Spaces:
Sleeping
Sleeping
Commit
ยท
c263c26
1
Parent(s):
29b89b3
Code error correction
Browse files
app.py
CHANGED
@@ -9,8 +9,13 @@ import urllib.request
|
|
9 |
import torchaudio
|
10 |
from scipy.spatial.distance import cosine
|
11 |
import json
|
12 |
-
import
|
13 |
-
import
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
# Simplified configuration parameters
|
16 |
SILENCE_THRESHS = [0, 0.4]
|
@@ -34,8 +39,9 @@ ABSOLUTE_MAX_SPEAKERS = 10
|
|
34 |
# Global variables
|
35 |
FAST_SENTENCE_END = True
|
36 |
SAMPLE_RATE = 16000
|
37 |
-
BUFFER_SIZE =
|
38 |
CHANNELS = 1
|
|
|
39 |
|
40 |
# Speaker colors
|
41 |
SPEAKER_COLORS = [
|
@@ -73,7 +79,7 @@ class SpeechBrainEncoder:
|
|
73 |
model_path = os.path.join(self.cache_dir, "embedding_model.ckpt")
|
74 |
|
75 |
if not os.path.exists(model_path):
|
76 |
-
|
77 |
urllib.request.urlretrieve(model_url, model_path)
|
78 |
|
79 |
return model_path
|
@@ -94,7 +100,7 @@ class SpeechBrainEncoder:
|
|
94 |
self.model_loaded = True
|
95 |
return True
|
96 |
except Exception as e:
|
97 |
-
|
98 |
return False
|
99 |
|
100 |
def embed_utterance(self, audio, sr=16000):
|
@@ -116,7 +122,7 @@ class SpeechBrainEncoder:
|
|
116 |
|
117 |
return embedding.squeeze().cpu().numpy()
|
118 |
except Exception as e:
|
119 |
-
|
120 |
return np.zeros(self.embedding_dim)
|
121 |
|
122 |
|
@@ -135,7 +141,7 @@ class AudioProcessor:
|
|
135 |
|
136 |
return embedding
|
137 |
except Exception as e:
|
138 |
-
|
139 |
return np.zeros(self.encoder.embedding_dim)
|
140 |
|
141 |
|
@@ -270,83 +276,105 @@ class SpeakerChangeDetector:
|
|
270 |
|
271 |
|
272 |
class WhisperTranscriber:
|
273 |
-
"""
|
274 |
def __init__(self, model_name="distil-large-v3"):
|
275 |
self.model = None
|
276 |
self.processor = None
|
277 |
self.model_name = model_name
|
278 |
-
self.
|
279 |
|
280 |
def load_model(self):
|
281 |
"""Load Whisper model"""
|
282 |
try:
|
283 |
from transformers import WhisperProcessor, WhisperForConditionalGeneration
|
284 |
|
285 |
-
|
286 |
-
|
287 |
-
self.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
288 |
|
|
|
289 |
return True
|
290 |
except Exception as e:
|
291 |
-
|
292 |
return False
|
293 |
|
294 |
def transcribe(self, audio_array, sample_rate=16000):
|
295 |
"""Transcribe audio array"""
|
|
|
|
|
|
|
296 |
try:
|
297 |
-
|
|
|
298 |
return ""
|
299 |
|
300 |
-
#
|
301 |
if sample_rate != 16000:
|
302 |
-
|
303 |
-
|
304 |
-
|
305 |
-
|
306 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
307 |
|
308 |
-
|
309 |
-
|
310 |
-
inputs = inputs.to(self.device)
|
311 |
|
312 |
-
# Generate transcription
|
313 |
with torch.no_grad():
|
314 |
-
predicted_ids = self.model.generate(
|
315 |
-
|
316 |
-
|
317 |
-
|
|
|
|
|
|
|
318 |
|
319 |
-
|
320 |
|
|
|
321 |
except Exception as e:
|
322 |
-
|
323 |
return ""
|
324 |
|
325 |
|
326 |
-
class
|
327 |
def __init__(self):
|
328 |
self.encoder = None
|
329 |
self.audio_processor = None
|
330 |
self.speaker_detector = None
|
331 |
self.transcriber = None
|
332 |
-
self.
|
333 |
self.processing_thread = None
|
334 |
-
self.sentence_queue = queue.Queue()
|
335 |
self.full_sentences = []
|
336 |
self.sentence_speakers = []
|
337 |
-
self.pending_sentences = []
|
338 |
-
self.displayed_text = ""
|
339 |
self.is_running = False
|
340 |
self.change_threshold = DEFAULT_CHANGE_THRESHOLD
|
341 |
self.max_speakers = DEFAULT_MAX_SPEAKERS
|
342 |
-
self.
|
343 |
-
self.
