RealTimeSTT 0.3.93__tar.gz → 0.3.94__tar.gz

{realtimestt-0.3.93 → realtimestt-0.3.94}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: RealtimeSTT
-Version: 0.3.93
+Version: 0.3.94
 Summary: A fast Voice Activity Detection and Transcription System
 Home-page: https://github.com/KoljaB/RealTimeSTT
 Author: Kolja Beigel
@@ -13,15 +13,15 @@ Requires-Python: >=3.6
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: PyAudio==0.2.14
-Requires-Dist: faster-whisper==1.1.0
+Requires-Dist: faster-whisper==1.1.1
 Requires-Dist: pvporcupine==1.9.5
 Requires-Dist: webrtcvad-wheels==2.0.14
 Requires-Dist: halo==0.0.31
 Requires-Dist: torch
 Requires-Dist: torchaudio
-Requires-Dist: scipy==1.14.1
+Requires-Dist: scipy==1.15.1
 Requires-Dist: openwakeword>=0.4.0
-Requires-Dist: websockets==14.1
+Requires-Dist: websockets==14.2
 Requires-Dist: websocket-client==1.8.0
 Requires-Dist: numpy<2.0.0
 

{realtimestt-0.3.93 → realtimestt-0.3.94}/RealtimeSTT/audio_input.py

@@ -1,22 +1,35 @@
 from colorama import init, Fore, Style
+from scipy.signal import butter, filtfilt, resample_poly
 import pyaudio
 import logging
-import time
 
+DESIRED_RATE = 16000
+CHUNK_SIZE = 1024
+AUDIO_FORMAT = pyaudio.paInt16
+CHANNELS = 1
 
 class AudioInput:
-    def __init__(self, input_device_index=None, debug_mode=False):
+    def __init__(
+            self,
+            input_device_index: int = None,
+            debug_mode: bool = False,
+            target_samplerate: int = DESIRED_RATE,
+            chunk_size: int = CHUNK_SIZE,
+            audio_format: int = AUDIO_FORMAT,
+            channels: int = CHANNELS,
+            resample_to_target: bool = True,
+        ):
+
         self.input_device_index = input_device_index
         self.debug_mode = debug_mode
         self.audio_interface = None
         self.stream = None
         self.device_sample_rate = None
-
-        # PyAudio constants
-        self.CHUNK = 1024
-        self.FORMAT = pyaudio.paInt16  
-        self.CHANNELS = 1
-        self.DESIRED_RATE = 16000
+        self.target_samplerate = target_samplerate
+        self.chunk_size = chunk_size
+        self.audio_format = audio_format
+        self.channels = channels
+        self.resample_to_target = resample_to_target
 
     def get_supported_sample_rates(self, device_index):
         """Test which standard sample rates are supported by the specified device."""
@@ -31,8 +44,8 @@ class AudioInput:
                 if self.audio_interface.is_format_supported(
                     rate,
                     input_device=device_index,  # Changed to input_device
-                    input_channels=max_channels,  # Changed to input_channels 
-                    input_format=self.FORMAT,  # Changed to input_format
+                    input_channels=max_channels,  # Changed to input_channels
+                    input_format=self.audio_format,  # Changed to input_format
                 ):
                     supported_rates.append(rate)
             except:
@@ -108,18 +121,18 @@ class AudioInput:
             if self.debug_mode:
                 print(f"Actual selected device index: {actual_device_index}")
             self.input_device_index = actual_device_index
-            self.device_sample_rate = self._get_best_sample_rate(actual_device_index, self.DESIRED_RATE)
+            self.device_sample_rate = self._get_best_sample_rate(actual_device_index, self.target_samplerate)
 
             if self.debug_mode:
                 print(f"Setting up audio on device {self.input_device_index} with sample rate {self.device_sample_rate}")
 
             try:
                 self.stream = self.audio_interface.open(
-                    format=self.FORMAT,
-                    channels=self.CHANNELS,
+                    format=self.audio_format,
+                    channels=self.channels,
                     rate=self.device_sample_rate,
                     input=True,
-                    frames_per_buffer=self.CHUNK,
+                    frames_per_buffer=self.chunk_size,
                     input_device_index=self.input_device_index,
                 )
                 if self.debug_mode:
@@ -135,6 +148,60 @@ class AudioInput:
                 self.audio_interface.terminate()
             return False
 
