Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -12,16 +12,33 @@ nltk.download('punkt')
|
|
12 |
class TTSDatasetCollector:
|
13 |
"""Manages TTS dataset collection and organization"""
|
14 |
|
15 |
-
def __init__(self
|
16 |
-
|
|
|
17 |
self.sentences = []
|
18 |
self.current_index = 0
|
19 |
self.setup_directories()
|
20 |
|
21 |
def setup_directories(self):
|
22 |
"""Create necessary directory structure"""
|
|
|
|
|
|
|
|
|
23 |
for subdir in ['audio', 'transcriptions', 'metadata']:
|
24 |
-
(self.root_path / subdir).mkdir(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
def load_text_file(self, file):
|
27 |
"""Process and load text file"""
|
@@ -30,8 +47,12 @@ class TTSDatasetCollector:
|
|
30 |
text = f.read()
|
31 |
self.sentences = nltk.sent_tokenize(text)
|
32 |
self.current_index = 0
|
|
|
|
|
|
|
33 |
return True, f"Loaded {len(self.sentences)} sentences"
|
34 |
except Exception as e:
|
|
|
35 |
return False, f"Error loading file: {str(e)}"
|
36 |
|
37 |
def generate_filenames(self, dataset_name: str, speaker_id: str) -> tuple:
|
@@ -57,11 +78,13 @@ class TTSDatasetCollector:
|
|
57 |
text_dir.mkdir(exist_ok=True)
|
58 |
|
59 |
# Save audio file
|
60 |
-
|
|
|
61 |
|
62 |
# Save transcription
|
|
|
63 |
self.save_transcription(
|
64 |
-
|
65 |
self.sentences[self.current_index],
|
66 |
{
|
67 |
'speaker_id': speaker_id,
|
@@ -71,9 +94,20 @@ class TTSDatasetCollector:
|
|
71 |
}
|
72 |
)
|
73 |
|
74 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
except Exception as e:
|
76 |
-
|
|
|
|
|
77 |
|
78 |
def save_transcription(self, file_path: Path, text: str, metadata: dict):
|
79 |
"""Save transcription with metadata"""
|
@@ -88,6 +122,50 @@ Timestamp: {metadata['timestamp']}
|
|
88 |
"""
|
89 |
with open(file_path, 'w', encoding='utf-8') as f:
|
90 |
f.write(content)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
91 |
|
92 |
def create_interface():
|
93 |
"""Create Gradio interface for TTS data collection"""
|
@@ -144,6 +222,24 @@ def create_interface():
|
|
144 |
label="Status",
|
145 |
interactive=False
|
146 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
147 |
|
148 |
# Event handlers
|
149 |
def load_file(file):
|
@@ -152,7 +248,8 @@ def create_interface():
|
|
152 |
current_text: "",
|
153 |
next_text: "",
|
154 |
progress: "",
|
155 |
-
status: "No file selected"
|
|
|
156 |
}
|
157 |
|
158 |
success, msg = collector.load_text_file(file)
|
@@ -161,14 +258,16 @@ def create_interface():
|
|
161 |
current_text: "",
|
162 |
next_text: "",
|
163 |
progress: "",
|
164 |
-
status: msg
|
|
|
165 |
}
|
166 |
|
167 |
return {
|
168 |
current_text: collector.sentences[0],
|
169 |
next_text: collector.sentences[1] if len(collector.sentences) > 1 else "",
|
170 |
progress: f"Sentence 1 of {len(collector.sentences)}",
|
171 |
-
status: msg
|
|
|
172 |
}
|
173 |
|
174 |
def update_display():
|
@@ -178,7 +277,8 @@ def create_interface():
|
|
178 |
current_text: "",
|
179 |
next_text: "",
|
180 |
progress: "",
|
181 |
-
status: "No text loaded"
|
|
|
182 |
}
|
183 |
|
184 |
next_idx = collector.current_index + 1
|
@@ -186,7 +286,8 @@ def create_interface():
|
|
186 |
current_text: collector.sentences[collector.current_index],
|
187 |
next_text: collector.sentences[next_idx] if next_idx < len(collector.sentences) else "",