|
|
|
|
|
344 |
|
345 |
def initialize_models(self):
|
346 |
"""Initialize the speaker encoder and transcription models"""
|
347 |
try:
|
348 |
device_str = "cuda" if torch.cuda.is_available() else "cpu"
|
349 |
-
|
350 |
|
351 |
# Initialize speaker encoder
|
352 |
self.encoder = SpeechBrainEncoder(device=device_str)
|
@@ -363,124 +391,131 @@ class RealtimeSpeakerDiarization:
|
|
363 |
change_threshold=self.change_threshold,
|
364 |
max_speakers=self.max_speakers
|
365 |
)
|
366 |
-
|
367 |
return True
|
368 |
else:
|
369 |
-
|
370 |
return False
|
371 |
except Exception as e:
|
372 |
-
|
373 |
return False
|
374 |
|
375 |
-
def
|
376 |
-
"""Process
|
377 |
-
if not self.is_running or
|
378 |
return
|
379 |
|
380 |
try:
|
381 |
-
#
|
382 |
-
if isinstance(
|
383 |
-
|
384 |
-
|
385 |
-
|
386 |
-
|
387 |
-
# Ensure audio is float32 and normalized
|
388 |
-
if audio_array.dtype != np.float32:
|
389 |
-
if audio_array.dtype == np.int16:
|
390 |
-
audio_array = audio_array.astype(np.float32) / 32768.0
|
391 |
-
else:
|
392 |
-
audio_array = audio_array.astype(np.float32)
|
393 |
-
|
394 |
-
# Ensure mono audio
|
395 |
-
if len(audio_array.shape) > 1 and audio_array.shape[1] > 1:
|
396 |
-
audio_array = np.mean(audio_array, axis=1)
|
397 |
-
|
398 |
-
# Add to buffer
|
399 |
-
self.audio_buffer.extend(audio_array.flatten())
|
400 |
-
|
401 |
-
# Process when we have enough audio (about 2 seconds)
|
402 |
-
target_length = int(sample_rate * 2.0)
|
403 |
-
if len(self.audio_buffer) >= target_length:
|
404 |
-
self.process_audio_chunk()
|
405 |
|
406 |
-
|
407 |
-
|
408 |
-
|
409 |
-
def process_audio_chunk(self):
|
410 |
-
"""Process accumulated audio chunk"""
|
411 |
-
try:
|
412 |
-
if len(self.audio_buffer) < SAMPLE_RATE: # Need at least 1 second
|
413 |
-
return
|
414 |
-
|
415 |
-
# Get audio chunk
|
416 |
-
audio_chunk = np.array(self.audio_buffer[:int(SAMPLE_RATE * 2)])
|
417 |
-
self.audio_buffer = self.audio_buffer[int(SAMPLE_RATE * 1.5):] # Keep some overlap
|
418 |
-
|
419 |
-
# Transcribe audio
|
420 |
-
transcription = self.transcriber.transcribe(audio_chunk, SAMPLE_RATE)
|
421 |
-
|
422 |
-
if transcription.strip():
|
423 |
-
# Extract speaker embedding
|
424 |
-
speaker_embedding = self.audio_processor.extract_embedding(audio_chunk)
|
425 |
|
426 |
-
|
427 |
-
|
428 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
429 |
except Exception as e:
|
430 |
-
|
431 |
|
432 |
-
def
|
433 |
-
"""Process
|
434 |
while self.is_running:
|
435 |
try:
|
436 |
-
|
437 |
-
|
438 |
-
# Store sentence and embedding
|
439 |
-
self.full_sentences.append((text, speaker_embedding))
|
440 |
|
441 |
-
#
|
442 |
-
|
443 |
-
self.sentence_speakers.append(0)
|
444 |
|
445 |
-
#
|
446 |
-
|
447 |
-
self.sentence_speakers.append(speaker_id)
|
448 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
449 |
except queue.Empty:
|
450 |
continue
|
451 |
except Exception as e:
|
452 |
-
|
453 |
|
454 |
def start_recording(self):
|
455 |
-
"""Start the recording and
|
456 |
-
if self.encoder is None:
|
457 |
return "Please initialize models first!"
|
458 |
|
459 |
try:
|
460 |
-
# Start sentence processing thread
|
461 |
self.is_running = True
|
462 |
-
self.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
463 |
self.processing_thread.start()
|
464 |
|
465 |
-
|
|
|
466 |
|
467 |
except Exception as e:
|
|
|
468 |
return f"Error starting recording: {e}"
|
469 |
|
470 |
def stop_recording(self):
|
471 |
"""Stop the recording process"""
|
472 |
self.is_running = False
|
473 |
-
|
474 |
return "Recording stopped!"