+    def lowpass_filter(self, signal, cutoff_freq, sample_rate):
+        """
+        Apply a low-pass Butterworth filter to prevent aliasing in the signal.
+
+        Args:
+            signal (np.ndarray): Input audio signal to filter
+            cutoff_freq (float): Cutoff frequency in Hz
+            sample_rate (float): Sampling rate of the input signal in Hz
+
+        Returns:
+            np.ndarray: Filtered audio signal
+
+        Notes:
+            - Uses a 5th order Butterworth filter
+            - Applies zero-phase filtering using filtfilt
+        """
+        # Calculate the Nyquist frequency (half the sample rate)
+        nyquist_rate = sample_rate / 2.0
+
+        # Normalize cutoff frequency to Nyquist rate (required by butter())
+        normal_cutoff = cutoff_freq / nyquist_rate
+
+        # Design the Butterworth filter
+        b, a = butter(5, normal_cutoff, btype='low', analog=False)
+
+        # Apply zero-phase filtering (forward and backward)
+        filtered_signal = filtfilt(b, a, signal)
+        return filtered_signal
+
+    def resample_audio(self, pcm_data, target_sample_rate, original_sample_rate):
+        """
+        Filter and resample audio data to a target sample rate.
+
+        Args:
+            pcm_data (np.ndarray): Input audio data
+            target_sample_rate (int): Desired output sample rate in Hz
+            original_sample_rate (int): Original sample rate of input in Hz
+
+        Returns:
+            np.ndarray: Resampled audio data
+
+        Notes:
+            - Applies anti-aliasing filter before resampling
+            - Uses polyphase filtering for high-quality resampling
+        """
+        if target_sample_rate < original_sample_rate:
+            # Downsampling with low-pass filter
+            pcm_filtered = self.lowpass_filter(pcm_data, target_sample_rate / 2, original_sample_rate)
+            resampled = resample_poly(pcm_filtered, target_sample_rate, original_sample_rate)
+        else:
+            # Upsampling without low-pass filter
+            resampled = resample_poly(pcm_data, target_sample_rate, original_sample_rate)
+        return resampled
+
     def read_chunk(self):
         """Read a chunk of audio data"""
         return self.stream.read(self.CHUNK, exception_on_overflow=False)

{realtimestt-0.3.93 → realtimestt-0.3.94}/RealtimeSTT/audio_recorder.py

@@ -120,7 +120,7 @@ class TranscriptionWorker:
                     data = self.conn.recv()
                     self.queue.put(data)
                 except Exception as e:
-                    logging.error(f"Error receiving data from connection: {e}")
+                    logging.error(f"Error receiving data from connection: {e}", exc_info=True)
             else:
                 time.sleep(TIME_SLEEP)
 
@@ -162,6 +162,7 @@ class TranscriptionWorker:
                 try:
                     audio, language = self.queue.get(timeout=0.1)
                     try:
+                        logging.debug(f"Transcribing audio with language {language}")
                         if self.batch_size > 0:
                             segments, info = model.transcribe(
                                 audio,
@@ -184,7 +185,7 @@ class TranscriptionWorker:
                         logging.debug(f"Final text detected with main model: {transcription}")
                         self.conn.send(('success', (transcription, info)))
                     except Exception as e:
-                        logging.error(f"General error in transcription: {e}")
+                        logging.error(f"General error in transcription: {e}", exc_info=True)
                         self.conn.send(('error', str(e)))
                 except queue.Empty:
                     continue
@@ -193,7 +194,7 @@ class TranscriptionWorker:
                     logging.debug("Transcription worker process finished due to KeyboardInterrupt")
                     break
                 except Exception as e:
-                    logging.error(f"General error in processing queue item: {e}")
+                    logging.error(f"General error in processing queue item: {e}", exc_info=True)
         finally:
             __builtins__['print'] = print  # Restore the original print function
             self.conn.close()
@@ -567,6 +568,8 @@ class AudioToTextRecorder:
         self.sample_rate = sample_rate
         self.recording_start_time = 0
         self.recording_stop_time = 0
+        self.last_recording_start_time = 0
+        self.last_recording_stop_time = 0
         self.wake_word_detect_time = 0
         self.silero_check_time = 0
         self.silero_working = False
@@ -590,6 +593,8 @@ class AudioToTextRecorder:
         self.stream = None
         self.start_recording_event = threading.Event()
         self.stop_recording_event = threading.Event()
+        self.backdate_stop_seconds = 0.0
+        self.backdate_resume_seconds = 0.0
         self.last_transcription_bytes = None
         self.last_transcription_bytes_b64 = None
         self.initial_prompt = initial_prompt
@@ -860,6 +865,7 @@ class AudioToTextRecorder:
                        0.3)
         )
         self.frames = []
+        self.last_frames = []
 
