dearwolf/aider
1
0
Fork 0
mirror of https://github.com/Aider-AI/aider.git synced 2025-06-04 11:45:00 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

Refactored voice recording to use a queue and write audio to a file.

Browse source
This commit is contained in:
Paul Gauthier 2023-08-10 16:10:02 -03:00
parent 1f42b0839f
commit 9b526d51e4
1 changed files with 21 additions and 17 deletions
Download patch file Download diff file Expand all files Collapse all files

38
aider/voice.py
View file

@ -3,7 +3,9 @@ import numpy as np
import keyboard import keyboard
import openai import openai
import io import io
import tempfile
import queue
import soundfile as sf
import os import os
def record_and_transcribe(api_key): def record_and_transcribe(api_key):
@ -12,25 +14,27 @@ def record_and_transcribe(api_key):
sample_rate = 16000 # 16kHz sample_rate = 16000 # 16kHz
duration = 10 # in seconds duration = 10 # in seconds
# Create a callback function to stop recording when a key is pressed def callback(indata, frames, time, status):
def on_key_press(e): """This is called (from a separate thread) for each audio block."""
print("Key pressed, stopping recording...") if status:
sd.stop() print(status, file=sys.stderr)
q.put(indata.copy())
# Start the recording
print("Recording started, press any key to stop...")
# Create an instance of InputStream with the callback
stream = sd.InputStream(samplerate=sample_rate, channels=1, callback=on_key_press)
stream.start()
recording = sd.rec(int(sample_rate * duration), samplerate=sample_rate, channels=1)
# Wait for a key press filename = tempfile.mktemp(prefix='delme_rec_unlimited_', suffix='.wav', dir='')
keyboard.wait()
# Convert the recording to bytes q = queue.Queue()
recording_bytes = io.BytesIO()
np.save(recording_bytes, recording, allow_pickle=False) # Make sure the file is opened before recording anything:
recording_bytes = recording_bytes.getvalue() with sf.SoundFile(filename, mode='x', samplerate=sample_rate, channels=1) as file:
with sd.InputStream(samplerate=sample_rate, channels=1, callback=callback):
input('Press enter when done')
while not q.empty():
print('.')
file.write(q.get())
print('done')
# Transcribe the audio using the Whisper API # Transcribe the audio using the Whisper API
response = openai.Whisper.asr.create(audio_data=recording_bytes) response = openai.Whisper.asr.create(audio_data=recording_bytes)