fork(fix): Clone AMLL 并修复 BUG

- 将AMLL Clone到本以地进行修复和优化(emm虽然这很不优雅但是暂时无时间做子模块和Fork)
- 修复在当前播放歌词行不可见的视口Seek会出现滚动偏移的问题

amll-local/packages/lyric/src/formats/ass.ts

@@ -0,0 +1,109 @@
+/**
+ * @fileoverview ASS 字幕格式导出。
+ * 注意导出会损失 10ms 以内精度（按厘秒四舍五入）。
+ *
+ * 格式示例：
+ * Dialogue: 0,0:00:12.34, 0:00:15.67, Default, v1,0,0,0,,{\k20}Hello{\k15} world
+ * Dialogue: 0,0:00:12.34, 0:00:15.67, Default, v1-trans,0,0,0,,你好 世界
+ * Dialogue: 0,0:00:12.34, 0:00:15.67, Default, v1-roman,0,0,0,,ni hao shi jie
+ */
+import type { LyricLine } from "../types";
+import { normalizeTimestamp } from "../utils";
+
+function writeASSTimestamp(ms: number): string {
+	const normalized = normalizeTimestamp(ms);
+	const milli = Math.round(normalized) % 1000;
+	const secTotal = Math.floor(Math.round(normalized) / 1000);
+	const sec = secTotal % 60;
+	const minTotal = Math.floor(secTotal / 60);
+	const hour = Math.floor(minTotal / 60);
+	return `${hour}:${String(minTotal % 60).padStart(2, "0")}:${String(sec).padStart(2, "0")}.${String(Math.floor(milli / 10)).padStart(2, "0")}`;
+}
+
+function getSpeakerName(line: LyricLine): string {
+	let name = line.isDuet ? "v2" : "v1";
+	if (line.isBG) name += "-bg";
+	return name;
+}
+
+function writeLyricDialogue(
+	result: string[],
+	startTime: number,
+	endTime: number,
+	name: string,
+	text: string,
+) {
+	result.push(
+		`Dialogue: 0,${writeASSTimestamp(startTime)}, ${writeASSTimestamp(endTime)}, Default, ${name},0,0,0,,${text}`,
+	);
+}
+
+/**
+ * 将歌词数组转换为 ASS 字幕格式字符串
+ * @param lines 歌词数组
+ * @returns ASS 字幕格式字符串
+ */
+export function stringifyAss(lines: LyricLine[]): string {
+	const result: string[] = [
+		"[Script Info]",
+		"[Events]",
+		"Formats: Marked, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text",
+	];
+
+	for (const line of lines) {
+		const timedWords = line.words
+			.map((w) => ({
+				...w,
+				startTime: normalizeTimestamp(w.startTime),
+				endTime: normalizeTimestamp(w.endTime),
+			}))
+			.filter((w) => w.endTime > w.startTime);
+		const startTime = Math.min(...timedWords.map((w) => w.startTime));
+		const endTime = Math.max(...timedWords.map((w) => w.endTime));
+		if (!Number.isFinite(startTime) || !Number.isFinite(endTime)) continue;
+
+		let lyricText = "";
+		let previousWordEndTime = startTime;
+		for (const word of line.words) {
+			const wordStart = normalizeTimestamp(word.startTime);
+			const wordEnd = normalizeTimestamp(word.endTime);
+			if (wordStart >= wordEnd) {
+				lyricText += word.word;
+				continue;
+			}
+
+			if (wordStart > previousWordEndTime) {
+				const gapDurationCS = Math.floor(
+					(wordStart - previousWordEndTime + 5) / 10,
+				);
+				if (gapDurationCS > 0) lyricText += `{\\k${gapDurationCS}}`;
+			}
+
+			const wordDurationCS = Math.floor((wordEnd - wordStart + 5) / 10);
+			if (wordDurationCS > 0) lyricText += `{\\k${wordDurationCS}}`;
+			lyricText += word.word;
+			previousWordEndTime = wordEnd;
+		}
+
+		const speaker = getSpeakerName(line);
+		writeLyricDialogue(result, startTime, endTime, speaker, lyricText);
+		if (line.translatedLyric)
+			writeLyricDialogue(
+				result,
+				startTime,
+				endTime,
+				`${speaker}-trans`,
+				line.translatedLyric,
+			);
+		if (line.romanLyric)
+			writeLyricDialogue(
+				result,
+				startTime,
+				endTime,
+				`${speaker}-roman`,
+				line.romanLyric,
+			);
+	}
+
+	return `${result.join("\n")}\n`;
+}