Commit
ยท
25dcfd9
1
Parent(s):
9e0d933
Changed gradio approx real-stream to FastRTC
Browse files
app.py
CHANGED
@@ -8,6 +8,7 @@ import os
|
|
8 |
import urllib.request
|
9 |
import torchaudio
|
10 |
from scipy.spatial.distance import cosine
|
|
|
11 |
import json
|
12 |
import io
|
13 |
import wave
|
@@ -57,9 +58,6 @@ SPEAKER_COLOR_NAMES = [
|
|
57 |
]
|
58 |
|
59 |
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
class SpeechBrainEncoder:
|
64 |
"""ECAPA-TDNN encoder from SpeechBrain for speaker embeddings"""
|
65 |
def __init__(self, device="cpu"):
|
@@ -70,11 +68,24 @@ class SpeechBrainEncoder:
|
|
70 |
self.cache_dir = os.path.join(os.path.expanduser("~"), ".cache", "speechbrain")
|
71 |
os.makedirs(self.cache_dir, exist_ok=True)
|
72 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
73 |
def load_model(self):
|
74 |
"""Load the ECAPA-TDNN model"""
|
75 |
try:
|
76 |
from speechbrain.pretrained import EncoderClassifier
|
77 |
|
|
|
|
|
78 |
self.model = EncoderClassifier.from_hparams(
|
79 |
source="speechbrain/spkrec-ecapa-voxceleb",
|
80 |
savedir=self.cache_dir,
|
@@ -82,10 +93,9 @@ class SpeechBrainEncoder:
|
|
82 |
)
|
83 |
|
84 |
self.model_loaded = True
|
85 |
-
print("ECAPA-TDNN model loaded successfully!")
|
86 |
return True
|
87 |
except Exception as e:
|
88 |
-
print(f"
|
89 |
return False
|
90 |
|
91 |
def embed_utterance(self, audio, sr=16000):
|
@@ -116,21 +126,16 @@ class AudioProcessor:
|
|
116 |
def __init__(self, encoder):
|
117 |
self.encoder = encoder
|
118 |
|
119 |
-
def extract_embedding(self,
|
120 |
try:
|
121 |
-
|
122 |
-
if audio_data.dtype == np.int16:
|
123 |
-
float_audio = audio_data.astype(np.float32) / 32768.0
|
124 |
-
else:
|
125 |
-
float_audio = audio_data.astype(np.float32)
|
126 |
|
127 |
-
# Normalize if needed
|
128 |
if np.abs(float_audio).max() > 1.0:
|
129 |
float_audio = float_audio / np.abs(float_audio).max()
|
130 |
|
131 |
-
embedding = self.encoder.embed_utterance(float_audio
|
132 |
-
return embedding
|
133 |
|
|
|
134 |
except Exception as e:
|
135 |
print(f"Embedding extraction error: {e}")
|
136 |
return np.zeros(self.encoder.embedding_dim)
|
@@ -266,14 +271,68 @@ class SpeakerChangeDetector:
|
|
266 |
}
|
267 |
|
268 |
|
269 |
-
class
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
270 |
def __init__(self):
|
271 |
self.encoder = None
|
272 |
self.audio_processor = None
|
273 |
self.speaker_detector = None
|
|
|
|
|
|
|
274 |
self.full_sentences = []
|
275 |
self.sentence_speakers = []
|
276 |
-
self.
|
|
|
|
|
|
|
277 |
self.change_threshold = DEFAULT_CHANGE_THRESHOLD
|
278 |
self.max_speakers = DEFAULT_MAX_SPEAKERS
|
279 |
|
@@ -283,7 +342,6 @@ class GradioSpeakerDiarization:
|
|
283 |
device_str = "cuda" if torch.cuda.is_available() else "cpu"
|
284 |
print(f"Using device: {device_str}")
|
285 |
|
286 |
-
# Load SpeechBrain encoder
|
287 |
self.encoder = SpeechBrainEncoder(device=device_str)
|
288 |
success = self.encoder.load_model()
|
289 |
|
@@ -294,62 +352,145 @@ class GradioSpeakerDiarization:
|
|
294 |
change_threshold=self.change_threshold,
|
295 |
max_speakers=self.max_speakers
|
296 |
)
|
297 |
-
self.
|
|
|
298 |
return True
|
299 |
else:
|
|
|
300 |
return False
|
301 |
-
|
302 |
except Exception as e:
|
303 |
print(f"Model initialization error: {e}")
|
304 |
return False
|
305 |
|
306 |
-
def
|
307 |
-
"""
|
308 |
-
|
309 |
-
|
310 |
-
|
311 |
-
|
312 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
313 |
|
314 |
try:
|
315 |
-
#
|
316 |
-
|
317 |
-
|
318 |
-
|
319 |
-
|
320 |
-
|
321 |
-
|
322 |
-
|
323 |
-
|
324 |
-
|
325 |
-
|
326 |
-
|
327 |
-
|
328 |
-
|
329 |
-
|
330 |
-
|
331 |
-
|
332 |
-
|
333 |
-
|
334 |
-
|
335 |
-
|
|
|
|
|
336 |
|
337 |
-
#
|
338 |
-
|
339 |
-
self.
