__name__)
|
12 |
|
13 |
+
app = Flask(__name__)
|
|
|
|
|
14 |
|
15 |
+
# Configuration
|
16 |
+
app.config['MAX_CONTENT_LENGTH'] = 100 * 1024 * 1024 # 100MB max file size
|
17 |
+
ALLOWED_EXTENSIONS = {'wav', 'mp3', 'mp4', 'mpeg', 'mpga', 'm4a', 'webm', 'flac'}
|
18 |
+
|
19 |
+
# Load Whisper model (you can change the model size: tiny, base, small, medium, large)
|
20 |
+
MODEL_SIZE = "base" # Change this to your preferred model size
|
21 |
+
logger.info(f"Loading Whisper model: {MODEL_SIZE}")
|
22 |
+
model = whisper.load_model(MODEL_SIZE)
|
23 |
+
logger.info("Whisper model loaded successfully")
|
24 |
+
|
25 |
+
def allowed_file(filename):
|
26 |
+
"""Check if the file extension is allowed"""
|
27 |
+
return '.' in filename and \
|
28 |
+
filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS
|
29 |
|
30 |
+
def format_timestamp(seconds):
|
31 |
+
"""Convert seconds to HH:MM:SS.mmm format"""
|
32 |
+
hours = int(seconds // 3600)
|
33 |
+
minutes = int((seconds % 3600) // 60)
|
34 |
+
secs = seconds % 60
|
35 |
+
return f"{hours:02d}:{minutes:02d}:{secs:06.3f}"
|
36 |
|
37 |
+
@app.route('/', methods=['GET'])
|
38 |
+
def health_check():
|
39 |
+
"""Health check endpoint"""
|
40 |
+
return jsonify({
|
41 |
+
"status": "healthy",
|
42 |
+
"message": "Whisper Transcription API is running",
|
43 |
+
"model": MODEL_SIZE,
|
44 |
+
"timestamp": datetime.now().isoformat()
|
45 |
+
})
|
46 |
|
47 |
+
@app.route('/transcribe', methods=['POST'])
|
48 |
+
def transcribe_audio():
|
49 |
+
"""
|
50 |
+
Transcribe audio file and return word-level timestamps
|
51 |
+
|
52 |
+
Expected form data:
|
53 |
+
- audio_file: The audio file to transcribe
|
54 |
+
- language (optional): Language code (e.g., 'en', 'es', 'fr')
|
55 |
+
- task (optional): 'transcribe' or 'translate' (default: transcribe)
|
56 |
+
"""
|
57 |
+
try:
|
58 |
+
# Check if audio file is present
|
59 |
+
if 'audio_file' not in request.files:
|
60 |
+
return jsonify({'error': 'No audio file provided'}), 400
|
61 |
+
|
62 |
+
file = request.files['audio_file']
|
63 |
+
|
64 |
+
if file.filename == '':
|
65 |
+
return jsonify({'error': 'No file selected'}), 400
|
66 |
+
|
67 |
+
if not allowed_file(file.filename):
|
68 |
+
return jsonify({
|
69 |
+
'error': f'File type not allowed. Supported formats: {", ".join(ALLOWED_EXTENSIONS)}'
|
70 |
+
}), 400
|
71 |
+
|
72 |
+
# Get optional parameters
|
73 |
+
language = request.form.get('language', None)
|
74 |
+
task = request.form.get('task', 'transcribe')
|
75 |
+
|
76 |
+
if task not in ['transcribe', 'translate']:
|
77 |
+
return jsonify({'error': 'Task must be either "transcribe" or "translate"'}), 400
|
78 |
+
|
79 |
+
# Save uploaded file temporarily
|
80 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=f".{file.filename.rsplit('.', 1)[1].lower()}") as tmp_file:
|
81 |
+
file.save(tmp_file.name)
|
82 |
+
temp_path = tmp_file.name
|
83 |
+
|
84 |
+
logger.info(f"Processing file: {file.filename}")
|
85 |
+
|
86 |
+
try:
|
87 |
+
# Transcribe with word-level timestamps
|
88 |
+
result = model.transcribe(
|
89 |
+
temp_path,
|
90 |
+
language=language,
|
91 |
+
task=task,
|
92 |
+
word_timestamps=True,
|
93 |
+
verbose=False
|
94 |
+
)
|
95 |
+
|
96 |
+
# Extract word-level data
|
97 |
+
word_segments = []
|
98 |
+
for segment in result.get("segments", []):
|
99 |
+
if "words" in segment:
|
100 |
+
for word_data in segment["words"]:
|
101 |
+
word_segments.append({
|
102 |
+
"word": word_data.get("word", "").strip(),
|
103 |
+
"start": word_data.get("start", 0),
|
104 |
+
"end": word_data.get("end", 0),
|
105 |
+
"start_formatted": format_timestamp(word_data.get("start", 0)),
|
106 |
+
"end_formatted": format_timestamp(word_data.get("end", 0)),