|
475 |
|
476 |
def clear_conversation(self):
|
477 |
"""Clear all conversation data"""
|
478 |
self.full_sentences = []
|
479 |
self.sentence_speakers = []
|
480 |
-
self.pending_sentences = []
|
481 |
-
self.displayed_text = ""
|
482 |
self.audio_buffer = []
|
483 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
484 |
if self.speaker_detector:
|
485 |
self.speaker_detector = SpeakerChangeDetector(
|
486 |
embedding_dim=self.encoder.embedding_dim,
|
@@ -504,26 +539,24 @@ class RealtimeSpeakerDiarization:
|
|
504 |
def get_formatted_conversation(self):
|
505 |
"""Get the formatted conversation with speaker colors"""
|
506 |
try:
|
|
|
|
|
|
|
507 |
sentences_with_style = []
|
508 |
|
509 |
-
#
|
510 |
-
for i, sentence in enumerate(self.full_sentences):
|
511 |
-
sentence_text, _ = sentence
|
512 |
if i >= len(self.sentence_speakers):
|
513 |
color = "#FFFFFF"
|
514 |
-
speaker_name = "
|
515 |
else:
|
516 |
-
speaker_id = self.sentence_speakers[i]
|
517 |
color = self.speaker_detector.get_color_for_speaker(speaker_id)
|
518 |
speaker_name = f"Speaker {speaker_id + 1}"
|
519 |
-
|
520 |
sentences_with_style.append(
|
521 |
-
f'<span style="color:{color};"
|
522 |
|
523 |
-
|
524 |
-
return "<br><br>".join(sentences_with_style)
|
525 |
-
else:
|
526 |
-
return "Waiting for speech input..."
|
527 |
|
528 |
except Exception as e:
|
529 |
return f"Error formatting conversation: {e}"
|
@@ -535,6 +568,7 @@ class RealtimeSpeakerDiarization:
|
|
535 |
|
536 |
try:
|
537 |
status = self.speaker_detector.get_status_info()
|
|
|
538 |
|
539 |
status_lines = [
|
540 |
f"**Current Speaker:** {status['current_speaker'] + 1}",
|
@@ -542,7 +576,8 @@ class RealtimeSpeakerDiarization:
|
|
542 |
f"**Last Similarity:** {status['last_similarity']:.3f}",
|
543 |
f"**Change Threshold:** {status['threshold']:.2f}",
|
544 |
f"**Total Sentences:** {len(self.full_sentences)}",
|
545 |
-
f"**
|
|
|
546 |
"",
|
547 |
"**Speaker Segment Counts:**"
|
548 |
]
|
@@ -558,7 +593,7 @@ class RealtimeSpeakerDiarization:
|
|
558 |
|
559 |
|
560 |
# Global instance
|
561 |
-
diarization_system =
|
562 |
|
563 |
|
564 |
def initialize_system():
|
@@ -600,49 +635,56 @@ def get_status():
|
|
600 |
return diarization_system.get_status_info()
|
601 |
|
602 |
|
603 |
-
def
|
604 |
-
"""Process audio from
|
605 |
-
if
|
606 |
-
sample_rate,
|
607 |
-
diarization_system.
|
|
|
608 |
return get_conversation(), get_status()
|
609 |
|
610 |
|
611 |
-
# Create Gradio interface
|
612 |
def create_interface():
|
613 |
-
with gr.Blocks(title="Real-time Speaker Diarization", theme=gr.themes.
|
614 |
-
gr.Markdown("# ๐ค Real-time Speech Recognition with Speaker Diarization")
|
615 |
-
gr.Markdown("This app
|
616 |
|
617 |
with gr.Row():
|
618 |
with gr.Column(scale=2):
|
619 |
-
# Audio input
|
620 |
audio_input = gr.Audio(
|
621 |
-
|
622 |
type="numpy",
|
623 |
streaming=True,
|
624 |
-
label="๐๏ธ Microphone Input"
|
|
|
|
|
|
|
|
|
625 |
)
|
626 |
|
627 |
# Main conversation display
|
628 |
conversation_output = gr.HTML(
|
629 |
-
value="<i>Click 'Initialize System' to
|
630 |
-
label="Live Conversation"
|
|
|
631 |
)
|
632 |
|
633 |
# Control buttons
|
634 |
with gr.Row():
|
635 |
-
init_btn = gr.Button("๐ง Initialize System", variant="secondary")
|
636 |
-
start_btn = gr.Button("๐๏ธ Start Recording", variant="primary", interactive=False)
|
637 |
-
stop_btn = gr.Button("โน๏ธ Stop Recording", variant="stop", interactive=False)
|
638 |
-
clear_btn = gr.Button("๐๏ธ Clear
|
639 |
|
640 |
# Status display
|
641 |
status_output = gr.Textbox(
|
642 |
label="System Status",
|
643 |
value="System not initialized",
|
644 |
lines=10,
|
645 |
-
interactive=False
|
|
|
646 |
)
|
647 |
|
648 |
with gr.Column(scale=1):
|
@@ -655,7 +697,7 @@ def create_interface():
|
|
655 |
step=0.05,
|
656 |
value=DEFAULT_CHANGE_THRESHOLD,
|
657 |
label="Speaker Change Sensitivity",
|
658 |
-
info="Lower
|
659 |
)
|
660 |
|
661 |
max_speakers_slider = gr.Slider(
|
@@ -666,26 +708,23 @@ def create_interface():
|
|
666 |
label="Maximum Number of Speakers"
|
667 |
)
|
668 |
|
669 |
-
update_settings_btn = gr.Button("Update Settings")
|
670 |
|
671 |
# Speaker color legend
|
672 |
gr.Markdown("## ๐จ Speaker Colors")
|
673 |
color_info = []
|
674 |
for i, (color, name) in enumerate(zip(SPEAKER_COLORS, SPEAKER_COLOR_NAMES)):
|
675 |
-
color_info.append(f'<span style="color:{color};"
|
676 |
|
677 |
gr.HTML("<br>".join(color_info[:DEFAULT_MAX_SPEAKERS]))
|
678 |
|
679 |
-
#
|
|
|
680 |
gr.Markdown("""
