transcribe_whisper.py

import sys
import subprocess
from pathlib import Path

ROOT = Path(__file__).resolve().parents[1]
OUT_DIR = ROOT / "data" / "output"
OUT_TXT = OUT_DIR / "transcript.txt"

def main(audio_path: str):
    OUT_DIR.mkdir(parents=True, exist_ok=True)

    audio = Path(audio_path).expanduser().resolve()
    if not audio.is_file():
        raise FileNotFoundError(f"Audio file not found: {audio}")

    # whisper-cli writes transcript to "<-of>.txt"
    out_prefix = OUT_DIR / "transcript"  # will create transcript.txt

    cmd = [
        "whisper-cli",
        "-m", str(ROOT / "whisper.cpp" / "models" / "ggml-base.en.bin"),
        "-f", str(audio),
        "-of", str(out_prefix),
        "-otxt",  # ensure txt output enabled
        "-nt",    # optional: no timestamps in terminal (file still ok)
    ]

    print("⏳ Loading ASR model (10%)")
    subprocess.run(cmd, check=True)

    # whisper-cli writes to transcript.txt (based on -of)
    generated = OUT_DIR / "transcript.txt"
    if not generated.exists():
        raise RuntimeError("Whisper did not generate transcript.txt (output file missing).")

    text = generated.read_text(encoding="utf-8", errors="ignore").strip()
    if not text:
        raise RuntimeError(
            "Whisper produced empty transcript.\n"
            "Common causes:\n"
            "1) The audio is silence / wrong file\n"
            "2) The input path is wrong\n"
            "3) whisper-cli failed but still printed timings\n"
        )

    # normalize final location/name (keep same file name)
    OUT_TXT.write_text(text + "\n", encoding="utf-8")
    print("✅ Transcription complete (40%)")
    print(f"✅ Saved transcript to {OUT_TXT}")

if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("Usage: python src/transcribe_whisper.py <audio.wav>")
        sys.exit(1)
    main(sys.argv[1])