111

scripts/nodejs/subtitle_keyword_drawtext.js

@@ -0,0 +1,363 @@
+"use strict";
+
+/**
+ * 关键词高亮 drawtext（对齐 Electron ZimuShengcheng inline 模式 + pulse-scale）
+ */
+
+const fs = require("fs");
+const path = require("path");
+const os = require("os");
+
+const PUNCT_RE =
+  /[，。！？；：、""''（）《》【】…—·,.!?;:'"()[\]{}<>]/g;
+
+const DEFAULT_GROUP_META = {
+  focus: {
+    name: "重点词",
+    effectId: "pulse-scale",
+    styleConfig: {
+      fontColor: "#FFD700",
+      outlineColor: "#FF8C00",
+      outlineWidth: 4,
+    },
+  },
+  describe: {
+    name: "描述词",
+    effectId: "pulse-scale",
+    styleConfig: {
+      fontColor: "#5AC8FA",
+      outlineColor: "#007AFF",
+      outlineWidth: 3,
+    },
+  },
+  action: {
+    name: "行动词",
+    effectId: "pulse-scale",
+    styleConfig: {
+      fontColor: "#34C759",
+      outlineColor: "#248A3D",
+      outlineWidth: 3,
+    },
+  },
+  emotion: {
+    name: "情感词",
+    effectId: "pulse-scale",
+    styleConfig: {
+      fontColor: "#FF6B6B",
+      outlineColor: "#C92A2A",
+      outlineWidth: 3,
+    },
+  },
+};
+
+function parseKeywordLines(text) {
+  return String(text || "")
+    .split(/[\n,，、]+/)
+    .map((s) => s.trim())
+    .filter(Boolean);
+}
+
+/**
+ * 从步骤 04 字段或显式 keywordGroups 构建分组
+ * @param {Record<string, unknown>} params
+ */
+function buildKeywordGroups(params) {
+  if (Array.isArray(params.keywordGroups) && params.keywordGroups.length > 0) {
+    return params.keywordGroups
+      .map((g) => normalizeKeywordGroup(g))
+      .filter((g) => g.keywords.length > 0);
+  }
+
+  const sources = [
+    ["focus", params.keywordsFocus],
+    ["describe", params.keywordsDescribe],
+    ["action", params.keywordsAction],
+    ["emotion", params.keywordsEmotion],
+  ];
+
+  return sources
+    .map(([id, raw]) => {
+      const meta = DEFAULT_GROUP_META[id] || DEFAULT_GROUP_META.focus;
+      const keywords = parseKeywordLines(raw);
+      if (!keywords.length) return null;
+      return {
+        id,
+        name: meta.name,
+        effectId: meta.effectId,
+        styleConfig: { ...meta.styleConfig },
+        keywords,
+      };
+    })
+    .filter(Boolean);
+}
+
+function normalizeKeywordGroup(group) {
+  const g = group && typeof group === "object" ? group : {};
+  const keywords = (Array.isArray(g.keywords) ? g.keywords : [])
+    .map((k) => (typeof k === "string" ? k : String(k?.keyword || "")).trim())
+    .filter(Boolean);
+  return {
+    id: String(g.id || g.groupId || "group"),
+    name: String(g.name || g.groupName || "关键词"),
+    effectId: String(g.effectId || g.effectConfig?.type || "pulse-scale"),
+    styleConfig: g.styleConfig || g.styleOverride || {},
+    keywords,
+  };
+}
+
+function normalizeDisplayText(text) {
+  return String(text || "")
+    .replace(PUNCT_RE, "")
+    .replace(/\s+/g, "");
+}
+
+function normalizeEffectType(effectId) {
+  const id = String(effectId || "pulse-scale").toLowerCase();
+  if (id.includes("pulse") || id.includes("scale")) return "pulse-scale";
+  if (id.includes("slide")) return "slide";
+  return "highlight";
+}
+
+function buildKeywordStyle(group, keywordItem) {
+  const base = group.styleConfig || {};
+  const override =
+    keywordItem && typeof keywordItem === "object" ? keywordItem.styleOverride : null;
+  const o = override || base;
+  return {
+    fontColor: o.fontColor || "#FFD700",
+    outlineColor: o.outlineColor || "#FF8C00",
+    outlineWidth: Number(o.outlineWidth) ?? 4,
+    fontSize: Number(o.fontSize) || 0,
+  };
+}
+
+/**
+ * @param {Array<{ text: string, start: number, end: number }>} records 毫秒
+ * @param {ReturnType<typeof buildKeywordGroups>} groups
+ */
+function matchKeywordsInRecords(records, groups) {
+  const matches = [];
+  const seen = new Set();
+
