accented.py

import numpy as np
import pyaudio
from faster_whisper import WhisperModel
import queue
import threading
import os
from dotenv import load_dotenv
load_dotenv()
HF_TOKEN = os.getenv("HF_TOKEN")
print(HF_TOKEN)

# Load Whisper model
model = WhisperModel("small", compute_type="int8")  # Changed compute_type from float16 to int8 for compatibility

# Audio settings
RATE = 16000
CHUNK = 1024
RECORD_SECONDS = 2
print("✅ Model loaded and ready to transcribe!")
audio_queue = queue.Queue()
print("✅ Model loaded and ready to transcribe!")

def audio_capture():
    p = pyaudio.PyAudio()
    stream = p.open(format=pyaudio.paInt16,
                    channels=1,
                    rate=RATE,
                    input=True,
                    frames_per_buffer=CHUNK)

    print("🎙️ Listening...")

    frames = []

    while True:
        data = stream.read(CHUNK)
        frames.append(data)

        if len(frames) >= int(RATE / CHUNK * RECORD_SECONDS):
            audio_data = b''.join(frames)
            frames = []
            audio_queue.put(audio_data)

def transcribe():
    while True:
        audio_data = audio_queue.get()
        audio_np = np.frombuffer(audio_data, np.int16).astype(np.float32) / 32768.0

        segments, _ = model.transcribe(audio_np, language="en")

        for segment in segments:
            print("📝", segment.text)

# Start threads
threading.Thread(target=audio_capture).start()
threading.Thread(target=transcribe).start()