feat(proofread): 支持角色命名、纠错与批量管理

docs/docs/features/proofreading.md

@@ -32,6 +32,14 @@ keywords: [字幕校对, 字幕编辑, 校对台, AI 润色, 字幕修改, 时
 
 常用快捷键：`↑↓` 上一条 / 下一条，`Tab` 在原文与译文间切换，`⌘S` 保存，`?` 查看全部快捷键。
 
+## 校正角色归属
+
+启用[角色分离](./subtitle-generation#角色分离)的任务会在每条字幕旁显示角色名和颜色。点击角色标签即可重新分配当前字幕、补充多人重叠角色或指定主要角色；没有可靠识别结果的字幕会显示为「未分配」。
+
+角色工具栏支持按角色、未分配和多人重叠筛选，也可以统一重命名、换色、移动全部字幕或合并误拆角色。角色名称和归属独立保存在校对数据中，相关修改都可撤销 / 重做，不会改动字幕正文或时间轴。
+
+保存时默认不把角色名写入字幕文件；需要交付带角色标记的字幕时，可勾选「在字幕中写入角色名」，导出内容会使用当前角色名称。
+
 ## AI 帮你改
 
 - **AI 优化**：选中单条让 AI 润色措辞

docs/docs/features/subtitle-generation.md

@@ -61,7 +61,7 @@ keywords:
 
 标准的「转写」和「转写 + 翻译」任务可在高级设置中开启**角色分离**。它会在转写、精修与翻译完成后，本地分析整段音频，把不同角色按时间轴对齐到每条字幕；处理耗时与音频长度大致成正比。
 
-首次使用前需在「资源中心 → 本地多模型引擎」确认运行库可用并下载角色分离模型。角色信息默认只保存在应用内校对数据中，不会改变导出的字幕；需要交付可读标签时，可同时开启「将角色标签写入字幕文件」，输出 `[Speaker 1]` 等前缀。
+首次使用前需在「资源中心 → 本地多模型引擎」确认运行库可用并下载角色分离模型。角色信息默认只保存在应用内校对数据中，不会改变导出的字幕；需要交付可读标签时，可同时开启「将角色标签写入字幕文件」，初始输出使用 `[Speaker 1]` 等前缀。进入[校对台](./proofreading#校正角色归属)后可以把编号改为真实角色名、纠正归属，并决定保存时是否写入当前角色名称。
 
 当前版本暂不在一条龙向导、自定义配方及含配音 / 合成的流程中开放角色分离；这些流程会在角色元数据与配音音色映射完成后再接入。
 

main/helpers/ipcHandlers.ts

@@ -20,6 +20,10 @@ import {
   readProofreadDataFile,
   updateProofreadDataFromSubtitles,
 } from './proofreadData';
+import {
+  prefixTextWithSpeakerNames,
+  type SpeakerInfo,
+} from '../../types/proofreadData';
 import {
   MANUSCRIPT_EXTENSIONS,
   ManuscriptFileError,
@@ -164,18 +168,33 @@ function buildSubtitleFileContent(
   filePath: string,
   subtitles: any[],
   contentType = 'source',
+  speakerOptions?: {
+    speakers?: SpeakerInfo[];
+    embedSpeakerNames?: boolean;
+  },
 ): string {
   const format = detectSubtitleFormat(filePath);
+  const withSpeakerPrefix = (subtitle: any, text: string): string => {
+    if (!speakerOptions?.embedSpeakerNames) return text;
+    return prefixTextWithSpeakerNames(
+      text,
+      subtitle,
+      speakerOptions.speakers || [],
+    );
+  };
   const buildText = (subtitle): string => {
+    let text: string;
     if (contentType === 'source') {
-      return subtitle.sourceContent ?? '';
+      text = subtitle.sourceContent ?? '';
+    } else {
+      const template =
+        CONTENT_TEMPLATES[contentType] || CONTENT_TEMPLATES.onlyTranslate;
+      text = renderTemplate(template, {
+        sourceContent: subtitle.sourceContent ?? '',
+        targetContent: subtitle.targetContent ?? '',
+      }).replace(/\n+$/, '');
     }
-    const template =
-      CONTENT_TEMPLATES[contentType] || CONTENT_TEMPLATES.onlyTranslate;
-    return renderTemplate(template, {
-      sourceContent: subtitle.sourceContent ?? '',
-      targetContent: subtitle.targetContent ?? '',
-    }).replace(/\n+$/, '');
+    return withSpeakerPrefix(subtitle, text);
   };
 