         # Recording control flags
         self.is_recording = False
@@ -925,7 +931,7 @@ class AudioToTextRecorder:
                 logging.info("KeyboardInterrupt in read from stdout detected, exiting...")
                 break
             except Exception as e:
-                logging.error(f"Unexpected error in read from stdout: {e}")
+                logging.error(f"Unexpected error in read from stdout: {e}", exc_info=True)
                 logging.error(traceback.format_exc())  # Log the full traceback here
                 break 
             time.sleep(0.1)
@@ -1074,7 +1080,7 @@ class AudioToTextRecorder:
                     return stream
 
                 except Exception as e:
-                    logging.error(f"Microphone connection failed: {e}. Retrying...")
+                    logging.error(f"Microphone connection failed: {e}. Retrying...", exc_info=True)
                     input_device_index = None
                     time.sleep(3)  # Wait before retrying
                     continue
@@ -1190,7 +1196,7 @@ class AudioToTextRecorder:
                     if e.errno == pyaudio.paInputOverflowed:
                         logging.warning("Input overflowed. Frame dropped.")
                     else:
-                        logging.error(f"OSError during recording: {e}")
+                        logging.error(f"OSError during recording: {e}", exc_info=True)
                         # Attempt to reinitialize the stream
                         logging.error("Attempting to reinitialize the audio stream...")
 
@@ -1308,13 +1314,60 @@ class AudioToTextRecorder:
                     if (self.stop_recording_event.wait(timeout=0.02)):
                         break
 
-            # Convert recorded frames to the appropriate audio format.
-            audio_array = np.frombuffer(b''.join(self.frames), dtype=np.int16)
-            self.audio = audio_array.astype(np.float32) / INT16_MAX_ABS_VALUE
+            frames = self.frames
+            if len(frames) == 0:
+                frames = self.last_frames
+
+            # Calculate samples needed for backdating resume
+            samples_to_keep = int(self.sample_rate * self.backdate_resume_seconds)
+
+            # First convert all current frames to audio array
+            full_audio_array = np.frombuffer(b''.join(frames), dtype=np.int16)
+            full_audio = full_audio_array.astype(np.float32) / INT16_MAX_ABS_VALUE
+
+            # Calculate how many samples we need to keep for backdating resume
+            if samples_to_keep > 0:
+                samples_to_keep = min(samples_to_keep, len(full_audio))
+                # Keep the last N samples for backdating resume
+                frames_to_read_audio = full_audio[-samples_to_keep:]
+
+                # Convert the audio back to int16 bytes for frames
+                frames_to_read_int16 = (frames_to_read_audio * INT16_MAX_ABS_VALUE).astype(np.int16)
+                frame_bytes = frames_to_read_int16.tobytes()
+
+                # Split into appropriate frame sizes (assuming standard frame size)
+                FRAME_SIZE = 2048  # Typical frame size
+                frames_to_read = []
+                for i in range(0, len(frame_bytes), FRAME_SIZE):
+                    frame = frame_bytes[i:i + FRAME_SIZE]
+                    if frame:  # Only add non-empty frames
+                        frames_to_read.append(frame)
+            else:
+                frames_to_read = []
+
+            # Process backdate stop seconds
+            samples_to_remove = int(self.sample_rate * self.backdate_stop_seconds)
+
+            if samples_to_remove > 0:
+                if samples_to_remove < len(full_audio):
+                    self.audio = full_audio[:-samples_to_remove]
+                    logging.debug(f"Removed {samples_to_remove} samples "
+                        f"({samples_to_remove/self.sample_rate:.3f}s) from end of audio")
+                else:
+                    self.audio = np.array([], dtype=np.float32)
+                    logging.debug("Cleared audio (samples_to_remove >= audio length)")
+            else:
+                self.audio = full_audio
+                logging.debug(f"No samples removed, final audio length: {len(self.audio)}")
+
             self.frames.clear()
+            self.last_frames.clear()
+            self.frames.extend(frames_to_read)
+
+            # Reset backdating parameters
+            self.backdate_stop_seconds = 0.0
+            self.backdate_resume_seconds = 0.0
 