|
681 |
-
|
682 |
-
|
683 |
-
|
684 |
-
|
685 |
-
4. **Speak naturally** - The system will detect different speakers
|
686 |
-
5. **Stop Recording** when done
|
687 |
-
|
688 |
-
**Note:** Processing happens in real-time with ~2 second chunks for better accuracy.
|
689 |
""")
|
690 |
|
691 |
# Event handlers
|
@@ -693,25 +732,25 @@ def create_interface():
|
|
693 |
result = initialize_system()
|
694 |
if "successfully" in result:
|
695 |
return (
|
696 |
-
result,
|
697 |
gr.update(interactive=True), # start_btn
|
698 |
gr.update(interactive=True), # clear_btn
|
699 |
-
get_conversation(),
|
700 |
-
get_status()
|
701 |
)
|
702 |
else:
|
703 |
return (
|
704 |
-
result,
|
705 |
gr.update(interactive=False), # start_btn
|
706 |
gr.update(interactive=False), # clear_btn
|
707 |
-
get_conversation(),
|
708 |
-
get_status()
|
709 |
)
|
710 |
|
711 |
def on_start():
|
712 |
result = start_recording()
|
713 |
return (
|
714 |
-
result,
|
715 |
gr.update(interactive=False), # start_btn
|
716 |
gr.update(interactive=True), # stop_btn
|
717 |
)
|
@@ -719,11 +758,15 @@ def create_interface():
|
|
719 |
def on_stop():
|
720 |
result = stop_recording()
|
721 |
return (
|
722 |
-
result,
|
723 |
gr.update(interactive=True), # start_btn
|
724 |
gr.update(interactive=False), # stop_btn
|
725 |
)
|
726 |
|
|
|
|
|
|
|
|
|
727 |
# Connect event handlers
|
728 |
init_btn.click(
|
729 |
on_initialize,
|
@@ -751,19 +794,19 @@ def create_interface():
|
|
751 |
outputs=[status_output]
|
752 |
)
|
753 |
|
754 |
-
#
|
755 |
audio_input.stream(
|
756 |
-
|
757 |
inputs=[audio_input],
|
758 |
outputs=[conversation_output, status_output],
|
759 |
-
|
760 |
-
|
761 |
)
|
762 |
|
763 |
-
# Auto-refresh
|
764 |
-
refresh_timer = gr.Timer(
|
765 |
refresh_timer.tick(
|
766 |
-
|
767 |
outputs=[conversation_output, status_output]
|
768 |
)
|
769 |
|
|
|
9 |
import torchaudio
|
10 |
from scipy.spatial.distance import cosine
|
11 |
import json
|
12 |
+
import asyncio
|
13 |
+
from typing import Iterator
|
14 |
+
import logging
|
15 |
+
|
16 |
+
# Configure logging
|
17 |
+
logging.basicConfig(level=logging.INFO)
|
18 |
+
logger = logging.getLogger(__name__)
|
19 |
|
20 |
# Simplified configuration parameters
|
21 |
SILENCE_THRESHS = [0, 0.4]
|
|
|
39 |
# Global variables
|
40 |
FAST_SENTENCE_END = True
|
41 |
SAMPLE_RATE = 16000
|
42 |
+
BUFFER_SIZE = 1024
|
43 |
CHANNELS = 1
|
44 |
+
CHUNK_DURATION_MS = 100 # 100ms chunks for FastRTC
|
45 |
|
46 |
# Speaker colors
|
47 |
SPEAKER_COLORS = [
|
|
|
79 |
model_path = os.path.join(self.cache_dir, "embedding_model.ckpt")
|
80 |
|
81 |
if not os.path.exists(model_path):
|
82 |
+
logger.info(f"Downloading ECAPA-TDNN model to {model_path}...")