   return (
@@ -218,8 +237,17 @@ async function writeSubtitleFile(
   filePath: string,
   subtitles: any[],
   contentType = 'source',
+  speakerOptions?: {
+    speakers?: SpeakerInfo[];
+    embedSpeakerNames?: boolean;
+  },
 ): Promise<void> {
-  const content = buildSubtitleFileContent(filePath, subtitles, contentType);
+  const content = buildSubtitleFileContent(
+    filePath,
+    subtitles,
+    contentType,
+    speakerOptions,
+  );
   await backupSubtitleFile(filePath);
   await fs.promises.writeFile(filePath, content, 'utf-8');
   logMessage(`保存字幕文件成功: ${filePath}`, 'info');
@@ -386,7 +414,10 @@ export function setupIpcHandlers(mainWindow: BrowserWindow) {
         return [];
       }
       const proofreadData = await readProofreadDataFile(filePath);
-      return proofreadDataToSubtitleRows(proofreadData);
+      return {
+        subtitles: proofreadDataToSubtitleRows(proofreadData),
+        speakers: proofreadData.speakers,
+      };
     } catch (error) {
       logMessage(`读取校对中间态错误: ${error.message}`, 'error');
       return [];
@@ -447,10 +478,14 @@ export function setupIpcHandlers(mainWindow: BrowserWindow) {
       {
         proofreadDataFile,
         subtitles,
+        speakers = [],
+        embedSpeakerNames = false,
         outputs = [],
       }: {
         proofreadDataFile: string;
         subtitles: any[];
+        speakers?: SpeakerInfo[];
+        embedSpeakerNames?: boolean;
         outputs: { filePath?: string; contentType?: string }[];
       },
     ) => {
@@ -461,7 +496,11 @@ export function setupIpcHandlers(mainWindow: BrowserWindow) {
           );
         }
 
-        await updateProofreadDataFromSubtitles(proofreadDataFile, subtitles);
+        const updated = await updateProofreadDataFromSubtitles(
+          proofreadDataFile,
+          subtitles,
+          speakers,
+        );
 
         const rendered = new Set<string>();
         for (const output of outputs) {
@@ -474,6 +513,10 @@ export function setupIpcHandlers(mainWindow: BrowserWindow) {
             filePath,
             subtitles,
             output.contentType || 'source',
+            {
+              speakers: updated.speakers,
+              embedSpeakerNames,
+            },
           );
         }
 

main/helpers/proofreadData.ts

@@ -12,38 +12,26 @@ import {
 import { logMessage } from './storeManager';
 import {
   speakerIdsForCues,
+  stripSpeakerLabelPrefix,
   type SpeakerDiarizationSegment,
 } from './speakerDiarization/alignment';
 import {
   mergeSpeakerIds,
   realignSpeakerIdsForCue,
 } from '../../types/speakerDiarization';
+import {
+  PROOFREAD_DATA_VERSION,
+  normalizePrimarySpeakerId,
+  normalizeProofreadData,
+  normalizeSpeakerIds,
+  normalizeSpeakerRoster,
+  type ProofreadDataCue,
+  type ProofreadDataFileV2,
+  type SpeakerInfo,
+} from '../../types/proofreadData';
 
