feat: 集成 SenseVoice 字幕引擎，支持多模型选择

.gitignore

@@ -34,14 +34,21 @@ vite.config.js
 vite.config.d.ts
 
 # Native capture build artifacts
-electron/native/wgc-capture/build/
-electron/native/cursor-monitor/build/
-electron/native/gpu-export-probe/build/
-electron/native/nvidia-cuda-compositor/build/
-electron/native/bin/*/whisper-*
-electron/native/bin/*/whisper-runtime.json
-
-# Local debug helpers
-tmp-*.ps1
-.tmp-*.ps1
-gpu-export-probe.mp4
+electron/native/wgc-capture/build/
+electron/native/cursor-monitor/build/
+electron/native/gpu-export-probe/build/
+electron/native/nvidia-cuda-compositor/build/
+electron/native/bin/*/whisper-*
+electron/native/bin/*/whisper-runtime.json
+
+# Bundled caption models (downloaded at postinstall)
+/models/
+
+# Local debug helpers
+tmp-*.ps1
+.tmp-*.ps1
+gpu-export-probe.mp4
+
+# 开发日志内部版（不公开）
+dev_docs/articles/internal/
+dev_docs/.drafts/

dev_docs/INDEX.md

@@ -0,0 +1,14 @@
+# cc-Recordly-cn — 开发日志
+> Build in public 系列
+
+## 文章
+| 日期 | 标题 | 类型 |
+|------|------|------|
+
+## 命令
+```bash
+journal push    # 生成草稿
+journal status  # 查看状态
+journal sync    # 同步到全局
+journal search  # 全局搜索
+```

electron-builder.json5

@@ -36,6 +36,10 @@
     {
       "from": "public/wallpapers",
       "to": "assets/wallpapers"
+    },
+    {
+      "from": "models",
+      "to": "assets/models"
     }
   ],
   "publish": [
@@ -53,15 +57,23 @@
     "entitlements": "build/entitlements.mac.plist",
     "entitlementsInherit": "build/entitlements.mac.inherit.plist",
     "target": [
-      {
-        "target": "dmg",
-        "arch": ["x64", "arm64"]
-      },
-      {
-        "target": "zip",
-        "arch": ["x64", "arm64"]
-      }
+        {
+            "target": "dmg",
+            "arch": ["x64", "arm64"]
+        },
+        {
+            "target": "zip",
+            "arch": ["x64", "arm64"]
+        }
     ],
+    "entitlements": "build/entitlements.mac.plist",
+    "entitlementsInherit": "build/entitlements.mac.inherit.plist",
+	"target": [
+		{
+			"target": "dir",
+			"arch": ["arm64"]
+		}
+	],
     "icon": "icons/icons/mac/icon.icns",
     "artifactName": "${productName}-${arch}.${ext}",
     "extendInfo": {

electron/electron-env.d.ts

@@ -673,21 +673,32 @@ interface Window {
 			canceled?: boolean;
 			error?: string;
 		}>;
-		getWhisperSmallModelStatus: () => Promise<{
+		// ── Multi-model API ─────────────────────────────────────────────
+		getAvailableModels: () => Promise<
+			Array<{
+				id: string;
+				name: string;
+				engine: string;
+				sizeLabel?: string;
+				languages: string[];
+				description: string;
+			}>
+		>;
+		getModelStatus: (modelId: string) => Promise<{
 			success: boolean;
 			exists: boolean;
 			path?: string | null;
 			error?: string;
 		}>;
-		downloadWhisperSmallModel: () => Promise<{
+		downloadModel: (modelId: string) => Promise<{
 			success: boolean;
 			path?: string;
-			alreadyDownloaded?: boolean;
 			error?: string;
 		}>;
-		deleteWhisperSmallModel: () => Promise<{ success: boolean; error?: string }>;
-		onWhisperSmallModelDownloadProgress: (
+		deleteModel: (modelId: string) => Promise<{ success: boolean; error?: string }>;
+		onModelDownloadProgress: (
 			callback: (state: {
+				modelId: string;
 				status: "idle" | "downloading" | "downloaded" | "error";
 				progress: number;
 				path?: string | null;
@@ -698,6 +709,7 @@ interface Window {
 			videoPath: string;
 			whisperExecutablePath?: string;
 			whisperModelPath: string;
+			modelId?: string;
 			language?: string;
 		}) => Promise<{
 			success: boolean;

