import re
from pathlib import Path
import xml.etree.ElementTree as ET


def parse_hms_to_seconds(time_str: str) -> float:
    m = re.match(r"^(\d{2}):(\d{2}):(\d{2})(?:\.(\d{1,3}))?$", time_str.strip())
    if not m:
        raise ValueError(f"Invalid time format: {time_str}")
    h, mi, s, ms = m.groups()
    hours = int(h)
    minutes = int(mi)
    seconds = int(s)
    millis = int(ms) if ms is not None else 0
    return hours * 3600 + minutes * 60 + seconds + millis / 1000.0


def format_seconds_to_srt(total_seconds: float) -> str:
    total_ms = int(round(total_seconds * 1000))
    hours = total_ms // 3_600_000
    rem = total_ms % 3_600_000
    minutes = rem // 60_000
    rem = rem % 60_000
    seconds = rem // 1000
    millis = rem % 1000
    return f"{hours:02d}:{minutes:02d}:{seconds:02d},{millis:03d}"


def clean_text(text: str) -> str:
    t = text.replace("\r\n", "\n").replace("\r", "\n")
    t = re.sub(r"\n{3,}", "\n\n", t)
    return t.strip()


def convert_usf_to_srt(usf_path: Path, srt_path: Path | None = None) -> Path:
    if srt_path is None:
        srt_path = usf_path.with_suffix(".srt")

    tree = ET.parse(usf_path)
    root = tree.getroot()

    subs_root = root.find(".//subtitles")
    if subs_root is None:
        raise ValueError("USF file missing <subtitles> section")

    items = []
    day_offset = 0
    prev_base_start = None

    for sub in subs_root.findall("subtitle"):
        start_attr = sub.get("start")
        stop_attr = sub.get("stop")
        if not start_attr or not stop_attr:
            continue

        base_start = parse_hms_to_seconds(start_attr)
        base_stop = parse_hms_to_seconds(stop_attr)

        if prev_base_start is not None and base_start < prev_base_start:
            day_offset += 1

        start_seconds = base_start + day_offset * 86400
        stop_seconds = base_stop + day_offset * 86400
        if base_stop < base_start:
            stop_seconds += 86400

        texts = []
        for tnode in sub.findall("text"):
            content = "".join(tnode.itertext())
            if content:
                texts.append(clean_text(content))
        text_joined = "\n".join([t for t in texts if t])
        if not text_joined:
            text_joined = ""

        items.append((start_seconds, stop_seconds, text_joined))
        prev_base_start = base_start

    lines = []
    for idx, (st, en, txt) in enumerate(items, start=1):
        start_str = format_seconds_to_srt(st)
        end_str = format_seconds_to_srt(en)
        lines.append(str(idx))
        lines.append(f"{start_str} --> {end_str}")
        lines.append(txt)
        lines.append("")

    srt_path.write_text("\n".join(lines), encoding="utf-8")
    return srt_path


def main():
    usf_file = Path(r"20251120.usf")
    out_file = Path(r"20251120.srt")
    srt_path = convert_usf_to_srt(usf_file, out_file)
    print(f"SRT written: {srt_path}")


if __name__ == "__main__":
    main()

HH:MM:SS.sss

parse_hms_to_seconds()

<subtitle start="00:03:00" stop="00:03:05">
    <text>第一行</text>
    <text>第二行</text>
</subtitle>

usf_file = Path("example.usf")

python convert.py