whisper speech-to-text

process_wav_whispers.py

from pathlib import Path

import whisper

# CHANGE THE MODEL HERE: use "base" or "medium"
MODEL_NAME = "base"

MODEL = whisper.load_model(MODEL_NAME)

with open(MODEL_NAME + ".txt", "a", encoding="utf-8") as result_file:
    for path in sorted(
        Path("wav").glob("*.wav"), key=lambda p: int(p.stem.replace(".wav", ""))
    ):
        print("processing " + str(path) + " ...")
        result = MODEL.transcribe(str(path), fp16=False, language="English")
        transcript = result["text"].strip()
        print(transcript)
        result_file.write(transcript)
        result_file.write("\n")