|
|
|
340 |
|
341 |
-
|
|
|
|
|
342 |
|
343 |
-
return
|
344 |
|
345 |
except Exception as e:
|
346 |
-
|
347 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
348 |
|
349 |
def clear_conversation(self):
|
350 |
"""Clear all conversation data"""
|
351 |
self.full_sentences = []
|
352 |
self.sentence_speakers = []
|
|
|
|
|
|
|
353 |
|
354 |
if self.speaker_detector:
|
355 |
self.speaker_detector = SpeakerChangeDetector(
|
@@ -358,7 +499,7 @@ class GradioSpeakerDiarization:
|
|
358 |
max_speakers=self.max_speakers
|
359 |
)
|
360 |
|
361 |
-
return "Conversation cleared!"
|
362 |
|
363 |
def update_settings(self, threshold, max_speakers):
|
364 |
"""Update speaker detection settings"""
|
@@ -369,22 +510,18 @@ class GradioSpeakerDiarization:
|
|
369 |
self.speaker_detector.set_change_threshold(threshold)
|
370 |
self.speaker_detector.set_max_speakers(max_speakers)
|
371 |
|
372 |
-
|
373 |
-
return status_msg, self.get_formatted_conversation(), self.get_status_info()
|
374 |
|
375 |
def get_formatted_conversation(self):
|
376 |
"""Get the formatted conversation with speaker colors"""
|
377 |
try:
|
378 |
-
if not self.full_sentences:
|
379 |
-
return "No audio processed yet. Upload an audio file or record using the microphone."
|
380 |
-
|
381 |
sentences_with_style = []
|
382 |
|
|
|
383 |
for i, sentence in enumerate(self.full_sentences):
|
384 |
sentence_text, _ = sentence
|
385 |
if i >= len(self.sentence_speakers):
|
386 |
color = "#FFFFFF"
|
387 |
-
speaker_name = "Unknown"
|
388 |
else:
|
389 |
speaker_id = self.sentence_speakers[i]
|
390 |
color = self.speaker_detector.get_color_for_speaker(speaker_id)
|
@@ -393,7 +530,15 @@ class GradioSpeakerDiarization:
|
|
393 |
sentences_with_style.append(
|
394 |
f'<span style="color:{color};"><b>{speaker_name}:</b> {sentence_text}</span>')
|
395 |
|
396 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
397 |
|
398 |
except Exception as e:
|
399 |
return f"Error formatting conversation: {e}"
|
@@ -411,7 +556,7 @@ class GradioSpeakerDiarization:
|
|
411 |
f"**Active Speakers:** {status['active_speakers']} of {status['max_speakers']}",
|
412 |
f"**Last Similarity:** {status['last_similarity']:.3f}",
|
413 |
f"**Change Threshold:** {status['threshold']:.2f}",
|
414 |
-
f"**Total
|
415 |
"",
|
416 |
"**Speaker Segment Counts:**"
|
417 |
]
|
@@ -427,21 +572,26 @@ class GradioSpeakerDiarization:
|
|
427 |
|
428 |
|
429 |
# Global instance
|
430 |
-
diarization_system =
|
431 |
|
432 |
|
433 |
def initialize_system():
|
434 |
"""Initialize the diarization system"""
|
435 |
success = diarization_system.initialize_models()
|
436 |
if success:
|
437 |
-
return "โ
System initialized successfully! Models loaded."
|
438 |
else:
|
439 |
-
return "โ Failed to initialize system. Please check the logs."
|
|
|
|
|
|
|
|
|
|
|
440 |
|
441 |
|
442 |
-
def
|
443 |
-
"""
|
444 |
-
return diarization_system.
|
445 |
|
446 |
|
447 |
def clear_conversation():
|
@@ -454,52 +604,59 @@ def update_settings(threshold, max_speakers):
|
|
454 |
return diarization_system.update_settings(threshold, max_speakers)
|
455 |
|
456 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
457 |
# Create Gradio interface
|
458 |
def create_interface():
|
459 |
-
with gr.Blocks(title="Speaker Diarization", theme=gr.themes.
