11

scripts/nodejs/pipeline_native.js

@@ -653,6 +653,136 @@ async function localAsr(audioPath, info) {
   });
 }
 
+const MIN_ENROLLMENT_SAMPLE_RATE = 16000;
+const VOICE_ENROLL_URL =
+  "https://dashscope.aliyuncs.com/api/v1/services/audio/tts/customization";
+
+async function ensureEnrollmentAudio(filePath) {
+  const ffmpeg = requireFfmpeg();
+  const normalized = (p) => (process.platform === "win32" ? p.replace(/\\/g, "/") : p);
+  let sampleRate = 0;
+  let channels = 0;
+  try {
+    const ffprobe = locateFfmpeg()?.replace(/ffmpeg(\.exe)?$/i, "ffprobe$1");
+    if (ffprobe && fs.existsSync(ffprobe)) {
+      const stdout = execFileSync(
+        ffprobe,
+        [
+          "-v",
+          "error",
+          "-select_streams",
+          "a:0",
+          "-show_entries",
+          "stream=sample_rate,channels",
+          "-of",
+          "json",
+          filePath,
+        ],
+        { encoding: "utf8", windowsHide: true },
+      );
+      const probe = JSON.parse(stdout || "{}");
+      const stream = (probe.streams && probe.streams[0]) || {};
+      sampleRate = parseInt(stream.sample_rate || "0", 10) || 0;
+      channels = parseInt(stream.channels || "0", 10) || 0;
+    }
+  } catch {
+    /* 转码兜底 */
+  }
+  const ext = path.extname(filePath).toLowerCase();
+  const needsConvert =
+    !sampleRate ||
+    sampleRate < MIN_ENROLLMENT_SAMPLE_RATE ||
+    channels !== 1 ||
+    ext !== ".wav";
+  if (!needsConvert) return filePath;
+
+  const parsed = path.parse(filePath);
+  const outputPath = path.join(parsed.dir, `${parsed.name}_enroll_16k.wav`);
+  await runCommand(ffmpeg, [
+    "-y",
+    "-i",
+    normalized(filePath),
+    "-vn",
+    "-acodec",
+    "pcm_s16le",
+    "-ar",
+    String(MIN_ENROLLMENT_SAMPLE_RATE),
+    "-ac",
+    "1",
+    normalized(outputPath),
+  ]);
+  return outputPath;
+}
+
+function audioFileToDataUri(filePath) {
+  const ext = path.extname(filePath).toLowerCase();
+  const mimeTypes = {
+    ".wav": "audio/wav",
+    ".mp3": "audio/mpeg",
+    ".m4a": "audio/m4a",
+    ".flac": "audio/flac",
+    ".ogg": "audio/ogg",
+    ".aac": "audio/aac",
+    ".webm": "audio/webm",
+  };
+  const mime = mimeTypes[ext] || "audio/wav";
+  const buf = fs.readFileSync(filePath);
+  return `data:${mime};base64,${buf.toString("base64")}`;
+}
+
+/**
+ * 阿里云声音复刻（对齐 Electron aliyun:cosyvoice:enroll）
+ * @param {{ apiKey: string, audioPath: string, name: string }} opts
+ */
+async function qwenVoiceEnroll(opts) {
+  const apiKey = String(opts.apiKey || "").trim();
+  const audioPath = String(opts.audioPath || "").trim();
+  const name = String(opts.name || "").trim();
+  if (!apiKey) return { success: false, error: "缺少 DashScope apiKey" };
+  if (!audioPath || !fs.existsSync(audioPath)) {
+    return { success: false, error: "参考音频文件不存在" };
+  }
+  if (!name) return { success: false, error: "请先输入音色名称" };
+
+  try {
+    const prepared = await ensureEnrollmentAudio(audioPath);
+    const dataUri = audioFileToDataUri(prepared);
+    let prefix = name.replace(/[^a-zA-Z0-9_]/g, "");
+    if (!prefix) prefix = "custom";
+    else if (prefix.length > 10) prefix = prefix.substring(0, 10);
+
+    const payload = {
+      model: "qwen-voice-enrollment",
+      input: {
+        action: "create",
+        target_model: "qwen3-tts-vc-2026-01-22",
+        preferred_name: prefix,
+        audio: { data: dataUri },
+      },
+    };
+
+    const data = await fetchJson(VOICE_ENROLL_URL, {
+      method: "POST",
+      headers: {
+        Authorization: `Bearer ${apiKey}`,
+        "Content-Type": "application/json",
+      },
+      body: JSON.stringify(payload),
+    });
+
+    const voiceId = data?.output?.voice || data?.output?.voice_id;
+    if (!voiceId) {
+      return {
+        success: false,
+        error: data?.message || data?.code || "复刻响应格式错误",
+      };
+    }
+    return { success: true, voiceId: String(voiceId) };
+  } catch (e) {
+    return { success: false, error: e.message || String(e) };
+  }
+}
+
 async function openaiChat(req) {
   let apiUrl = (req.apiUrl || "").replace(/\/$/, "");
   if (!apiUrl.includes("/chat/completions")) {
@@ -698,5 +828,6 @@ module.exports = {
   resolveTtsModel,
   splitTextForTts,
   qwenTtsSynthesize,
+  qwenVoiceEnroll,
   locateFfmpeg,
 };