-            # Reset recording-related timestamps
-            self.recording_stop_time = 0
             self.listen_start = 0
 
             self._set_state("inactive")
@@ -1388,7 +1441,7 @@ class AudioToTextRecorder:
                     logging.error(f"Transcription error: {result}")
                     raise Exception(result)
             except Exception as e:
-                logging.error(f"Error during transcription: {str(e)}")
+                logging.error(f"Error during transcription: {str(e)}", exc_info=True)
                 raise e
 
     def _process_wakeword(self, data):
@@ -1476,7 +1529,16 @@ class AudioToTextRecorder:
             return self.transcribe()
 
 
-    def start(self):
+    def format_number(self, num):
+        # Convert the number to a string
+        num_str = f"{num:.10f}"  # Ensure precision is sufficient
+        # Split the number into integer and decimal parts
+        integer_part, decimal_part = num_str.split('.')
+        # Take the last two digits of the integer part and the first two digits of the decimal part
+        result = f"{integer_part[-2:]}.{decimal_part[:2]}"
+        return result
+
+    def start(self, frames = None):
         """
         Starts recording audio directly without waiting for voice activity.
         """
@@ -1498,7 +1560,10 @@ class AudioToTextRecorder:
         self.wakeword_detected = False
         self.wake_word_detect_time = 0
         self.frames = []
+        if frames:
+            self.frames = frames
         self.is_recording = True
+
         self.recording_start_time = time.time()
         self.is_silero_speech_active = False
         self.is_webrtc_speech_active = False
@@ -1510,9 +1575,19 @@ class AudioToTextRecorder:
 
         return self
 
-    def stop(self):
+    def stop(self,
+             backdate_stop_seconds: float = 0.0,
+             backdate_resume_seconds: float = 0.0,
+        ):
         """
         Stops recording audio.
+
+        Args:
+        - backdate_stop_seconds (float, default="0.0"): Specifies the number of
+            seconds to backdate the stop time. This is useful when the stop
+            command is issued after the actual stop time.
+        - backdate_resume_seconds (float, default="0.0"): Specifies the number
+            of seconds to backdate the time relistening is initiated.
         """
 
         # Ensure there's a minimum interval
@@ -1525,6 +1600,9 @@ class AudioToTextRecorder:
             return self
 
         logging.info("recording stopped")
+        self.last_frames = copy.deepcopy(self.frames)
+        self.backdate_stop_seconds = backdate_stop_seconds
+        self.backdate_resume_seconds = backdate_resume_seconds
         self.is_recording = False
         self.recording_stop_time = time.time()
         self.is_silero_speech_active = False
@@ -1533,6 +1611,9 @@ class AudioToTextRecorder:
         self.start_recording_event.clear()
         self.stop_recording_event.set()
 
+        self.last_recording_start_time = self.recording_start_time
+        self.last_recording_stop_time = self.recording_stop_time
+
         if self.on_recording_stop:
             self.on_recording_stop()
 
@@ -1607,7 +1688,6 @@ class AudioToTextRecorder:
                 return
 
             print("\033[91mRealtimeSTT shutting down\033[0m")
-            # logging.debug("RealtimeSTT shutting down")
 
             # Force wait_audio() and text() to exit
             self.is_shut_down = True
@@ -1736,7 +1816,7 @@ class AudioToTextRecorder:
                             data = self.audio_queue.get()
 
                 except BrokenPipeError:
-                    logging.error("BrokenPipeError _recording_worker")
+                    logging.error("BrokenPipeError _recording_worker", exc_info=True)
                     self.is_running = False
                     break
 