|
107 |
+
"confidence": word_data.get("probability", 0)
|
108 |
+
})
|
109 |
+
|
110 |
+
# Prepare response
|
111 |
+
response_data = {
|
112 |
+
"success": True,
|
113 |
+
"filename": secure_filename(file.filename),
|
114 |
+
"language": result.get("language", "unknown"),
|
115 |
+
"task": task,
|
116 |
+
"duration": result.get("segments", [{}])[-1].get("end", 0) if result.get("segments") else 0,
|
117 |
+
"text": result.get("text", ""),
|
118 |
+
"word_count": len(word_segments),
|
119 |
+
"segments": result.get("segments", []),
|
120 |
+
"words": word_segments,
|
121 |
+
"model_used": MODEL_SIZE,
|
122 |
+
"processing_time": None # You can add timing if needed
|
123 |
+
}
|
124 |
+
|
125 |
+
logger.info(f"Successfully transcribed {len(word_segments)} words from {file.filename}")
|
126 |
+
return jsonify(response_data)
|
127 |
+
|
128 |
+
except Exception as e:
|
129 |
+
logger.error(f"Transcription error: {str(e)}")
|
130 |
+
return jsonify({'error': f'Transcription failed: {str(e)}'}), 500
|
131 |
+
|
132 |
+
finally:
|
133 |
+
# Clean up temporary file
|
134 |
+
if os.path.exists(temp_path):
|
135 |
+
os.unlink(temp_path)
|
136 |
+
|
137 |
except Exception as e:
|
138 |
+
logger.error(f"API error: {str(e)}")
|
139 |
+
return jsonify({'error': f'Server error: {str(e)}'}), 500
|
140 |
+
|
141 |
+
@app.route('/models', methods=['GET'])
|
142 |
+
def available_models():
|
143 |
+
"""Get information about available Whisper models"""
|
144 |
+
models_info = {
|
145 |
+
"current_model": MODEL_SIZE,
|
146 |
+
"available_models": {
|
147 |
+
"tiny": {"size": "~39 MB", "speed": "~32x", "accuracy": "lowest"},
|
148 |
+
"base": {"size": "~74 MB", "speed": "~16x", "accuracy": "low"},
|
149 |
+
"small": {"size": "~244 MB", "speed": "~6x", "accuracy": "medium"},
|
150 |
+
"medium": {"size": "~769 MB", "speed": "~2x", "accuracy": "high"},
|
151 |
+
"large": {"size": "~1550 MB", "speed": "~1x", "accuracy": "highest"}
|
152 |
+
},
|
153 |
+
"supported_languages": [
|
154 |
+
"en", "zh", "de", "es", "ru", "ko", "fr", "ja", "pt", "tr", "pl", "ca", "nl",
|
155 |
+
"ar", "sv", "it", "id", "hi", "fi", "vi", "he", "uk", "el", "ms", "cs", "ro",
|
156 |
+
"da", "hu", "ta", "no", "th", "ur", "hr", "bg", "lt", "la", "mi", "ml", "cy",
|
157 |
+
"sk", "te", "fa", "lv", "bn", "sr", "az", "sl", "kn", "et", "mk", "br", "eu",
|
158 |
+
"is", "hy", "ne", "mn", "bs", "kk", "sq", "sw", "gl", "mr", "pa", "si", "km",
|
159 |
+
"sn", "yo", "so", "af", "oc", "ka", "be", "tg", "sd", "gu", "am", "yi", "lo",
|
160 |
+
"uz", "fo", "ht", "ps", "tk", "nn", "mt", "sa", "lb", "my", "bo", "tl", "mg",
|
161 |
+
"as", "tt", "haw", "ln", "ha", "ba", "jw", "su"
|
162 |
+
]
|
163 |
+
}
|
164 |
+
return jsonify(models_info)
|
165 |
+
|
166 |
+
@app.errorhandler(413)
|
167 |
+
def too_large(e):
|
168 |
+
return jsonify({'error': 'File too large. Maximum size is 100MB'}), 413
|
169 |
+
|
170 |
+
@app.errorhandler(404)
|
171 |
+
def not_found(e):
|
172 |
+
return jsonify({'error': 'Endpoint not found'}), 404
|
173 |
+
|
174 |
+
@app.errorhandler(500)
|
175 |
+
def internal_error(e):
|
176 |
+
return jsonify({'error': 'Internal server error'}), 500
|
177 |
+
|
178 |
+
if __name__ == '__main__':
|
179 |
+
print(f"""
|
180 |
+
Whisper Transcription API Server
|
181 |
+
================================
|
182 |
+
Model: {MODEL_SIZE}
|
183 |
+
Endpoints:
|
184 |
+
- GET / : Health check
|
185 |
+
- POST /transcribe : Transcribe audio file
|
186 |
+
- GET /models : Available models info
|
187 |
+
|
188 |
+
Supported formats: {', '.join(ALLOWED_EXTENSIONS)}
|
189 |
+
Max file size: 100MB
|
190 |
+
""")
|
191 |
+
|
192 |
+
app.run(debug=True, host='0.0.0.0', port=5000)
|