recognizer_tunisian_vosk.py

import wave
import json
from vosk import Model as VoskModel, KaldiRecognizer as VoskRecognizer
import os

class RecognizerTunisianVosk:
    def __init__(self, recognizer_name: str = "vosk", vosk_model_dir: str = "model/vosk-model-small-ar-tn-0.1-linto"):
        self.recognizer_name = recognizer_name
        self.vosk_model_dir = vosk_model_dir
        if not os.path.exists(self.vosk_model_dir):
            raise ValueError(f"Vosk model directory '{self.vosk_model_dir}' does not exist.")
        self.vosk_model = VoskModel(self.vosk_model_dir)

    def transcribe(self, audio_path: str) -> str:
        """
        Transcribe speech from an audio file.
        
        :param audio_path: Path to the WAV file.
        :return: Transcribed text.
        """
        with wave.open(audio_path, "rb") as wf:
            if wf.getnchannels() != 1 or wf.getsampwidth() != 2 or wf.getcomptype() != "NONE":
                raise ValueError("Audio file must be WAV format mono PCM (16-bit, mono, uncompressed).")

            recognizer = VoskRecognizer(self.vosk_model, wf.getframerate())
            recognizer.AcceptWaveform(wf.readframes(wf.getnframes()))
            result = recognizer.FinalResult()
            return json.loads(result)["text"]