+  for (const record of records) {
+    const display = normalizeDisplayText(record.text);
+    if (!display) continue;
+
+    const segStartSec = record.start / 1000;
+    const segEndSec = record.end / 1000;
+    const dur = Math.max(0.05, segEndSec - segStartSec);
+
+    for (const group of groups) {
+      for (const kw of group.keywords) {
+        const raw = typeof kw === "string" ? kw : String(kw?.keyword || "");
+        const needle = normalizeDisplayText(raw);
+        if (!needle) continue;
+
+        let searchFrom = 0;
+        while (searchFrom < display.length) {
+          const idx = display.indexOf(needle, searchFrom);
+          if (idx === -1) break;
+
+          const dedupeKey = `${segStartSec.toFixed(3)}-${segEndSec.toFixed(3)}-${needle}-${idx}`;
+          if (!seen.has(dedupeKey)) {
+            seen.add(dedupeKey);
+            const ratioStart = idx / display.length;
+            const ratioEnd = (idx + needle.length) / display.length;
+            const startSec = segStartSec + dur * ratioStart;
+            const endSec = Math.max(startSec + 0.08, segStartSec + dur * ratioEnd);
+
+            matches.push({
+              keyword: needle,
+              text: needle,
+              lineText: display,
+              lineLength: display.length,
+              occurrenceStart: idx,
+              occurrenceEnd: idx + needle.length,
+              startSec,
+              endSec,
+              groupId: group.id,
+              groupName: group.name,
+              effectType: normalizeEffectType(
+                (typeof kw === "object" && kw?.effectId) || group.effectId,
+              ),
+              style: buildKeywordStyle(group, kw),
+            });
+          }
+          searchFrom = idx + Math.max(1, needle.length);
+        }
+      }
+    }
+  }
+
+  return matches.sort((a, b) => a.startSec - b.startSec);
+}
+
+const WIN_FONT_MAP = {
+  "microsoft yahei": ["msyh.ttc", "msyhbd.ttc"],
+  "microsoft yahei bold": ["msyhbd.ttc", "msyh.ttc"],
+  simhei: ["simhei.ttf"],
+  "sim sun": ["simsun.ttc"],
+  "source han sans": ["SourceHanSansSC-Regular.otf"],
+};
+
+function resolveFontFile(fontName) {
+  const fromEnv = process.env.AICLIENT_SUBTITLE_FONT_PATH;
+  if (fromEnv && fs.existsSync(fromEnv)) return fromEnv;
+
+  const key = String(fontName || "Microsoft YaHei")
+    .trim()
+    .toLowerCase();
+  const candidates = WIN_FONT_MAP[key] || WIN_FONT_MAP["microsoft yahei"];
+
+  if (process.platform === "win32") {
+    const fontsDir = path.join(process.env.WINDIR || "C:\\Windows", "Fonts");
+    for (const file of candidates) {
+      const full = path.join(fontsDir, file);
+      if (fs.existsSync(full)) return full;
+    }
+  }
+
+  const fallbacks = [
+    "/usr/share/fonts/truetype/noto/NotoSansCJK-Regular.ttc",
+    "/System/Library/Fonts/PingFang.ttc",
+    "/System/Library/Fonts/Supplemental/Arial Unicode.ttf",
+  ];
+  for (const p of fallbacks) {
+    if (fs.existsSync(p)) return p;
+  }
+  return null;
+}
+
+function escapeDrawtextPath(filePath) {
+  return filePath
+    .replace(/\\/g, "/")
+    .replace(/:/g, "\\:")
+    .replace(/'/g, "\\'");
+}
+
+function escapeDrawtextText(text) {
+  return String(text || "")
+    .replace(/\\/g, "\\\\")
+    .replace(/'/g, "\\'")
+    .replace(/:/g, "\\:")
+    .replace(/,/g, "\\,");
+}
+
+function hexToDrawtextColor(hex) {
+  const h = String(hex || "#FFFFFF").replace("#", "").slice(0, 6);
+  if (h.length !== 6) return "white";
+  return `0x${h.toUpperCase()}`;
+}
+
+function resolveSubtitleY(position, videoHeight) {
+  const h = Math.max(360, Number(videoHeight) || 1080);
+  if (position === "top") return "h*0.12";
+  if (position === "center") return "(h-text_h)/2";
+  return "h*0.82-text_h";
+}
+
+/**