|
460 |
-
gr.Markdown("# ๐ค
|
461 |
-
gr.Markdown("
|
462 |
|
463 |
with gr.Row():
|
464 |
with gr.Column(scale=2):
|
465 |
-
#
|
466 |
-
|
467 |
-
|
468 |
-
|
469 |
-
|
470 |
-
|
471 |
-
with gr.Tab("Upload Audio File"):
|
472 |
-
audio_file = gr.Audio(
|
473 |
-
label="Upload Audio File",
|
474 |
-
type="filepath",
|
475 |
-
sources=["upload"]
|
476 |
-
)
|
477 |
-
process_file_btn = gr.Button("Process Audio File", variant="secondary")
|
478 |
-
|
479 |
-
with gr.Tab("Record Audio"):
|
480 |
-
audio_mic = gr.Audio(
|
481 |
-
label="Record Audio",
|
482 |
-
type="numpy",
|
483 |
-
sources=["microphone"]
|
484 |
-
)
|
485 |
-
process_mic_btn = gr.Button("Process Recording", variant="secondary")
|
486 |
-
|
487 |
-
# Results display
|
488 |
-
status_output = gr.Textbox(
|
489 |
-
label="Status",
|
490 |
-
value="Click 'Initialize System' to start...",
|
491 |
-
lines=2,
|
492 |
-
interactive=False
|
493 |
)
|
494 |
|
|
|
495 |
conversation_output = gr.HTML(
|
496 |
-
value="<i>System
|
497 |
-
label="
|
498 |
)
|
499 |
|
500 |
# Control buttons
|
501 |
with gr.Row():
|
502 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
503 |
|
504 |
with gr.Column(scale=1):
|
505 |
# Settings panel
|
@@ -511,7 +668,7 @@ def create_interface():
|
|
511 |
step=0.05,
|
512 |
value=DEFAULT_CHANGE_THRESHOLD,
|
513 |
label="Speaker Change Sensitivity",
|
514 |
-
info="Lower = more sensitive to speaker changes"
|
515 |
)
|
516 |
|
517 |
max_speakers_slider = gr.Slider(
|
@@ -522,51 +679,106 @@ def create_interface():
|
|
522 |
label="Maximum Number of Speakers"
|
523 |
)
|
524 |
|
525 |
-
update_settings_btn = gr.Button("Update Settings"
|
526 |
|
527 |
-
#
|
528 |
-
|
529 |
-
|
530 |
-
|
531 |
-
|
532 |
-
|
533 |
-
|
|
|
|
|
|
|
534 |
|
535 |
# Speaker color legend
|
536 |
gr.Markdown("## ๐จ Speaker Colors")
|
537 |
color_info = []
|
538 |
-
for i, (color, name) in enumerate(zip(SPEAKER_COLORS
|
539 |
-
color_info.append(f'<span style="color:{color};"
|
540 |
|
541 |
-
gr.HTML("<br>".join(color_info))
|
|
|
|
|
|
|
|
|
542 |
|
543 |
# Event handlers
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
544 |
init_btn.click(
|
545 |
-
|
546 |
-
outputs=[status_output, conversation_output,
|
547 |
)
|
548 |
|
549 |
-
|
550 |
-
|
551 |
-
|
552 |
-
outputs=[status_output, conversation_output, system_status]
|
553 |
)
|
554 |
|
555 |
-
|
556 |
-
|
557 |
-
|
558 |
-
outputs=[status_output, conversation_output, system_status]
|
559 |
)
|
560 |
|
561 |
clear_btn.click(
|
562 |
clear_conversation,
|
563 |
-
outputs=[status_output
|
564 |
)
|
565 |
|
566 |
update_settings_btn.click(
|
567 |
update_settings,
|
568 |
inputs=[threshold_slider, max_speakers_slider],
|
569 |
-
outputs=[status_output
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
570 |
)
|
571 |
|
572 |
return app
|
@@ -576,5 +788,6 @@ if __name__ == "__main__":
|
|
576 |
app = create_interface()
|
577 |
app.launch(
|
578 |
server_name="0.0.0.0",
|
579 |
-
server_port=7860
|
|
|
580 |
)
|
|
|
8 |
import urllib.request
|
9 |
import torchaudio
|
10 |
from scipy.spatial.distance import cosine
|
11 |
+
from RealtimeSTT import AudioToTextRecorder
|
12 |
import json
|
13 |
import io
|
14 |
import wave
|
|
|
58 |
]
|
59 |
|
60 |
|
|
|
|
|
|
|
61 |
class SpeechBrainEncoder:
|
62 |
"""ECAPA-TDNN encoder from SpeechBrain for speaker embeddings"""
|
63 |
def __init__(self, device="cpu"):
|
|
|
68 |
self.cache_dir = os.path.join(os.path.expanduser("~"), ".cache", "speechbrain")
|
69 |
os.makedirs(self.cache_dir, exist_ok=True)
|
70 |
|
71 |
+
def _download_model(self):
|
72 |
+
"""Download pre-trained SpeechBrain ECAPA-TDNN model if not present"""
|
73 |
+
model_url = "https://huggingface.co/speechbrain/spkrec-ecapa-voxceleb/resolve/main/embedding_model.ckpt"
|
74 |
+
model_path = os.path.join(self.cache_dir, "embedding_model.ckpt")
|
75 |
+
|
76 |
+
if not os.path.exists(model_path):
|
77 |
+
print(f"Downloading ECAPA-TDNN model to {model_path}...")
