github/max_headroom
1
0
Fork 0
mirror of https://github.com/maglore9900/max_headroom.git synced 2025-06-06 03:25:34 +00:00
Code Issues Packages Projects Releases Wiki Activity

fixed whisper

Browse Source
This commit is contained in:
maglore9900 2024-10-04 15:32:03 -04:00
parent c7e632c0cf
commit a213f30084
1 changed files with 7 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
modules/speak.py
View File

@ -77,8 +77,9 @@ class Speak:
def transcribe(self, audio_data): def transcribe(self, audio_data):
"""Transcribe the audio data using the selected model.""" """Transcribe the audio data using the selected model."""
if self.model_name == "whisper": if self.model_name == "whisper":
# Whisper expects float32 data # Whisper expects float32 audio data
energy_threshold = 0.001 energy_threshold = 0.0001
# Convert int16 PCM audio data to float32
audio_np = np.frombuffer(audio_data, np.int16).astype(np.float32) / 32768.0 audio_np = np.frombuffer(audio_data, np.int16).astype(np.float32) / 32768.0
# Calculate energy of the audio to determine if it should be transcribed # Calculate energy of the audio to determine if it should be transcribed
@ -86,17 +87,17 @@ class Speak:
# Only transcribe if energy exceeds the threshold # Only transcribe if energy exceeds the threshold
if energy > energy_threshold: if energy > energy_threshold:
# Transcribe using Whisper model (assumed to be already loaded in self.whisper_model)
segments, _ = self.whisper_model.transcribe(audio_np, beam_size=5) segments, _ = self.whisper_model.transcribe(audio_np, beam_size=5)
transcription = " ".join([segment.text for segment in segments]) transcription = " ".join([segment.text for segment in segments])
print(f"Whisper Transcription: {transcription}") print(f"Whisper Transcription: {transcription}")
return transcription return transcription
else: else:
print("Audio energy below threshold; no transcription performed.")
return "" return ""
else: else:
# Convert raw audio_data to PCM WAV format using an in-memory buffer # Google SpeechRecognition code (no changes here)
recognizer = sr.Recognizer() recognizer = sr.Recognizer()
# Convert audio_data to a WAV file in memory
audio_buffer = io.BytesIO() audio_buffer = io.BytesIO()
with wave.open(audio_buffer, 'wb') as wav_file: with wave.open(audio_buffer, 'wb') as wav_file:
@ -122,6 +123,7 @@ class Speak:
print(f"Could not request results; {e}") print(f"Could not request results; {e}")
return "" return ""
def listen(self, time_listen=8): def listen(self, time_listen=8):
"""Main transcoder function that handles listening, noise cancellation, and transcription.""" """Main transcoder function that handles listening, noise cancellation, and transcription."""
# Listen to the microphone and get both raw audio and ambient noise # Listen to the microphone and get both raw audio and ambient noise