+ * @param {ReturnType<typeof matchKeywordsInRecords>[number]} match
+ * @param {{ videoWidth: number, videoHeight: number, baseFontSize: number, fontfile: string, position: string }} layout
+ */
+function buildDrawtextForMatch(match, layout) {
+  const { videoWidth, videoHeight, baseFontSize, fontfile, position } = layout;
+  const style = match.style || {};
+  const baseSize = style.fontSize > 0 ? style.fontSize : Math.round(baseFontSize * 1.45);
+  const outlineW = Math.max(1, Math.round(style.outlineWidth || 3));
+  const fontColor = hexToDrawtextColor(style.fontColor);
+  const borderColor = hexToDrawtextColor(style.outlineColor);
+
+  const lineLen = Math.max(1, match.lineLength);
+  const charWidth = baseSize * 0.55;
+  const lineWidth = Math.min(lineLen * charWidth, videoWidth * 0.88);
+  const keywordMid = (match.occurrenceStart + match.occurrenceEnd) / 2;
+  const offsetPx = (keywordMid / lineLen) * lineWidth - lineWidth / 2;
+  const xExpr = `(w-text_w)/2+${offsetPx.toFixed(1)}`;
+  const yExpr = resolveSubtitleY(position, videoHeight);
+
+  const t0 = match.startSec.toFixed(3);
+  const t1 = match.endSec.toFixed(3);
+  const enable = `between(t\\,${t0}\\,${t1})`;
+
+  let fontsizePart = String(baseSize);
+  if (match.effectType === "pulse-scale") {
+    fontsizePart = `${baseSize}*(1+0.1*sin(2*PI*(t-${t0})*4))`;
+  }
+
+  const parts = [
+    `fontfile='${escapeDrawtextPath(fontfile)}'`,
+    `text='${escapeDrawtextText(match.text)}'`,
+    `fontsize='${fontsizePart}'`,
+    `fontcolor=${fontColor}`,
+    `borderw=${outlineW}`,
+    `bordercolor=${borderColor}`,
+    `x='${xExpr}'`,
+    `y='${yExpr}'`,
+    `enable='${enable}'`,
+  ];
+  return `drawtext=${parts.join(":")}`;
+}
+
+/**
+ * @param {ReturnType<typeof matchKeywordsInRecords>} matches
+ * @param {{ videoWidth: number, videoHeight: number, subtitleStyle?: Record<string, unknown> }} opts
+ * @returns {string | null} ffmpeg -vf 链（逗号分隔的 drawtext）
+ */
+function buildKeywordDrawtextFilter(matches, opts) {
+  if (!matches.length) return null;
+
+  const videoWidth = Math.max(320, Number(opts.videoWidth) || 1920);
+  const videoHeight = Math.max(320, Number(opts.videoHeight) || 1080);
+  const style = opts.subtitleStyle || {};
+  const scale = Math.max(0.2, Math.min(2, (Number(style.fontSizeScale) || 100) / 100));
+  const baseFontSize = Math.max(14, Math.round((Number(style.fontSize) || 35) * scale));
+  const fontfile = resolveFontFile(style.fontName);
+  if (!fontfile) return null;
+
+  const layout = {
+    videoWidth,
+    videoHeight,
+    baseFontSize,
+    fontfile,
+    position: String(style.position || "bottom"),
+  };
+
+  return matches.map((m) => buildDrawtextForMatch(m, layout)).join(",");
+}
+
+/**
+ * 从 ASS 底字中移除关键词字符，避免与 drawtext 高亮重复
+ * @param {string} displayText 已去标点文本
+ * @param {Array<{ occurrenceStart: number, occurrenceEnd: number }>} matches
+ */
+function stripKeywordsFromDisplay(displayText, matches) {
+  const text = String(displayText || "");
+  if (!text || !matches.length) return text;
+  const chars = [...text];
+  for (const m of matches) {
+    const start = Math.max(0, m.occurrenceStart);
+    const end = Math.min(chars.length, m.occurrenceEnd);
+    for (let i = start; i < end; i++) chars[i] = "";
+  }
+  return chars.join("").trim();
+}
+
+/** @param {{ start: number, end: number }} record 毫秒 */
+function filterMatchesForRecord(record, matches) {
+  const segStart = record.start / 1000;
+  const segEnd = record.end / 1000;
+  return matches.filter(
+    (m) => m.startSec >= segStart - 0.02 && m.endSec <= segEnd + 0.02,
+  );
+}
+
+module.exports = {
+  buildKeywordGroups,
+  matchKeywordsInRecords,
+  buildKeywordDrawtextFilter,
+  normalizeDisplayText,
+  stripKeywordsFromDisplay,
+  filterMatchesForRecord,
+  resolveFontFile,
+};