|
78 |
+
urllib.request.urlretrieve(model_url, model_path)
|
79 |
+
|
80 |
+
return model_path
|
81 |
+
|
82 |
def load_model(self):
|
83 |
"""Load the ECAPA-TDNN model"""
|
84 |
try:
|
85 |
from speechbrain.pretrained import EncoderClassifier
|
86 |
|
87 |
+
model_path = self._download_model()
|
88 |
+
|
89 |
self.model = EncoderClassifier.from_hparams(
|
90 |
source="speechbrain/spkrec-ecapa-voxceleb",
|
91 |
savedir=self.cache_dir,
|
|
|
93 |
)
|
94 |
|
95 |
self.model_loaded = True
|
|
|
96 |
return True
|
97 |
except Exception as e:
|
98 |
+
print(f"Error loading ECAPA-TDNN model: {e}")
|
99 |
return False
|
100 |
|
101 |
def embed_utterance(self, audio, sr=16000):
|
|
|
126 |
def __init__(self, encoder):
|
127 |
self.encoder = encoder
|
128 |
|
129 |
+
def extract_embedding(self, audio_int16):
|
130 |
try:
|
131 |
+
float_audio = audio_int16.astype(np.float32) / 32768.0
|
|
|
|
|
|
|
|
|
132 |
|
|
|
133 |
if np.abs(float_audio).max() > 1.0:
|
134 |
float_audio = float_audio / np.abs(float_audio).max()
|
135 |
|
136 |
+
embedding = self.encoder.embed_utterance(float_audio)
|
|
|
137 |
|
138 |
+
return embedding
|
139 |
except Exception as e:
|
140 |
print(f"Embedding extraction error: {e}")
|
141 |
return np.zeros(self.encoder.embedding_dim)
|
|
|
271 |
}
|
272 |
|
273 |
|
274 |
+
class WebRTCAudioProcessor:
|
275 |
+
"""Processes WebRTC audio streams for speaker diarization"""
|
276 |
+
def __init__(self, diarization_system):
|
277 |
+
self.diarization_system = diarization_system
|
278 |
+
self.audio_buffer = []
|
279 |
+
self.buffer_lock = threading.Lock()
|
280 |
+
self.processing_thread = None
|
281 |
+
self.is_processing = False
|
282 |
+
|
283 |
+
def process_audio(self, audio_data, sample_rate):
|
284 |
+
"""Process incoming audio data from WebRTC"""
|
285 |
+
try:
|
286 |
+
# Convert audio data to numpy array if needed
|
287 |
+
if isinstance(audio_data, bytes):
|
288 |
+
audio_array = np.frombuffer(audio_data, dtype=np.int16)
|
289 |
+
elif isinstance(audio_data, tuple):
|
290 |
+
# Handle tuple format (sample_rate, audio_array)
|
291 |
+
sample_rate, audio_array = audio_data
|
292 |
+
if isinstance(audio_array, np.ndarray):
|
293 |
+
if audio_array.dtype != np.int16:
|
294 |
+
audio_array = (audio_array * 32767).astype(np.int16)
|
295 |
+
else:
|
296 |
+
audio_array = np.array(audio_array, dtype=np.int16)
|
297 |
+
else:
|
298 |
+
audio_array = np.array(audio_data, dtype=np.int16)
|
299 |
+
|
300 |
+
# Ensure mono audio
|
301 |
+
if len(audio_array.shape) > 1:
|
302 |
+
audio_array = audio_array[:, 0]
|
303 |
+
|
304 |
+
# Add to buffer
|
305 |
+
with self.buffer_lock:
|
306 |
+
self.audio_buffer.extend(audio_array)
|
307 |
+
|
308 |
+
# Process buffer when it's large enough (1 second of audio)
|
309 |
+
if len(self.audio_buffer) >= sample_rate:
|
310 |
+
buffer_to_process = np.array(self.audio_buffer[:sample_rate])
|
311 |
+
self.audio_buffer = self.audio_buffer[sample_rate//2:] # Keep 50% overlap
|
312 |
+
|
313 |
+
# Feed to recorder in separate thread
|
314 |
+
if self.diarization_system.recorder:
|
315 |
+
audio_bytes = buffer_to_process.tobytes()
|
316 |
+
self.diarization_system.recorder.feed_audio(audio_bytes)
|
317 |
+
|
318 |
+
except Exception as e:
|
319 |
+
print(f"Error processing WebRTC audio: {e}")
|
320 |
+
|
321 |
+
|
322 |
+
class RealtimeSpeakerDiarization:
|
323 |
def __init__(self):
|
324 |
self.encoder = None
|
325 |
self.audio_processor = None
|
326 |
self.speaker_detector = None
|
327 |
+
self.recorder = None
|
328 |
+
self.webrtc_processor = None
|
329 |
+
self.sentence_queue = queue.Queue()
|
330 |
self.full_sentences = []
|
331 |
self.sentence_speakers = []
|
332 |
+
self.pending_sentences = []
|
333 |
+
self.displayed_text = ""
|
334 |
+
self.last_realtime_text = ""
|
335 |
+
self.is_running = False
|
336 |
self.change_threshold = DEFAULT_CHANGE_THRESHOLD
|
337 |
self.max_speakers = DEFAULT_MAX_SPEAKERS
|
338 |
|
|
|
342 |
device_str = "cuda" if torch.cuda.is_available() else "cpu"
|
343 |
print(f"Using device: {device_str}")
|
344 |
|
|
|
345 |
self.encoder = SpeechBrainEncoder(device=device_str)
|
346 |
success = self.encoder.load_model()
|
347 |
|
|
|
352 |
change_threshold=self.change_threshold,
|
353 |
max_speakers=self.max_speakers
|
354 |
)
|
355 |
+
self.webrtc_processor = WebRTCAudioProcessor(self)
|
356 |
+
print("ECAPA-TDNN model loaded successfully!")