-export interface ProofreadDataCue {
-  id: string;
-  startMs: number;
-  endMs: number;
-  source: string;
-  target: string;
-  /** 一基角色编号，与可选文本标签 `[Speaker N]` 的 N 一致。 */
-  speakerIds?: number[];
-}
-
-export interface ProofreadDataFile {
-  version: 1;
-  meta: {
-    createdAt: string;
-    updatedAt: string;
-    sourceLanguage?: string;
-    targetLanguage?: string;
-    translateContent?: string;
-    outputFormat?: string;
-    sourceFile?: string;
-    targetFile?: string;
-    finalTargetFile?: string;
-  };
-  cues: ProofreadDataCue[];
-}
+export type { ProofreadDataCue, SpeakerInfo } from '../../types/proofreadData';
+export type ProofreadDataFile = ProofreadDataFileV2;
 
 export interface ProofreadSubtitleRow {
   id: string;
@@ -55,6 +43,7 @@ export interface ProofreadSubtitleRow {
   endTimeInSeconds: number;
   isEditing: boolean;
   speakerIds?: number[];
+  primarySpeakerId?: number;
 }
 
 export type ProofreadDataWriteResult =
@@ -125,14 +114,23 @@ function buildCues(
       targetByTime.get(sourceEntry.startEndTime) || targetEntries[index];
     const { startMs, endMs } = parseStartEndTime(sourceEntry.startEndTime);
 
+    const assignedSpeakerIds = normalizeSpeakerIds(speakerIds[index]);
+    const cleanSpeakerPrefix = assignedSpeakerIds.length > 0;
     const cue: ProofreadDataCue = {
       id: sourceEntry.id || String(index + 1),
       startMs,
       endMs,
-      source: entryText(sourceEntry),
-      target: entryText(targetEntry),
+      source: cleanSpeakerPrefix
+        ? stripSpeakerLabelPrefix(entryText(sourceEntry))
+        : entryText(sourceEntry),
+      target: cleanSpeakerPrefix
+        ? stripSpeakerLabelPrefix(entryText(targetEntry))
+        : entryText(targetEntry),
     };
-    if (speakerIds[index]?.length) cue.speakerIds = speakerIds[index];
+    if (assignedSpeakerIds.length) {
+      cue.speakerIds = assignedSpeakerIds;
+      cue.primarySpeakerId = assignedSpeakerIds[0];
+    }
     return cue;
   });
 }
@@ -170,8 +168,9 @@ export async function writeProofreadDataFromFiles({
 
     const targetEntries = await readSubtitleEntries(targetFile);
     const now = new Date().toISOString();
+    const cues = buildCues(sourceEntries, targetEntries, speakerSegments);
     const proofreadData: ProofreadDataFile = {
-      version: 1,
+      version: PROOFREAD_DATA_VERSION,
       meta: {
         createdAt: now,
         updatedAt: now,
@@ -183,7 +182,8 @@ export async function writeProofreadDataFromFiles({
         targetFile,
         finalTargetFile,
       },
-      cues: buildCues(sourceEntries, targetEntries, speakerSegments),
+      speakers: normalizeSpeakerRoster([], cues),
+      cues,
     };
 
     const proofreadDataFile = getProofreadDataPath(file);
@@ -211,11 +211,27 @@ export async function readProofreadDataFile(
   filePath: string,
 ): Promise<ProofreadDataFile> {
   const content = await fs.promises.readFile(filePath, 'utf-8');
-  const parsed = JSON.parse(content) as ProofreadDataFile;
-  if (parsed?.version !== 1 || !Array.isArray(parsed.cues)) {
+  try {
+    const raw = JSON.parse(content);
+    const normalized = normalizeProofreadData(raw);
+    // v1 was created while technical labels could still be embedded in the
+    // source/target text. Keep migration idempotent and only strip labels from
+    // cues that already carry structured speaker assignments.
+    if (raw?.version === 1) {
+      normalized.cues = normalized.cues.map((cue) =>
+        cue.speakerIds?.length
+          ? {
+              ...cue,
+              source: stripSpeakerLabelPrefix(cue.source),
+              target: stripSpeakerLabelPrefix(cue.target),
+            }
+          : cue,
+      );
+    }
+    return normalized;
+  } catch {
     throw new Error(`Invalid proofread data file: ${filePath}`);
   }
-  return parsed;
 }
 
 export function proofreadDataToSubtitleRows(
@@ -235,23 +251,29 @@ export function proofreadDataToSubtitleRows(
       endTimeInSeconds: cue.endMs / 1000,
       isEditing: false,
       ...(cue.speakerIds?.length ? { speakerIds: [...cue.speakerIds] } : {}),
+      ...(cue.primarySpeakerId
+        ? { primarySpeakerId: cue.primarySpeakerId }
+        : {}),
     };
   });
 }
 