|
83 |
urllib.request.urlretrieve(model_url, model_path)
|
84 |
|
85 |
return model_path
|
|
|
100 |
self.model_loaded = True
|
101 |
return True
|
102 |
except Exception as e:
|
103 |
+
logger.error(f"Error loading ECAPA-TDNN model: {e}")
|
104 |
return False
|
105 |
|
106 |
def embed_utterance(self, audio, sr=16000):
|
|
|
122 |
|
123 |
return embedding.squeeze().cpu().numpy()
|
124 |
except Exception as e:
|
125 |
+
logger.error(f"Error extracting embedding: {e}")
|
126 |
return np.zeros(self.embedding_dim)
|
127 |
|
128 |
|
|
|
141 |
|
142 |
return embedding
|
143 |
except Exception as e:
|
144 |
+
logger.error(f"Embedding extraction error: {e}")
|
145 |
return np.zeros(self.encoder.embedding_dim)
|
146 |
|
147 |
|
|
|
276 |
|
277 |
|
278 |
class WhisperTranscriber:
|
279 |
+
"""Whisper transcriber using transformers with FastRTC optimization"""
|
280 |
def __init__(self, model_name="distil-large-v3"):
|
281 |
self.model = None
|
282 |
self.processor = None
|
283 |
self.model_name = model_name
|
284 |
+
self.model_loaded = False
|
285 |
|
286 |
def load_model(self):
|
287 |
"""Load Whisper model"""
|
288 |
try:
|
289 |
from transformers import WhisperProcessor, WhisperForConditionalGeneration
|
290 |
|
291 |
+
model_id = f"distil-whisper/distil-{self.model_name}" if "distil" in self.model_name else f"openai/whisper-{self.model_name}"
|
292 |
+
|
293 |
+
self.processor = WhisperProcessor.from_pretrained(model_id)
|
294 |
+
self.model = WhisperForConditionalGeneration.from_pretrained(
|
295 |
+
model_id,
|
296 |
+
torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
|
297 |
+
low_cpu_mem_usage=True,
|
298 |
+
use_safetensors=True
|
299 |
+
)
|
300 |
+
|
301 |
+
if torch.cuda.is_available():
|
302 |
+
self.model = self.model.cuda()
|
303 |
|
304 |
+
self.model_loaded = True
|
305 |
return True
|
306 |
except Exception as e:
|
307 |
+
logger.error(f"Error loading Whisper model: {e}")
|
308 |
return False
|
309 |
|
310 |
def transcribe(self, audio_array, sample_rate=16000):
|
311 |
"""Transcribe audio array"""
|
312 |
+
if not self.model_loaded:
|
313 |
+
return ""
|
314 |
+
|
315 |
try:
|
316 |
+
# Ensure audio is the right length and format
|
317 |
+
if len(audio_array) < 1600: # Less than 0.1 seconds
|
318 |
return ""
|
319 |
|
320 |
+
# Resample if needed
|
321 |
if sample_rate != 16000:
|
322 |
+
import torchaudio.functional as F
|
323 |
+
audio_tensor = torch.tensor(audio_array, dtype=torch.float32)
|
324 |
+
audio_array = F.resample(audio_tensor, sample_rate, 16000).numpy()
|
325 |
+
|
326 |
+
# Process with Whisper
|
327 |
+
inputs = self.processor(
|
328 |
+
audio_array,
|
329 |
+
sampling_rate=16000,
|
330 |
+
return_tensors="pt",
|
331 |
+
truncation=False,
|
332 |
+
padding=True
|
333 |
+
)
|
334 |
|
335 |
+
if torch.cuda.is_available():
|
336 |
+
inputs = {k: v.cuda() for k, v in inputs.items()}
|
|
|
337 |
|
|
|
338 |
with torch.no_grad():
|
339 |
+
predicted_ids = self.model.generate(
|
340 |
+
inputs["input_features"],
|
341 |
+
max_length=448,
|
342 |
+
num_beams=1,
|
343 |
+
do_sample=False,
|
344 |
+
use_cache=True
|
345 |
+
)
|
346 |
|
347 |
+
transcription = self.processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
|
348 |
|
349 |
+
return transcription.strip()
|
350 |
except Exception as e:
|
351 |
+
logger.error(f"Transcription error: {e}")
|
352 |
return ""
|
353 |
|
354 |
|
355 |
+
class FastRTCSpeakerDiarization:
|
356 |
def __init__(self):
|
357 |
self.encoder = None
|
358 |
self.audio_processor = None
|
359 |
self.speaker_detector = None
|
360 |
self.transcriber = None
|
361 |
+
self.audio_queue = queue.Queue(maxsize=100)
|
362 |
self.processing_thread = None
|
|
|
363 |
self.full_sentences = []
|
364 |
self.sentence_speakers = []
|
|
|
|
|
365 |
self.is_running = False
|
366 |
self.change_threshold = DEFAULT_CHANGE_THRESHOLD
|
367 |
self.max_speakers = DEFAULT_MAX_SPEAKERS
|
368 |
+
self.audio_buffer = []
|
369 |
+
self.buffer_duration = 3.0 # seconds
|
370 |
+
self.last_transcription_time = time.time()
|
371 |
+
self.chunk_size = int(SAMPLE_RATE * CHUNK_DURATION_MS / 1000)
|
372 |
|
373 |
def initialize_models(self):
|
374 |
"""Initialize the speaker encoder and transcription models"""
|
375 |
try:
|
376 |
device_str = "cuda" if torch.cuda.is_available() else "cpu"
|
377 |
+
logger.info(f"Using device: {device_str}")
|
378 |
|
379 |
# Initialize speaker encoder
|
380 |
self.encoder = SpeechBrainEncoder(device=device_str)
|
|
|
391 |
change_threshold=self.change_threshold,
|
392 |
max_speakers=self.max_speakers
|
393 |
)
|
394 |
+
logger.info("Models loaded successfully!")