|
357 |
return True
|
358 |
else:
|
359 |
+
print("Failed to load ECAPA-TDNN model")
|
360 |
return False
|
|
|
361 |
except Exception as e:
|
362 |
print(f"Model initialization error: {e}")
|
363 |
return False
|
364 |
|
365 |
+
def live_text_detected(self, text):
|
366 |
+
"""Callback for real-time transcription updates"""
|
367 |
+
text = text.strip()
|
368 |
+
if text:
|
369 |
+
sentence_delimiters = '.?!ใ'
|
370 |
+
prob_sentence_end = (
|
371 |
+
len(self.last_realtime_text) > 0
|
372 |
+
and text[-1] in sentence_delimiters
|
373 |
+
and self.last_realtime_text[-1] in sentence_delimiters
|
374 |
+
)
|
375 |
+
|
376 |
+
self.last_realtime_text = text
|
377 |
+
|
378 |
+
if prob_sentence_end and FAST_SENTENCE_END:
|
379 |
+
self.recorder.stop()
|
380 |
+
elif prob_sentence_end:
|
381 |
+
self.recorder.post_speech_silence_duration = SILENCE_THRESHS[0]
|
382 |
+
else:
|
383 |
+
self.recorder.post_speech_silence_duration = SILENCE_THRESHS[1]
|
384 |
+
|
385 |
+
def process_final_text(self, text):
|
386 |
+
"""Process final transcribed text with speaker embedding"""
|
387 |
+
text = text.strip()
|
388 |
+
if text:
|
389 |
+
try:
|
390 |
+
bytes_data = self.recorder.last_transcription_bytes
|
391 |
+
self.sentence_queue.put((text, bytes_data))
|
392 |
+
self.pending_sentences.append(text)
|
393 |
+
except Exception as e:
|
394 |
+
print(f"Error processing final text: {e}")
|
395 |
+
|
396 |
+
def process_sentence_queue(self):
|
397 |
+
"""Process sentences in the queue for speaker detection"""
|
398 |
+
while self.is_running:
|
399 |
+
try:
|
400 |
+
text, bytes_data = self.sentence_queue.get(timeout=1)
|
401 |
+
|
402 |
+
# Convert audio data to int16
|
403 |
+
audio_int16 = np.int16(bytes_data * 32767)
|
404 |
+
|
405 |
+
# Extract speaker embedding
|
406 |
+
speaker_embedding = self.audio_processor.extract_embedding(audio_int16)
|
407 |
+
|
408 |
+
# Store sentence and embedding
|
409 |
+
self.full_sentences.append((text, speaker_embedding))
|
410 |
+
|
411 |
+
# Fill in missing speaker assignments
|
412 |
+
while len(self.sentence_speakers) < len(self.full_sentences) - 1:
|
413 |
+
self.sentence_speakers.append(0)
|
414 |
+
|
415 |
+
# Detect speaker changes
|
416 |
+
speaker_id, similarity = self.speaker_detector.add_embedding(speaker_embedding)
|
417 |
+
self.sentence_speakers.append(speaker_id)
|
418 |
+
|
419 |
+
# Remove from pending
|
420 |
+
if text in self.pending_sentences:
|
421 |
+
self.pending_sentences.remove(text)
|
422 |
+
|
423 |
+
except queue.Empty:
|
424 |
+
continue
|
425 |
+
except Exception as e:
|
426 |
+
print(f"Error processing sentence: {e}")
|
427 |
+
|
428 |
+
def start_recording(self):
|
429 |
+
"""Start the recording and transcription process"""
|
430 |
+
if self.encoder is None:
|
431 |
+
return "Please initialize models first!"