|
188 |
progress: f"Sentence {collector.current_index + 1} of {len(collector.sentences)}",
|
189 |
-
status: "Ready for recording"
|
|
|
190 |
}
|
191 |
|
192 |
def next_sentence():
|
@@ -204,44 +305,43 @@ def create_interface():
|
|
204 |
def save_recording(audio, spk_id, ds_name):
|
205 |
"""Handle saving recording"""
|
206 |
if not audio:
|
207 |
-
return {status: "No audio recorded"}
|
208 |
if not spk_id:
|
209 |
-
return {status: "Speaker ID required"}
|
210 |
if not ds_name:
|
211 |
-
return {status: "Dataset name required"}
|
212 |
|
213 |
success, msg = collector.save_recording(audio, spk_id, ds_name)
|
214 |
-
return {
|
|
|
|
|
|
|
215 |
|
216 |
# Connect event handlers
|
217 |
file_input.change(
|
218 |
load_file,
|
219 |
inputs=[file_input],
|
220 |
-
outputs=[current_text, next_text, progress, status]
|
221 |
)
|
222 |
|
223 |
next_btn.click(
|
224 |
next_sentence,
|
225 |
-
outputs=[current_text, next_text, progress, status]
|
226 |
)
|
227 |
|
228 |
prev_btn.click(
|
229 |
prev_sentence,
|
230 |
-
outputs=[current_text, next_text, progress, status]
|
231 |
)
|
232 |
|
233 |
save_btn.click(
|
234 |
save_recording,
|
235 |
inputs=[audio_recorder, speaker_id, dataset_name],
|
236 |
-
outputs=[status]
|
237 |
)
|
238 |
|
239 |
return interface
|
240 |
|
241 |
if __name__ == "__main__":
|
242 |
interface = create_interface()
|
243 |
-
interface.launch(
|
244 |
-
server_name="0.0.0.0",
|
245 |
-
server_port=7860,
|
246 |
-
share=True
|
247 |
-
)
|
|
|
12 |
class TTSDatasetCollector:
|
13 |
"""Manages TTS dataset collection and organization"""
|
14 |
|
15 |
+
def __init__(self):
|
16 |
+
# Get the directory where app.py is located
|
17 |
+
self.root_path = Path(os.path.dirname(os.path.abspath(__file__))) / "dataset"
|
18 |
self.sentences = []
|
19 |
self.current_index = 0
|
20 |
self.setup_directories()
|
21 |
|
22 |
def setup_directories(self):
|
23 |
"""Create necessary directory structure"""
|
24 |
+
# Create main dataset directory
|
25 |
+
self.root_path.mkdir(exist_ok=True)
|
26 |
+
|
27 |
+
# Create subdirectories
|
28 |
for subdir in ['audio', 'transcriptions', 'metadata']:
|
29 |
+
(self.root_path / subdir).mkdir(exist_ok=True)
|
30 |
+
|
31 |
+
# Create a log file to track operations
|
32 |
+
log_file = self.root_path / 'dataset_log.txt'
|
33 |
+
if not log_file.exists():
|
34 |
+
with open(log_file, 'w', encoding='utf-8') as f:
|
35 |
+
f.write(f"Dataset collection started on {datetime.now().isoformat()}\n")
|
36 |
+
|
37 |
+
def log_operation(self, message: str):
|
38 |
+
"""Log operations to keep track of dataset collection"""
|
39 |
+
log_file = self.root_path / 'dataset_log.txt'
|
40 |
+
with open(log_file, 'a', encoding='utf-8') as f:
|
41 |
+
f.write(f"[{datetime.now().isoformat()}] {message}\n")
|
42 |
|
43 |
def load_text_file(self, file):
|
44 |
"""Process and load text file"""
|
|
|
47 |
text = f.read()
|
48 |
self.sentences = nltk.sent_tokenize(text)
|
49 |
self.current_index = 0
|
50 |
+
|
51 |
+
# Log the file loading
|
52 |
+
self.log_operation(f"Loaded text file with {len(self.sentences)} sentences")
|
53 |
return True, f"Loaded {len(self.sentences)} sentences"
|
54 |
except Exception as e:
|
55 |
+
self.log_operation(f"Error loading file: {str(e)}")
|
56 |