|
395 |
return True
|
396 |
else:
|
397 |
+
logger.error("Failed to load models")
|
398 |
return False
|
399 |
except Exception as e:
|
400 |
+
logger.error(f"Model initialization error: {e}")
|
401 |
return False
|
402 |
|
403 |
+
def process_audio_chunk(self, audio_chunk: np.ndarray, sample_rate: int):
|
404 |
+
"""Process individual audio chunk from FastRTC"""
|
405 |
+
if not self.is_running or audio_chunk is None:
|
406 |
return
|
407 |
|
408 |
try:
|
409 |
+
# Ensure audio chunk is in correct format
|
410 |
+
if isinstance(audio_chunk, np.ndarray):
|
411 |
+
# Ensure mono audio
|
412 |
+
if len(audio_chunk.shape) > 1:
|
413 |
+
audio_chunk = audio_chunk.mean(axis=1)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
414 |
|
415 |
+
# Normalize audio
|
416 |
+
if audio_chunk.dtype != np.float32:
|
417 |
+
audio_chunk = audio_chunk.astype(np.float32)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
418 |
|
419 |
+
if np.abs(audio_chunk).max() > 1.0:
|
420 |
+
audio_chunk = audio_chunk / np.abs(audio_chunk).max()
|
421 |
|
422 |
+
# Add to buffer
|
423 |
+
self.audio_buffer.extend(audio_chunk)
|
424 |
+
|
425 |
+
# Keep buffer to specified duration
|
426 |
+
max_buffer_length = int(self.buffer_duration * sample_rate)
|
427 |
+
if len(self.audio_buffer) > max_buffer_length:
|
428 |
+
self.audio_buffer = self.audio_buffer[-max_buffer_length:]
|
429 |
+
|
430 |
+
# Process if enough audio accumulated and enough time passed
|
431 |
+
current_time = time.time()
|
432 |
+
if (current_time - self.last_transcription_time > 1.5 and
|
433 |
+
len(self.audio_buffer) > sample_rate * 0.8): # At least 0.8 seconds
|
434 |
+
|
435 |
+
if not self.audio_queue.full():
|
436 |
+
self.audio_queue.put((np.array(self.audio_buffer[-int(sample_rate * 2):]), sample_rate))
|
437 |
+
self.last_transcription_time = current_time
|
438 |
+
|
439 |
except Exception as e:
|
440 |
+
logger.error(f"Audio chunk processing error: {e}")
|
441 |
|
442 |
+
def process_audio_queue(self):
|
443 |
+
"""Process audio from the queue"""
|
444 |
while self.is_running:
|
445 |
try:
|
446 |
+
audio_data, sample_rate = self.audio_queue.get(timeout=1)
|
|
|
|
|
|
|
447 |
|
448 |
+
if len(audio_data) < 1600: # Skip very short audio
|
449 |
+
continue
|
|
|
450 |
|
451 |
+
# Transcribe audio
|
452 |
+
transcription = self.transcriber.transcribe(audio_data, sample_rate)
|
|
|
453 |
|
454 |
+
if transcription and len(transcription.strip()) > 0:
|
455 |
+
# Extract speaker embedding
|
456 |
+
speaker_embedding = self.audio_processor.extract_embedding(audio_data)
|
457 |
+
|
458 |
+
# Detect speaker
|
459 |
+
speaker_id, similarity = self.speaker_detector.add_embedding(speaker_embedding)
|
460 |
+
|
461 |
+
# Store results
|
462 |
+
self.full_sentences.append(transcription.strip())
|
463 |
+
self.sentence_speakers.append(speaker_id)
|
464 |
+
|
465 |
+
logger.info(f"Processed: Speaker {speaker_id + 1}: {transcription.strip()[:50]}...")
|
466 |
+
|
467 |
except queue.Empty:
|
468 |
continue
|
469 |
except Exception as e:
|
470 |
+
logger.error(f"Error processing audio queue: {e}")
|
471 |
|
472 |
def start_recording(self):
|
473 |
+
"""Start the recording and processing"""
|
474 |
+
if self.encoder is None or self.transcriber is None:
|
475 |
return "Please initialize models first!"
|
476 |
|
477 |
try:
|
|
|
478 |
self.is_running = True
|
479 |
+
self.audio_buffer = []
|
480 |
+
self.last_transcription_time = time.time()
|
481 |
+
|
482 |
+
# Clear the queue
|
483 |
+
while not self.audio_queue.empty():
|
484 |
+
try:
|
485 |
+
self.audio_queue.get_nowait()
|
486 |
+
except queue.Empty:
|
487 |
+
break
|
488 |
+
|
489 |
+
# Start processing thread
|
490 |
+
self.processing_thread = threading.Thread(target=self.process_audio_queue, daemon=True)
|
491 |
self.processing_thread.start()
|
492 |
|
493 |
+
logger.info("Recording started successfully!")
|
494 |
+
return "Recording started successfully!"