|
432 |
|
433 |
try:
|
434 |
+
# Setup recorder configuration for WebRTC input
|
435 |
+
recorder_config = {
|
436 |
+
'spinner': False,
|
437 |
+
'use_microphone': False, # We'll feed audio manually
|
438 |
+
'model': FINAL_TRANSCRIPTION_MODEL,
|
439 |
+
'language': TRANSCRIPTION_LANGUAGE,
|
440 |
+
'silero_sensitivity': SILERO_SENSITIVITY,
|
441 |
+
'webrtc_sensitivity': WEBRTC_SENSITIVITY,
|
442 |
+
'post_speech_silence_duration': SILENCE_THRESHS[1],
|
443 |
+
'min_length_of_recording': MIN_LENGTH_OF_RECORDING,
|
444 |
+
'pre_recording_buffer_duration': PRE_RECORDING_BUFFER_DURATION,
|
445 |
+
'min_gap_between_recordings': 0,
|
446 |
+
'enable_realtime_transcription': True,
|
447 |
+
'realtime_processing_pause': 0,
|
448 |
+
'realtime_model_type': REALTIME_TRANSCRIPTION_MODEL,
|
449 |
+
'on_realtime_transcription_update': self.live_text_detected,
|
450 |
+
'beam_size': FINAL_BEAM_SIZE,
|
451 |
+
'beam_size_realtime': REALTIME_BEAM_SIZE,
|
452 |
+
'buffer_size': BUFFER_SIZE,
|
453 |
+
'sample_rate': SAMPLE_RATE,
|
454 |
+
}
|
455 |
+
|
456 |
+
self.recorder = AudioToTextRecorder(**recorder_config)
|
457 |
|
458 |
+
# Start sentence processing thread
|
459 |
+
self.is_running = True
|
460 |
+
self.sentence_thread = threading.Thread(target=self.process_sentence_queue, daemon=True)
|
461 |
+
self.sentence_thread.start()
|
462 |
|
463 |
+
# Start transcription thread
|
464 |
+
self.transcription_thread = threading.Thread(target=self.run_transcription, daemon=True)
|
465 |
+
self.transcription_thread.start()
|
466 |
|
467 |
+
return "Recording started successfully! WebRTC audio input ready."
|
468 |
|
469 |
except Exception as e:
|
470 |
+
return f"Error starting recording: {e}"
|
471 |
+
|
472 |
+
def run_transcription(self):
|
473 |
+
"""Run the transcription loop"""
|
474 |
+
try:
|
475 |
+
while self.is_running:
|
476 |
+
self.recorder.text(self.process_final_text)
|
477 |
+
except Exception as e:
|
478 |
+
print(f"Transcription error: {e}")
|
479 |
+
|
480 |
+
def stop_recording(self):
|
481 |
+
"""Stop the recording process"""
|
482 |
+
self.is_running = False
|
483 |
+
if self.recorder:
|
484 |
+
self.recorder.stop()
|
485 |
+
return "Recording stopped!"
|
486 |
|
487 |
def clear_conversation(self):
|
488 |
"""Clear all conversation data"""
|
489 |
self.full_sentences = []
|
490 |
self.sentence_speakers = []
|
491 |
+
self.pending_sentences = []
|
492 |
+
self.displayed_text = ""
|
493 |
+
self.last_realtime_text = ""
|
494 |
|
495 |
if self.speaker_detector:
|
496 |
self.speaker_detector = SpeakerChangeDetector(
|
|
|
499 |
max_speakers=self.max_speakers
|
500 |
)
|
501 |
|
502 |
+
return "Conversation cleared!"
|
503 |
|
504 |
def update_settings(self, threshold, max_speakers):
|
505 |
"""Update speaker detection settings"""
|
|
|
510 |
self.speaker_detector.set_change_threshold(threshold)
|
511 |
self.speaker_detector.set_max_speakers(max_speakers)
|
512 |
|
513 |
+
return f"Settings updated: Threshold={threshold:.2f}, Max Speakers={max_speakers}"
|
|
|
514 |
|
515 |
def get_formatted_conversation(self):
|
516 |
"""Get the formatted conversation with speaker colors"""
|
517 |
try:
|
|
|
|
|
|
|
518 |
sentences_with_style = []
|
519 |
|
520 |
+
# Process completed sentences
|
521 |
for i, sentence in enumerate(self.full_sentences):
|
522 |
sentence_text, _ = sentence
|
523 |
if i >= len(self.sentence_speakers):
|
524 |
color = "#FFFFFF"
|
|
|
525 |
else:
|
526 |
speaker_id = self.sentence_speakers[i]
|
527 |
color = self.speaker_detector.get_color_for_speaker(speaker_id)
|
|
|
530 |
sentences_with_style.append(
|
531 |
f'<span style="color:{color};"><b>{speaker_name}:</b> {sentence_text}</span>')
|
532 |
|
533 |
+
# Add pending sentences
|
534 |
+
for pending_sentence in self.pending_sentences:
|
535 |
+
sentences_with_style.append(
|
536 |
+
f'<span style="color:#60FFFF;"><b>Processing:</b> {pending_sentence}</span>')
|
537 |
+
|
538 |
+
if sentences_with_style:
|
539 |
+
return "<br><br>".join(sentences_with_style)
|
540 |
+
else:
|
541 |
+
return "Waiting for speech input..."
