Use Windows system TTS for local mode

src/discord-main.ts

@@ -18,6 +18,7 @@ import { Logger } from "./logger.js";
 import { LocalFasterWhisperSttService } from "./services/local-stt.js";
 import { LocalKokoroTtsService } from "./services/local-tts.js";
 import { OllamaLlmService } from "./services/ollama-llm.js";
+import { WindowsSystemTtsService } from "./services/windows-system-tts.js";
 
 export async function runDiscordBot(config: DiscordRuntimeConfig, logger: Logger): Promise<void> {
   const commands = [
@@ -38,7 +39,10 @@ export async function runDiscordBot(config: DiscordRuntimeConfig, logger: Logger
   });
 
   const stt = new LocalFasterWhisperSttService(config, logger);
-  const tts = new LocalKokoroTtsService(config, logger);
+  const tts =
+    process.platform === "win32"
+      ? new WindowsSystemTtsService(config.LOCAL_TTS_SPEED)
+      : new LocalKokoroTtsService(config, logger);
   const llm = new OllamaLlmService(config);
   const sessions = new Map<string, GuildVoiceSession>();
 

src/local-main.ts

@@ -8,6 +8,7 @@ import { requireFfmpegPath } from "./audio/ffmpeg-path.js";
 import { LocalFasterWhisperSttService } from "./services/local-stt.js";
 import { LocalKokoroTtsService } from "./services/local-tts.js";
 import { OllamaLlmService } from "./services/ollama-llm.js";
+import { WindowsSystemTtsService } from "./services/windows-system-tts.js";
 
 export async function printLocalAudioDevices(): Promise<void> {
   if (process.platform === "win32") {
@@ -68,7 +69,10 @@ export async function printLocalAudioDevices(): Promise<void> {
 
 export async function runLocalAssistant(config: AssistantRuntimeConfig, logger: Logger): Promise<void> {
   const stt = new LocalFasterWhisperSttService(config, logger);
-  const tts = new LocalKokoroTtsService(config, logger);
+  const tts =
+    process.platform === "win32"
+      ? new WindowsSystemTtsService(config.LOCAL_TTS_SPEED)
+      : new LocalKokoroTtsService(config, logger);
   const llm = new OllamaLlmService(config);
 
   await stt.warmup();

src/services/python-json-worker.ts

@@ -117,6 +117,7 @@ export class PythonJsonWorker {
         HF_HOME: cachePath,
         TRANSFORMERS_CACHE: cachePath,
         PYTHONIOENCODING: "utf-8",
+        HF_HUB_DISABLE_SYMLINKS_WARNING: "1",
         BOT_DEFAULT_LANGUAGE: this.config.BOT_DEFAULT_LANGUAGE,
         ...this.workerEnv,
       },
@@ -132,6 +133,9 @@ export class PythonJsonWorker {
 
       let payload: WorkerResponse;
       try {
+        if (!line.startsWith("{")) {
+          return;
+        }
         payload = JSON.parse(line) as WorkerResponse;
       } catch (error) {
         this.logger.warn(`${this.label} stdout parse failed`, error);