|
495 |
|
496 |
except Exception as e:
|
497 |
+
logger.error(f"Error starting recording: {e}")
|
498 |
return f"Error starting recording: {e}"
|
499 |
|
500 |
def stop_recording(self):
|
501 |
"""Stop the recording process"""
|
502 |
self.is_running = False
|
503 |
+
logger.info("Recording stopped!")
|
504 |
return "Recording stopped!"
|
505 |
|
506 |
def clear_conversation(self):
|
507 |
"""Clear all conversation data"""
|
508 |
self.full_sentences = []
|
509 |
self.sentence_speakers = []
|
|
|
|
|
510 |
self.audio_buffer = []
|
511 |
|
512 |
+
# Clear the queue
|
513 |
+
while not self.audio_queue.empty():
|
514 |
+
try:
|
515 |
+
self.audio_queue.get_nowait()
|
516 |
+
except queue.Empty:
|
517 |
+
break
|
518 |
+
|
519 |
if self.speaker_detector:
|
520 |
self.speaker_detector = SpeakerChangeDetector(
|
521 |
embedding_dim=self.encoder.embedding_dim,
|
|
|
539 |
def get_formatted_conversation(self):
|
540 |
"""Get the formatted conversation with speaker colors"""
|
541 |
try:
|
542 |
+
if not self.full_sentences:
|
543 |
+
return "Waiting for speech input... ๐ค"
|
544 |
+
|
545 |
sentences_with_style = []
|
546 |
|
547 |
+
for i, sentence in enumerate(self.full_sentences[-10:]): # Show last 10 sentences
|
|
|
|
|
548 |
if i >= len(self.sentence_speakers):
|
549 |
color = "#FFFFFF"
|
550 |
+
speaker_name = "Unknown"
|
551 |
else:
|
552 |
+
speaker_id = self.sentence_speakers[-(10-i) if len(self.sentence_speakers) >= 10 else i]
|
553 |
color = self.speaker_detector.get_color_for_speaker(speaker_id)
|
554 |
speaker_name = f"Speaker {speaker_id + 1}"
|
555 |
+
|
556 |
sentences_with_style.append(
|
557 |
+
f'<p><span style="color:{color}; font-weight: bold;">{speaker_name}:</span> {sentence}</p>')
|
558 |
|
559 |
+
return "".join(sentences_with_style)
|
|
|
|
|
|
|
560 |
|
561 |
except Exception as e:
|
562 |
return f"Error formatting conversation: {e}"
|
|
|
568 |
|
569 |
try:
|
570 |
status = self.speaker_detector.get_status_info()
|
571 |
+
queue_size = self.audio_queue.qsize()
|
572 |
|
573 |
status_lines = [
|
574 |
f"**Current Speaker:** {status['current_speaker'] + 1}",
|
|
|
576 |
f"**Last Similarity:** {status['last_similarity']:.3f}",
|
577 |
f"**Change Threshold:** {status['threshold']:.2f}",
|
578 |
f"**Total Sentences:** {len(self.full_sentences)}",
|
579 |
+
f"**Buffer Length:** {len(self.audio_buffer)} samples",
|
580 |
+
f"**Queue Size:** {queue_size}",
|
581 |
"",
|
582 |
"**Speaker Segment Counts:**"
|
583 |
]
|
|
|
593 |
|
594 |
|
595 |
# Global instance
|
596 |
+
diarization_system = FastRTCSpeakerDiarization()
|
597 |
|
598 |
|
599 |
def initialize_system():
|
|
|
635 |
return diarization_system.get_status_info()
|
636 |
|
637 |
|
638 |
+
def process_audio_stream(audio_stream):
|
639 |
+
"""Process streaming audio from FastRTC"""
|
640 |
+
if audio_stream is not None and diarization_system.is_running:
|
641 |
+
sample_rate, audio_data = audio_stream
|
642 |
+
diarization_system.process_audio_chunk(audio_data, sample_rate)
|
643 |
+
|
644 |
return get_conversation(), get_status()
|
645 |
|
646 |
|
647 |
+
# Create Gradio interface with FastRTC
|
648 |
def create_interface():
|
649 |
+
with gr.Blocks(title="FastRTC Real-time Speaker Diarization", theme=gr.themes.Soft()) as app:
|
650 |
+
gr.Markdown("# ๐ค FastRTC Real-time Speech Recognition with Speaker Diarization")
|
651 |
+
gr.Markdown("This app uses Hugging Face FastRTC for real-time audio streaming with automatic speaker identification and color-coding.")
