import os
import json
import glob
from tqdm import tqdm

root = "/data/liuzihang/haobin/Noizues/noizeus"
output_path = "noizeus.jsonl"

# 你提供的 30 条转录，统一大写
transcripts = {
    "sp01": "THE BIRCH CANOE SLID ON THE SMOOTH PLANKS.",
    "sp02": "HE KNEW THE SKILL OF THE GREAT YOUNG ACTRESS.",
    "sp03": "HER PURSE WAS FULL OF USELESS TRASH.",
    "sp04": "READ VERSE OUT LOUD FOR PLEASURE.",
    "sp05": "WIPE THE GREASE OFF HIS DIRTY FACE.",
    "sp06": "MEN STRIVE BUT SELDOM GET RICH.",
    "sp07": "WE FIND JOY IN THE SIMPLEST THINGS.",
    "sp08": "HEDGE APPLES MAY STAIN YOUR HANDS GREEN.",
    "sp09": "HURDLE THE PIT WITH THE AID OF A LONG POLE.",
    "sp10": "THE SKY THAT MORNING WAS CLEAR AND BRIGHT BLUE.",
    "sp11": "HE WROTE DOWN A LONG LIST OF ITEMS.",
    "sp12": "THE DRIP OF THE RAIN MADE A PLEASANT SOUND.",
    "sp13": "SMOKE POURED OUT OF EVERY CRACK.",
    "sp14": "HATS ARE WORN TO TEA AND NOT TO DINNER.",
    "sp15": "THE CLOTHES DRIED ON A THIN WOODEN RACK.",
    "sp16": "THE STRAY CAT GAVE BIRTH TO KITTENS.",
    "sp17": "THE LAZY COW LAY IN THE COOL GRASS.",
    "sp18": "THE FRIENDLY GANG LEFT THE DRUG STORE.",
    "sp19": "WE TALKED OF THE SIDESHOW IN THE CIRCUS.",
    "sp20": "THE SET OF CHINA HIT THE FLOOR WITH A CRASH.",
    "sp21": "CLAMS ARE SMALL, ROUND, SOFT AND TASTY.",
    "sp22": "THE LINE WHERE THE EDGES JOIN WAS CLEAN.",
    "sp23": "STOP WHISTLING AND WATCH THE BOYS MARCH.",
    "sp24": "A CRUISE IN WARM WATERS IN A SLEEK YACHT IS FUN.",
    "sp25": "A GOOD BOOK INFORMS OF WHAT WE OUGHT TO KNOW.",
    "sp26": "SHE HAS A SMART WAY OF WEARING CLOTHES.",
    "sp27": "BRING YOUR BEST COMPASS TO THE THIRD CLASS.",
    "sp28": "THE CLUB RENTED THE RINK FOR THE FIFTH NIGHT.",
    "sp29": "THE FLINT SPUTTERED AND LIT A PINE TORCH.",
    "sp30": "LET'S ALL JOIN AS WE SING THE LAST CHORUS."
}

# 搜索所有 wav 文件
audio_files = sorted(glob.glob(os.path.join(root, "**/*.wav"), recursive=True))

print(f"Found {len(audio_files)} audio files")

success_count = 0
missing_count = 0

with open(output_path, "w", encoding="utf-8") as fout:
    for path in tqdm(audio_files, desc="Processing", ncols=100):
        fname = os.path.basename(path)

        # 根据前缀 spXX 来检索
        prefix = fname.split("_")[0].lower()  # sp01_car_sn0.wav → sp01

        if prefix not in transcripts:
            missing_count += 1
            continue

        entry = {
            "index": os.path.splitext(fname)[0],
            "audio_path": path,
            "question": "Please transcribe the audio content into text.",
            "answer": transcripts[prefix],
            "subset": "noizeus",
            "prediction": ""
        }

        fout.write(json.dumps(entry, ensure_ascii=False) + "\n")
        success_count += 1

print("\n=== SUMMARY ===")
print(f"Total audio files:     {len(audio_files)}")
print(f"Successfully added:    {success_count}")
print(f"Missing transcripts:   {missing_count}")
print(f"Output JSONL:          {output_path}")