 export async function updateProofreadDataFromSubtitles(
   filePath: string,
   subtitles: ProofreadSubtitleRow[],
+  speakers?: SpeakerInfo[],
 ): Promise<ProofreadDataFile> {
   const existing = await readProofreadDataFile(filePath);
   const existingById = new Map(existing.cues.map((cue) => [cue.id, cue]));
   const now = new Date().toISOString();
   const updated: ProofreadDataFile = {
     ...existing,
+    version: PROOFREAD_DATA_VERSION,
     meta: {
       ...existing.meta,
       updatedAt: now,
     },
+    speakers: normalizeSpeakerRoster(speakers || existing.speakers, subtitles),
     cues: subtitles.map((subtitle, index) => {
       const { startMs, endMs } = parseStartEndTime(subtitle.startEndTime);
       const source =
@@ -268,13 +290,23 @@ export async function updateProofreadDataFromSubtitles(
             subtitle.speakerIds || previous?.speakerIds,
           )
         : mergeSpeakerIds(subtitle.speakerIds || previous?.speakerIds);
+      const normalizedSpeakerIds = normalizeSpeakerIds(speakerIds);
+      const primarySpeakerId = normalizePrimarySpeakerId(
+        subtitle.primarySpeakerId || previous?.primarySpeakerId,
+        normalizedSpeakerIds,
+      );
       return {
         id: subtitle.id || String(index + 1),
         startMs,
         endMs,
         source,
         target: subtitle.targetContent ?? '',
-        ...(speakerIds?.length ? { speakerIds: [...speakerIds] } : {}),
+        ...(normalizedSpeakerIds.length
+          ? {
+              speakerIds: normalizedSpeakerIds,
+              primarySpeakerId,
+            }
+          : {}),
       };
     }),
   };

package.json

@@ -38,6 +38,7 @@
     "test:custom-languages": "tsc scripts/test-custom-languages.ts --outDir node_modules/.cache/custom-language-tests --module commonjs --moduleResolution node --target es2019 --esModuleInterop --skipLibCheck && node node_modules/.cache/custom-language-tests/scripts/test-custom-languages.js",
     "test:refine": "tsc scripts/test-refine-units.ts --outDir node_modules/.cache/refine-tests --module commonjs --moduleResolution node --target es2019 --esModuleInterop --skipLibCheck --resolveJsonModule && node node_modules/.cache/refine-tests/scripts/test-refine-units.js",
     "test:speaker-diarization": "node scripts/test-speaker-diarization-config.cjs && tsc scripts/test-speaker-diarization.ts --outDir node_modules/.cache/speaker-diarization-tests --module commonjs --moduleResolution node --target es2019 --esModuleInterop --skipLibCheck && node node_modules/.cache/speaker-diarization-tests/scripts/test-speaker-diarization.js",
+    "test:proofread-speakers": "tsc scripts/test-proofread-speakers.ts --outDir node_modules/.cache/proofread-speaker-tests --module commonjs --moduleResolution node --target es2019 --esModuleInterop --skipLibCheck && node node_modules/.cache/proofread-speaker-tests/scripts/test-proofread-speakers.js",
     "test:manuscript": "tsc scripts/test-manuscript-matching.ts --outDir node_modules/.cache/manuscript-tests --module commonjs --moduleResolution node --target es2022 --esModuleInterop --skipLibCheck --resolveJsonModule && node node_modules/.cache/manuscript-tests/scripts/test-manuscript-matching.js",
     "longgap:gen": "tsc scripts/longgap/gen-audio.ts --outDir node_modules/.cache/longgap --module commonjs --moduleResolution node --target es2019 --esModuleInterop --skipLibCheck --resolveJsonModule && node node_modules/.cache/longgap/scripts/longgap/gen-audio.js",
     "test:longgap": "tsc scripts/longgap/run.ts --outDir node_modules/.cache/longgap --module commonjs --moduleResolution node --target es2019 --esModuleInterop --skipLibCheck --resolveJsonModule && node node_modules/.cache/longgap/scripts/longgap/run.js",