|
652 |
|
653 |
with gr.Row():
|
654 |
with gr.Column(scale=2):
|
655 |
+
# FastRTC Audio input for real-time streaming
|
656 |
audio_input = gr.Audio(
|
657 |
+
sources=["microphone"],
|
658 |
type="numpy",
|
659 |
streaming=True,
|
660 |
+
label="๐๏ธ FastRTC Microphone Input",
|
661 |
+
format="wav",
|
662 |
+
show_download_button=False,
|
663 |
+
container=True,
|
664 |
+
elem_id="fastrtc_audio"
|
665 |
)
|
666 |
|
667 |
# Main conversation display
|
668 |
conversation_output = gr.HTML(
|
669 |
+
value="<i>Click 'Initialize System' and then 'Start Recording' to begin...</i>",
|
670 |
+
label="Live Conversation",
|
671 |
+
elem_id="conversation_display"
|
672 |
)
|
673 |
|
674 |
# Control buttons
|
675 |
with gr.Row():
|
676 |
+
init_btn = gr.Button("๐ง Initialize System", variant="secondary", size="lg")
|
677 |
+
start_btn = gr.Button("๐๏ธ Start Recording", variant="primary", interactive=False, size="lg")
|
678 |
+
stop_btn = gr.Button("โน๏ธ Stop Recording", variant="stop", interactive=False, size="lg")
|
679 |
+
clear_btn = gr.Button("๐๏ธ Clear", interactive=False, size="lg")
|
680 |
|
681 |
# Status display
|
682 |
status_output = gr.Textbox(
|
683 |
label="System Status",
|
684 |
value="System not initialized",
|
685 |
lines=10,
|
686 |
+
interactive=False,
|
687 |
+
show_copy_button=True
|
688 |
)
|
689 |
|
690 |
with gr.Column(scale=1):
|
|
|
697 |
step=0.05,
|
698 |
value=DEFAULT_CHANGE_THRESHOLD,
|
699 |
label="Speaker Change Sensitivity",
|
700 |
+
info="Lower = more sensitive to changes"
|
701 |
)
|
702 |
|
703 |
max_speakers_slider = gr.Slider(
|
|
|
708 |
label="Maximum Number of Speakers"
|
709 |
)
|
710 |
|
711 |
+
update_settings_btn = gr.Button("Update Settings", variant="secondary")
|
712 |
|
713 |
# Speaker color legend
|
714 |
gr.Markdown("## ๐จ Speaker Colors")
|
715 |
color_info = []
|
716 |
for i, (color, name) in enumerate(zip(SPEAKER_COLORS, SPEAKER_COLOR_NAMES)):
|
717 |
+
color_info.append(f'<span style="color:{color}; font-size: 16px;">โ</span> Speaker {i+1} ({name})')
|
718 |
|
719 |
gr.HTML("<br>".join(color_info[:DEFAULT_MAX_SPEAKERS]))
|
720 |
|
721 |
+
# Performance info
|
722 |
+
gr.Markdown("## ๐ Performance")
|
723 |
gr.Markdown("""
|
724 |
+
- **FastRTC**: Low-latency audio streaming
|
725 |
+
- **Whisper**: distil-large-v3 for transcription
|
726 |
+
- **ECAPA-TDNN**: Speaker embeddings
|
727 |
+
- **Real-time**: ~100ms processing chunks
|
|
|
|
|
|
|
|
|
728 |
""")
|
729 |
|
730 |
# Event handlers
|
|
|
732 |
result = initialize_system()
|
733 |
if "successfully" in result:
|
734 |
return (
|
735 |
+
result, # status_output
|
736 |
gr.update(interactive=True), # start_btn
|
737 |
gr.update(interactive=True), # clear_btn
|
738 |
+
get_conversation(), # conversation_output
|
739 |
+
get_status() # status_output update
|
740 |
)
|
741 |
else:
|
742 |
return (
|
743 |
+
result, # status_output
|
744 |
gr.update(interactive=False), # start_btn
|
745 |
gr.update(interactive=False), # clear_btn
|
746 |
+
get_conversation(), # conversation_output
|
747 |
+
get_status() # status_output update
|
748 |
)
|
749 |
|
750 |
def on_start():
|
751 |
result = start_recording()
|
752 |
return (
|
753 |
+
result, # status_output
|
754 |
gr.update(interactive=False), # start_btn
|
755 |
gr.update(interactive=True), # stop_btn
|
756 |
)
|
|
|
758 |
def on_stop():
|
759 |
result = stop_recording()
|
760 |
return (
|
761 |
+
result, # status_output
|
762 |
gr.update(interactive=True), # start_btn
|
763 |
gr.update(interactive=False), # stop_btn
|
764 |
)
|
765 |
|
766 |
+
# Auto-refresh function
|
767 |
+
def refresh_display():
|
768 |
+
return get_conversation(), get_status()
|
769 |
+
|
770 |
# Connect event handlers
|
771 |
init_btn.click(
|
772 |
on_initialize,
|
|
|
794 |
outputs=[status_output]
|
795 |
)
|
796 |
|
797 |
+
# FastRTC streaming audio processing
|
798 |
audio_input.stream(
|
799 |
+
process_audio_stream,
|
800 |
inputs=[audio_input],
|
801 |
outputs=[conversation_output, status_output],
|
802 |
+
stream_every=0.1, # Process every 100ms
|
803 |
+
time_limit=None
|
804 |
)
|
805 |
|
806 |
+
# Auto-refresh timer
|
807 |
+
refresh_timer = gr.Timer(2.0)
|
808 |
refresh_timer.tick(
|
809 |
+
refresh_display,
|
810 |
outputs=[conversation_output, status_output]
|
811 |
)
|
812 |
|