|
542 |
|
543 |
except Exception as e:
|
544 |
return f"Error formatting conversation: {e}"
|
|
|
556 |
f"**Active Speakers:** {status['active_speakers']} of {status['max_speakers']}",
|
557 |
f"**Last Similarity:** {status['last_similarity']:.3f}",
|
558 |
f"**Change Threshold:** {status['threshold']:.2f}",
|
559 |
+
f"**Total Sentences:** {len(self.full_sentences)}",
|
560 |
"",
|
561 |
"**Speaker Segment Counts:**"
|
562 |
]
|
|
|
572 |
|
573 |
|
574 |
# Global instance
|
575 |
+
diarization_system = RealtimeSpeakerDiarization()
|
576 |
|
577 |
|
578 |
def initialize_system():
|
579 |
"""Initialize the diarization system"""
|
580 |
success = diarization_system.initialize_models()
|
581 |
if success:
|
582 |
+
return "โ
System initialized successfully! Models loaded."
|
583 |
else:
|
584 |
+
return "โ Failed to initialize system. Please check the logs."
|
585 |
+
|
586 |
+
|
587 |
+
def start_recording():
|
588 |
+
"""Start recording and transcription"""
|
589 |
+
return diarization_system.start_recording()
|
590 |
|
591 |
|
592 |
+
def stop_recording():
|
593 |
+
"""Stop recording and transcription"""
|
594 |
+
return diarization_system.stop_recording()
|
595 |
|
596 |
|
597 |
def clear_conversation():
|
|
|
604 |
return diarization_system.update_settings(threshold, max_speakers)
|
605 |
|
606 |
|
607 |
+
def get_conversation():
|
608 |
+
"""Get the current conversation"""
|
609 |
+
return diarization_system.get_formatted_conversation()
|
610 |
+
|
611 |
+
|
612 |
+
def get_status():
|
613 |
+
"""Get system status"""
|
614 |
+
return diarization_system.get_status_info()
|
615 |
+
|
616 |
+
|
617 |
+
def process_audio_stream(audio):
|
618 |
+
"""Process audio stream from WebRTC"""
|
619 |
+
if diarization_system.webrtc_processor and diarization_system.is_running:
|
620 |
+
diarization_system.webrtc_processor.process_audio(audio, SAMPLE_RATE)
|
621 |
+
return None
|
622 |
+
|
623 |
+
|
624 |
# Create Gradio interface
|
625 |
def create_interface():
|
626 |
+
with gr.Blocks(title="Real-time Speaker Diarization", theme=gr.themes.Dark()) as app:
|
627 |
+
gr.Markdown("# ๐ค Real-time Speech Recognition with Speaker Diarization")
|
628 |
+
gr.Markdown("This app performs real-time speech recognition with automatic speaker identification and color-coding using WebRTC.")
|
629 |
|
630 |
with gr.Row():
|
631 |
with gr.Column(scale=2):
|
632 |
+
# WebRTC Audio Input
|
633 |
+
audio_input = gr.Audio(
|
634 |
+
sources=["microphone"],
|
635 |
+
streaming=True,
|
636 |
+
label="๐๏ธ Microphone Input",
|
637 |
+
type="numpy"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
638 |
)
|
639 |
|
640 |
+
# Main conversation display
|
641 |
conversation_output = gr.HTML(
|
642 |
+
value="<i>Click 'Initialize System' to start...</i>",
|
643 |
+
label="Live Conversation"
|
644 |
)
|
645 |
|
646 |
# Control buttons
|
647 |
with gr.Row():
|
648 |
+
init_btn = gr.Button("๐ง Initialize System", variant="secondary")
|
649 |
+
start_btn = gr.Button("๐๏ธ Start Recording", variant="primary", interactive=False)
|
650 |
+
stop_btn = gr.Button("โน๏ธ Stop Recording", variant="stop", interactive=False)
|
651 |
+
clear_btn = gr.Button("๐๏ธ Clear Conversation", interactive=False)
|
652 |
+
|
653 |
+
# Status display
|
654 |
+
status_output = gr.Textbox(
|
655 |
+
label="System Status",
|
656 |
+
value="System not initialized",
|
657 |
+
lines=8,
|
658 |
+
interactive=False
|
659 |
+
)
|
660 |
|
661 |
with gr.Column(scale=1):
|
662 |
# Settings panel
|
|
|
668 |
step=0.05,
|
669 |
value=DEFAULT_CHANGE_THRESHOLD,
|