@@ -1813,11 +1893,11 @@ class AudioToTextRecorder:
 
                         except struct.error:
                             logging.error("Error unpacking audio data "
-                                        "for wake word processing.")
+                                        "for wake word processing.", exc_info=True)
                             continue
 
                         except Exception as e:
-                            logging.error(f"Wake word processing error: {e}")
+                            logging.error(f"Wake word processing error: {e}", exc_info=True)
                             continue
 
                         if self.use_extended_logging:
@@ -2044,7 +2124,7 @@ class AudioToTextRecorder:
         except Exception as e:
             logging.debug('Debug: Caught exception in main try block')
             if not self.interrupt_stop_event.is_set():
-                logging.error(f"Unhandled exeption in _recording_worker: {e}")
+                logging.error(f"Unhandled exeption in _recording_worker: {e}", exc_info=True)
                 raise
 
         if self.use_extended_logging:
@@ -2111,7 +2191,7 @@ class AudioToTextRecorder:
                                     logging.warning("Realtime transcription timed out")
                                     continue
                             except Exception as e:
-                                logging.error(f"Error in realtime transcription: {str(e)}")
+                                logging.error(f"Error in realtime transcription: {str(e)}", exc_info=True)
                                 continue
                     else:
                         # Perform transcription and assemble the text
@@ -2145,7 +2225,6 @@ class AudioToTextRecorder:
                     if self.is_recording and time.time() - \
                             self.recording_start_time > self.init_realtime_after_seconds:
 
-                        # logging.debug('Starting realtime transcription')
                         self.realtime_transcription_text = realtime_text
                         self.realtime_transcription_text = \
                             self.realtime_transcription_text.strip()
@@ -2225,7 +2304,7 @@ class AudioToTextRecorder:
                     time.sleep(TIME_SLEEP)
 
         except Exception as e:
-            logging.error(f"Unhandled exeption in _realtime_worker: {e}")
+            logging.error(f"Unhandled exeption in _realtime_worker: {e}", exc_info=True)
             raise
 
     def _is_silero_speech(self, chunk):

{realtimestt-0.3.93 → realtimestt-0.3.94}/RealtimeSTT.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: RealtimeSTT
-Version: 0.3.93
+Version: 0.3.94
 Summary: A fast Voice Activity Detection and Transcription System
 Home-page: https://github.com/KoljaB/RealTimeSTT
 Author: Kolja Beigel
@@ -13,15 +13,15 @@ Requires-Python: >=3.6
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: PyAudio==0.2.14
-Requires-Dist: faster-whisper==1.1.0
+Requires-Dist: faster-whisper==1.1.1
 Requires-Dist: pvporcupine==1.9.5
 Requires-Dist: webrtcvad-wheels==2.0.14
 Requires-Dist: halo==0.0.31
 Requires-Dist: torch
 Requires-Dist: torchaudio
-Requires-Dist: scipy==1.14.1
+Requires-Dist: scipy==1.15.1
 Requires-Dist: openwakeword>=0.4.0
-Requires-Dist: websockets==14.1
+Requires-Dist: websockets==14.2
 Requires-Dist: websocket-client==1.8.0
 Requires-Dist: numpy<2.0.0
 

{realtimestt-0.3.93 → realtimestt-0.3.94}/RealtimeSTT.egg-info/requires.txt

@@ -1,12 +1,12 @@
 PyAudio==0.2.14
-faster-whisper==1.1.0
+faster-whisper==1.1.1
 pvporcupine==1.9.5
 webrtcvad-wheels==2.0.14
 halo==0.0.31
 torch
 torchaudio
-scipy==1.14.1
+scipy==1.15.1
 openwakeword>=0.4.0
-websockets==14.1
+websockets==14.2
 websocket-client==1.8.0
 numpy<2.0.0

{realtimestt-0.3.93 → realtimestt-0.3.94}/setup.py

@@ -9,7 +9,7 @@ with open('requirements.txt') as f:
 
 setuptools.setup(
     name="RealtimeSTT",
-    version="0.3.93",
+    version="0.3.94",
     author="Kolja Beigel",
     author_email="kolja.beigel@web.de",
     description="A fast Voice Activity Detection and Transcription System",