electron/ipc/captions/engine.ts

@@ -0,0 +1,55 @@
+import type { CaptionModel } from "./models";
+
+/**
+ * Result from caption generation.
+ */
+export interface CaptionWordPayload {
+	text: string;
+	startMs: number;
+	endMs: number;
+}
+
+export interface CaptionCuePayload {
+	id: string;
+	text: string;
+	startMs: number;
+	endMs: number;
+	words?: CaptionWordPayload[];
+}
+
+export interface GenerateCaptionResult {
+	success: boolean;
+	cues: CaptionCuePayload[];
+	message?: string;
+	error?: string;
+}
+
+/**
+ * Options passed to an engine for caption generation.
+ */
+export interface GenerateCaptionOptions {
+	/** Path to the pre-extracted 16kHz mono WAV file */
+	audioPath: string;
+	/** Absolute path to the downloaded model file */
+	modelPath: string;
+	/** The model metadata */
+	model: CaptionModel;
+	/** Language hint (BCP-47 code or "auto") */
+	language?: string;
+	/** Temp directory for intermediate files */
+	tempDir: string;
+}
+
+/**
+ * Abstract engine interface.
+ * Each engine (whisper, sensevoice) implements this.
+ */
+export interface CaptionEngine {
+	readonly engineType: string;
+
+	/**
+	 * Run caption generation.
+	 * Receives fully resolved paths — no model resolution needed.
+	 */
+	generate(options: GenerateCaptionOptions): Promise<GenerateCaptionResult>;
+}

electron/ipc/captions/generate.ts

@@ -96,6 +96,13 @@ export async function resolveCaptionAudioCandidates(videoPath: string) {
 		candidates.push({ path: normalizedCandidatePath, label });
 	};
 
+	// Companion audio (mic/system) should be tried FIRST - they have the best audio quality
+	const videoDir = path.dirname(videoPath);
+	const videoExt = path.extname(videoPath);
+	const videoBase = path.basename(videoPath, videoExt);
+	pushCandidate(path.join(videoDir, `${videoBase}.mic.wav`), "microphone");
+	pushCandidate(path.join(videoDir, `${videoBase}.system.wav`), "system audio");
+
 	pushCandidate(videoPath, "recording");
 
 	const requestedRecordingSession = await resolveRecordingSession(videoPath);
@@ -190,8 +197,47 @@ export async function generateAutoCaptionsFromVideo(options: {
 	videoPath: string;
 	whisperExecutablePath?: string;
 	whisperModelPath: string;
+	modelId?: string;
 	language?: string;
 }) {
+	// ── Engine dispatch: SenseVoice ──────────────────────────────────────
+	if (options.modelId) {
+		const { getModelById } = await import("./models");
+		const model = getModelById(options.modelId);
+		if (model?.engine === "sensevoice") {
+			const { SenseVoiceEngine } = await import("./sensevoice");
+			const engine = new SenseVoiceEngine();
+			const ffmpegPath = getFfmpegBinaryPath();
+			const tempDir = app.getPath("temp");
+			const wavPath = path.join(tempDir, `sensevoice-${Date.now()}.wav`);
+
+			// Use the same audio extraction as whisper path
+			const audioSource = await extractCaptionAudioSource({
+				videoPath: options.videoPath,
+				ffmpegPath,
+				wavPath,
+			});
+
+			const result = await engine.generate({
+				audioPath: wavPath,
+				modelPath: options.whisperModelPath,
+				model,
+				language: options.language,
+				tempDir,
+			});
+
+			if (!result.success) {
+				await fs.rm(wavPath, { force: true }).catch(() => undefined);
+				throw new Error(result.error || "SenseVoice caption generation failed.");
+			}
+
+			await fs.rm(wavPath, { force: true }).catch(() => undefined);
+
+			return { success: true, cues: result.cues, audioSourceLabel: audioSource.label };
+		}
+	}
+
+	// ── Engine: whisper.cpp (default) ────────────────────────────────────
 	const ffmpegPath = getFfmpegBinaryPath();
 	const normalizedVideoPath = normalizeVideoSourcePath(options.videoPath);
 	if (!normalizedVideoPath) {
@@ -292,6 +338,7 @@ export async function generateAutoCaptionsFromVideo(options: {
 		}
 
 		return {
+			success: true,
 			cues: cuesToReturn,
 			audioSourceLabel: audioSource.label,
 		};