return False, f"Error loading file: {str(e)}"
|
57 |
|
58 |
def generate_filenames(self, dataset_name: str, speaker_id: str) -> tuple:
|
|
|
78 |
text_dir.mkdir(exist_ok=True)
|
79 |
|
80 |
# Save audio file
|
81 |
+
audio_path = audio_dir / audio_name
|
82 |
+
shutil.copy2(audio_file, audio_path)
|
83 |
|
84 |
# Save transcription
|
85 |
+
text_path = text_dir / text_name
|
86 |
self.save_transcription(
|
87 |
+
text_path,
|
88 |
self.sentences[self.current_index],
|
89 |
{
|
90 |
'speaker_id': speaker_id,
|
|
|
94 |
}
|
95 |
)
|
96 |
|
97 |
+
# Update metadata
|
98 |
+
self.update_metadata(speaker_id, dataset_name)
|
99 |
+
|
100 |
+
# Log the save operation
|
101 |
+
self.log_operation(
|
102 |
+
f"Saved recording: Speaker={speaker_id}, Dataset={dataset_name}, "
|
103 |
+
f"Audio={audio_name}, Text={text_name}"
|
104 |
+
)
|
105 |
+
|
106 |
+
return True, f"Recording saved successfully as {audio_name}"
|
107 |
except Exception as e:
|
108 |
+
error_msg = f"Error saving recording: {str(e)}"
|
109 |
+
self.log_operation(error_msg)
|
110 |
+
return False, error_msg
|
111 |
|
112 |
def save_transcription(self, file_path: Path, text: str, metadata: dict):
|
113 |
"""Save transcription with metadata"""
|
|
|
122 |
"""
|
123 |
with open(file_path, 'w', encoding='utf-8') as f:
|
124 |
f.write(content)
|
125 |
+
|
126 |
+
def update_metadata(self, speaker_id: str, dataset_name: str):
|
127 |
+
"""Update dataset metadata file"""
|
128 |
+
metadata_file = self.root_path / 'metadata' / 'dataset_info.json'
|
129 |
+
|
130 |
+
try:
|
131 |
+
if metadata_file.exists():
|
132 |
+
with open(metadata_file, 'r') as f:
|
133 |
+
metadata = json.load(f)
|
134 |
+
else:
|
135 |
+
metadata = {'speakers': {}, 'last_updated': None}
|
136 |
+
|
137 |
+
# Update speaker data
|
138 |
+
if speaker_id not in metadata['speakers']:
|
139 |
+
metadata['speakers'][speaker_id] = {
|
140 |
+
'total_recordings': 0,
|
141 |
+
'datasets': {}
|
142 |
+
}
|
143 |
+
|
144 |
+
if dataset_name not in metadata['speakers'][speaker_id]['datasets']:
|
145 |
+
metadata['speakers'][speaker_id]['datasets'][dataset_name] = {
|
146 |
+
'recordings': 0,
|
147 |
+
'sentences': len(self.sentences),
|
148 |
+
'first_recording': datetime.now().isoformat(),
|
149 |
+
'last_recording': None
|
150 |
+
}
|
151 |
+
|
152 |
+
# Update counts and timestamps
|
153 |
+
metadata['speakers'][speaker_id]['total_recordings'] += 1
|
154 |
+
metadata['speakers'][speaker_id]['datasets'][dataset_name]['recordings'] += 1
|
155 |
+
metadata['speakers'][speaker_id]['datasets'][dataset_name]['last_recording'] = \
|
156 |
+
datetime.now().isoformat()
|
157 |
+
metadata['last_updated'] = datetime.now().isoformat()
|
158 |
+
|
159 |
+
# Save updated metadata
|
160 |
+
with open(metadata_file, 'w') as f:
|
161 |
+
json.dump(metadata, f, indent=2)
|
162 |
+
|
163 |
+
self.log_operation(f"Updated metadata for {speaker_id} in {dataset_name}")
|
164 |
+
|
165 |
+
except Exception as e:
|
166 |
+
error_msg = f"Error updating metadata: {str(e)}"
|
167 |
+
self.log_operation(error_msg)
|
168 |
+
print(error_msg)
|
169 |
|
170 |
def create_interface():
|
171 |
"""Create Gradio interface for TTS data collection"""
|
|
|
222 |
label="Status",
|
223 |
interactive=False
|
224 |
)
|
225 |
+
|
226 |
+
# Dataset Info
|
227 |
+
with gr.Row():