src/services/windows-system-tts.ts

@@ -0,0 +1,126 @@
+import { spawn } from "node:child_process";
+import { createReadStream } from "node:fs";
+import { unlink } from "node:fs/promises";
+import os from "node:os";
+import path from "node:path";
+
+import prism from "prism-media";
+
+import { resolveFfmpegPath } from "../audio/ffmpeg-path.js";
+import type { PreparedSpeechAudio, TtsService } from "./tts.js";
+
+function escapePowerShellSingleQuoted(text: string): string {
+  return text.replace(/'/g, "''");
+}
+
+function toSpeechRate(speed: number): number {
+  const mapped = Math.round((speed - 1) * 8);
+  return Math.max(-10, Math.min(10, mapped));
+}
+
+export class WindowsSystemTtsService implements TtsService {
+  constructor(private readonly speed: number) {
+    const resolvedFfmpegPath = resolveFfmpegPath();
+    if (resolvedFfmpegPath && !process.env.FFMPEG_PATH) {
+      process.env.FFMPEG_PATH = resolvedFfmpegPath;
+    }
+  }
+
+  async warmup(): Promise<void> {
+    return;
+  }
+
+  async preparePlayback(text: string, signal?: AbortSignal): Promise<PreparedSpeechAudio> {
+    const tempPath = path.join(os.tmpdir(), `realtime-voice-bot-tts-${Date.now()}.wav`);
+    const rate = toSpeechRate(this.speed);
+    const script = [
+      "Add-Type -AssemblyName System.Speech;",
+      "$synth = New-Object System.Speech.Synthesis.SpeechSynthesizer;",
+      "$koVoice = $synth.GetInstalledVoices() | Where-Object { $_.VoiceInfo.Culture.Name -like 'ko*' } | Select-Object -First 1;",
+      "if ($koVoice) { $synth.SelectVoice($koVoice.VoiceInfo.Name) }",
+      `$synth.Rate = ${rate};`,
+      `$synth.SetOutputToWaveFile('${escapePowerShellSingleQuoted(tempPath)}');`,
+      `$synth.Speak('${escapePowerShellSingleQuoted(text)}');`,
+      "$synth.Dispose();",
+    ].join(" ");
+
+    await new Promise<void>((resolve, reject) => {
+      const child = spawn("powershell", ["-NoProfile", "-Command", script], {
+        stdio: ["ignore", "ignore", "pipe"],
+      });
+
+      let stderr = "";
+      child.stderr.on("data", (chunk: Buffer) => {
+        stderr += chunk.toString();
+      });
+
+      signal?.addEventListener(
+        "abort",
+        () => {
+          if (!child.killed) {
+            child.kill("SIGKILL");
+          }
+        },
+        { once: true },
+      );
+
+      child.on("exit", (code) => {
+        if (signal?.aborted) {
+          reject(new Error("tts aborted"));
+          return;
+        }
+        if (code === 0) {
+          resolve();
+          return;
+        }
+        reject(new Error(stderr.trim() || `powershell tts exited with code ${code ?? "null"}`));
+      });
+      child.on("error", reject);
+    }).catch(async (error) => {
+      await unlink(tempPath).catch(() => null);
+      throw error;
+    });
+
+    const input = createReadStream(tempPath);
+    const ffmpeg = new prism.FFmpeg({
+      args: [
+        "-analyzeduration",
+        "0",
+        "-loglevel",
+        "0",
+        "-i",
+        tempPath,
+        "-f",
+        "s16le",
+        "-ar",
+        "48000",
+        "-ac",
+        "2",
+        "pipe:1",
+      ],
+    });
+
+    signal?.addEventListener(
+      "abort",
+      () => {
+        input.destroy();
+        ffmpeg.destroy();
+        void unlink(tempPath).catch(() => null);
+      },
+      { once: true },
+    );
+
+    return {
+      stream: ffmpeg,
+      dispose: () => {
+        input.destroy();
+        ffmpeg.destroy();
+        void unlink(tempPath).catch(() => null);
+      },
+    };
+  }
+
+  async destroy(): Promise<void> {
+    return;
+  }
+}

src/setup-local-ai.ts

@@ -82,11 +82,16 @@ async function main(): Promise<void> {
   const cachePath = resolveLocalAiCachePath(config);
   const ttsModelPath = resolveLocalAiTtsModelPath(config);
   const ttsVoicesPath = resolveLocalAiTtsVoicesPath(config);
-  const requirementsPath = path.resolve(process.cwd(), "python", "requirements.txt");
+  const requirementsPath = path.resolve(
+    process.cwd(),
+    "python",
+    process.platform === "win32" ? "requirements-windows.txt" : "requirements.txt",
+  );
   const baseEnv = {
     HF_HOME: cachePath,
     TRANSFORMERS_CACHE: cachePath,
     PYTHONIOENCODING: "utf-8",
+    HF_HUB_DISABLE_SYMLINKS_WARNING: "1",
   };
 
   await mkdir(cachePath, { recursive: true });
@@ -106,9 +111,11 @@ async function main(): Promise<void> {
   console.log("로컬 AI 의존성 설치를 시작합니다.");
   await run(venvPython, ["-m", "pip", "install", "--upgrade", "pip", "setuptools", "wheel"], baseEnv);
   await run(venvPython, ["-m", "pip", "install", "-r", requirementsPath], baseEnv);
-  console.log("로컬 TTS 모델 파일을 확인합니다.");
-  await ensureDownload(KOKORO_MODEL_URL, ttsModelPath);
-  await ensureDownload(KOKORO_VOICES_URL, ttsVoicesPath);
+  if (process.platform !== "win32") {
+    console.log("로컬 TTS 모델 파일을 확인합니다.");
+    await ensureDownload(KOKORO_MODEL_URL, ttsModelPath);
+    await ensureDownload(KOKORO_VOICES_URL, ttsVoicesPath);
+  }
 
   console.log("설치가 끝났습니다.");
   console.log("다음 순서:");