tested

BasedHardware · Aug 17, 2024 · 91c8d5d · 91c8d5d
1 parent c758076
commit 91c8d5d
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 8 deletions.
diff --git a/backend/routers/transcribe.py b/backend/routers/transcribe.py
@@ -12,7 +12,7 @@
 import opuslib
 
 from utils.redis_utils import get_user_speech_profile, get_user_speech_profile_duration
-from utils.stt.deepgram_util import process_audio_dg, send_initial_file2, transcribe_file_deepgram, convert_audio_bytes_to_resampled_bytes
+from utils.stt.deepgram_util import process_audio_dg, send_initial_file2, transcribe_file_deepgram
 from utils.stt.vad import VADIterator, model, get_speech_state, SpeechState, vad_is_empty, is_speech_present
 
 router = APIRouter()
@@ -54,7 +54,7 @@ async def _websocket_util(
     websocket_active = True
     duration = 0
     is_speech_active = False
-    speech_timeout = 0.7
+    speech_timeout = 0.7 # Good for now but better dynamically adjust it by user behaviour
     last_speech_time = 0
     try:
         if language == 'en' and codec == 'opus' and include_speech_profile:
@@ -77,7 +77,7 @@ async def _websocket_util(
         await websocket.close()
         return
 
-    threshold = 0.7
+    threshold = 0.6
     vad_iterator = VADIterator(model, sampling_rate=sample_rate, threshold=threshold) 
     window_size_samples = 256 if sample_rate == 8000 else 512
     if codec == 'opus':

diff --git a/backend/utils/stt/vad.py b/backend/utils/stt/vad.py
@@ -19,11 +19,6 @@
 (get_speech_timestamps, save_audio, read_audio, VADIterator, collect_chunks) = utils
 
 
-class SpeechState(str, Enum):
-    has_speech = 'has_speech'
-    no_speech = 'no_speech'
-
-
 def get_speech_state(data, vad_iterator, window_size_samples=256):
     has_start, has_end = False, False
     for i in range(0, len(data), window_size_samples):