|
228 |
+
dataset_info = gr.JSON(
|
229 |
+
label="Dataset Statistics",
|
230 |
+
value={}
|
231 |
+
)
|
232 |
+
|
233 |
+
def update_dataset_info():
|
234 |
+
"""Update dataset statistics display"""
|
235 |
+
try:
|
236 |
+
metadata_file = collector.root_path / 'metadata' / 'dataset_info.json'
|
237 |
+
if metadata_file.exists():
|
238 |
+
with open(metadata_file, 'r') as f:
|
239 |
+
return json.load(f)
|
240 |
+
return {}
|
241 |
+
except Exception:
|
242 |
+
return {}
|
243 |
|
244 |
# Event handlers
|
245 |
def load_file(file):
|
|
|
248 |
current_text: "",
|
249 |
next_text: "",
|
250 |
progress: "",
|
251 |
+
status: "No file selected",
|
252 |
+
dataset_info: update_dataset_info()
|
253 |
}
|
254 |
|
255 |
success, msg = collector.load_text_file(file)
|
|
|
258 |
current_text: "",
|
259 |
next_text: "",
|
260 |
progress: "",
|
261 |
+
status: msg,
|
262 |
+
dataset_info: update_dataset_info()
|
263 |
}
|
264 |
|
265 |
return {
|
266 |
current_text: collector.sentences[0],
|
267 |
next_text: collector.sentences[1] if len(collector.sentences) > 1 else "",
|
268 |
progress: f"Sentence 1 of {len(collector.sentences)}",
|
269 |
+
status: msg,
|
270 |
+
dataset_info: update_dataset_info()
|
271 |
}
|
272 |
|
273 |
def update_display():
|
|
|
277 |
current_text: "",
|
278 |
next_text: "",
|
279 |
progress: "",
|
280 |
+
status: "No text loaded",
|
281 |
+
dataset_info: update_dataset_info()
|
282 |
}
|
283 |
|
284 |
next_idx = collector.current_index + 1
|
|
|
286 |
current_text: collector.sentences[collector.current_index],
|
287 |
next_text: collector.sentences[next_idx] if next_idx < len(collector.sentences) else "",
|
288 |
progress: f"Sentence {collector.current_index + 1} of {len(collector.sentences)}",
|
289 |
+
status: "Ready for recording",
|
290 |
+
dataset_info: update_dataset_info()
|
291 |
}
|
292 |
|
293 |
def next_sentence():
|
|
|
305 |
def save_recording(audio, spk_id, ds_name):
|
306 |
"""Handle saving recording"""
|
307 |
if not audio:
|
308 |
+
return {status: "No audio recorded", dataset_info: update_dataset_info()}
|
309 |
if not spk_id:
|
310 |
+
return {status: "Speaker ID required", dataset_info: update_dataset_info()}
|
311 |
if not ds_name:
|
312 |
+
return {status: "Dataset name required", dataset_info: update_dataset_info()}
|
313 |
|
314 |
success, msg = collector.save_recording(audio, spk_id, ds_name)
|
315 |
+
return {
|
316 |
+
status: msg,
|
317 |
+
dataset_info: update_dataset_info()
|
318 |
+
}
|
319 |
|
320 |
# Connect event handlers
|
321 |
file_input.change(
|
322 |
load_file,
|
323 |
inputs=[file_input],
|
324 |
+
outputs=[current_text, next_text, progress, status, dataset_info]
|
325 |
)
|
326 |
|
327 |
next_btn.click(
|
328 |
next_sentence,
|
329 |
+
outputs=[current_text, next_text, progress, status, dataset_info]
|
330 |
)
|
331 |
|
332 |
prev_btn.click(
|
333 |
prev_sentence,
|
334 |
+
outputs=[current_text, next_text, progress, status, dataset_info]
|
335 |
)
|
336 |
|
337 |
save_btn.click(
|
338 |
save_recording,
|
339 |
inputs=[audio_recorder, speaker_id, dataset_name],
|
340 |
+
outputs=[status, dataset_info]
|
341 |
)
|
342 |
|
343 |
return interface
|
344 |
|
345 |
if __name__ == "__main__":
|
346 |
interface = create_interface()
|
347 |
+
interface.launch()
|
|
|
|
|
|
|
|