670 |
label="Speaker Change Sensitivity",
|
671 |
+
info="Lower values = more sensitive to speaker changes"
|
672 |
)
|
673 |
|
674 |
max_speakers_slider = gr.Slider(
|
|
|
679 |
label="Maximum Number of Speakers"
|
680 |
)
|
681 |
|
682 |
+
update_settings_btn = gr.Button("Update Settings")
|
683 |
|
684 |
+
# Instructions
|
685 |
+
gr.Markdown("## ๐ Instructions")
|
686 |
+
gr.Markdown("""
|
687 |
+
1. Click **Initialize System** to load models
|
688 |
+
2. Click **Start Recording** to begin processing
|
689 |
+
3. Allow microphone access when prompted
|
690 |
+
4. Speak into your microphone
|
691 |
+
5. Watch real-time transcription with speaker labels
|
692 |
+
6. Adjust settings as needed
|
693 |
+
""")
|
694 |
|
695 |
# Speaker color legend
|
696 |
gr.Markdown("## ๐จ Speaker Colors")
|
697 |
color_info = []
|
698 |
+
for i, (color, name) in enumerate(zip(SPEAKER_COLORS, SPEAKER_COLOR_NAMES)):
|
699 |
+
color_info.append(f'<span style="color:{color};">โ </span> Speaker {i+1} ({name})')
|
700 |
|
701 |
+
gr.HTML("<br>".join(color_info[:DEFAULT_MAX_SPEAKERS]))
|
702 |
+
|
703 |
+
# Auto-refresh conversation and status
|
704 |
+
def refresh_display():
|
705 |
+
return get_conversation(), get_status()
|
706 |
|
707 |
# Event handlers
|
708 |
+
def on_initialize():
|
709 |
+
result = initialize_system()
|
710 |
+
if "successfully" in result:
|
711 |
+
return (
|
712 |
+
result,
|
713 |
+
gr.update(interactive=True), # start_btn
|
714 |
+
gr.update(interactive=True), # clear_btn
|
715 |
+
get_conversation(),
|
716 |
+
get_status()
|
717 |
+
)
|
718 |
+
else:
|
719 |
+
return (
|
720 |
+
result,
|
721 |
+
gr.update(interactive=False), # start_btn
|
722 |
+
gr.update(interactive=False), # clear_btn
|
723 |
+
get_conversation(),
|
724 |
+
get_status()
|
725 |
+
)
|
726 |
+
|
727 |
+
def on_start():
|
728 |
+
result = start_recording()
|
729 |
+
return (
|
730 |
+
result,
|
731 |
+
gr.update(interactive=False), # start_btn
|
732 |
+
gr.update(interactive=True), # stop_btn
|
733 |
+
)
|
734 |
+
|
735 |
+
def on_stop():
|
736 |
+
result = stop_recording()
|
737 |
+
return (
|
738 |
+
result,
|
739 |
+
gr.update(interactive=True), # start_btn
|
740 |
+
gr.update(interactive=False), # stop_btn
|
741 |
+
)
|
742 |
+
|
743 |
+
# Connect event handlers
|
744 |
init_btn.click(
|
745 |
+
on_initialize,
|
746 |
+
outputs=[status_output, start_btn, clear_btn, conversation_output, status_output]
|
747 |
)
|
748 |
|
749 |
+
start_btn.click(
|
750 |
+
on_start,
|
751 |
+
outputs=[status_output, start_btn, stop_btn]
|
|
|
752 |
)
|
753 |
|
754 |
+
stop_btn.click(
|
755 |
+
on_stop,
|
756 |
+
outputs=[status_output, start_btn, stop_btn]
|
|
|
757 |
)
|
758 |
|
759 |
clear_btn.click(
|
760 |
clear_conversation,
|
761 |
+
outputs=[status_output]
|
762 |
)
|
763 |
|
764 |
update_settings_btn.click(
|
765 |
update_settings,
|
766 |
inputs=[threshold_slider, max_speakers_slider],
|
767 |
+
outputs=[status_output]
|
768 |
+
)
|
769 |
+
|
770 |
+
# Connect WebRTC audio stream to processing
|
771 |
+
audio_input.stream(
|
772 |
+
process_audio_stream,
|
773 |
+
inputs=[audio_input],
|
774 |
+
outputs=[]
|
775 |
+
)
|
776 |
+
|
777 |
+
# Auto-refresh every 2 seconds when recording
|
778 |
+
refresh_timer = gr.Timer(2.0)
|
779 |
+
refresh_timer.tick(
|
780 |
+
refresh_display,
|
781 |
+
outputs=[conversation_output, status_output]
|
782 |
)
|
783 |
|
784 |
return app
|
|
|
788 |
app = create_interface()
|
789 |
app.launch(
|
790 |
server_name="0.0.0.0",
|
791 |
+
server_port=7860,
|
792 |
+
share=True
|
793 |
)
|