增加语音对话功能

2026-08-29 22:13:55 +08:00
parent 1ffa069151
commit 3a2438f566
51 changed files with 6207 additions and 3 deletions
@@ -0,0 +1,93 @@
+import { joinPath } from "../../platform/paths";
+import {
+  defaultVoiceForLang,
+  removeIfExists,
+  resolveExecutable,
+  writeTempTextFile,
+} from "../../platform/executable";
+import { runCommand } from "../../platform/subprocess";
+import type { TtsProvider, TtsSynthesizeOptions, TtsResult } from "../types";
+import { TtsError } from "../types";
+import { getPref } from "../../../../../utils/prefs";
+
+async function estimateWavDurationMs(path: string): Promise<number> {
+  try {
+    const stat = await IOUtils.stat(path);
+    return Math.max(1000, Math.floor((stat.size / 32000) * 1000));
+  } catch {
+    return 0;
+  }
+}
+
+export function createWindowsSapiProvider(): TtsProvider {
+  return {
+    id: "system-win",
+    label: "Windows SAPI",
+    kind: "local",
+    supportsTimestamps: false,
+    async synthesize(options: TtsSynthesizeOptions): Promise<TtsResult> {
+      const powershell = await resolveExecutable("powershell");
+      const voice =
+        options.voice ||
+        (getPref("ttsVoice") as string) ||
+        defaultVoiceForLang(options.lang);
+
+      const outWav = options.outputPath.endsWith(".wav")
+        ? options.outputPath
+        : options.outputPath.replace(/\.\w+$/, ".wav");
+
+      const textPath = await writeTempTextFile(options.text, "sapi-text");
+      const scriptPath = joinPath(
+        PathUtils.tempDir,
+        `chatpapers-sapi-${Date.now()}.ps1`,
+      );
+
+      const psScript = [
+        "Add-Type -AssemblyName System.Speech",
+        "$text = Get-Content -LiteralPath $args[0] -Raw -Encoding UTF8",
+        "$out = $args[1]",
+        "$voice = $args[2]",
+        "$synth = New-Object System.Speech.Synthesis.SpeechSynthesizer",
+        "if ($voice) { try { $synth.SelectVoice($voice) } catch {} }",
+        "$synth.SetOutputToWaveFile($out)",
+        "$synth.Speak($text)",
+        "$synth.Dispose()",
+      ].join("\n");
+
+      await IOUtils.writeUTF8(scriptPath, psScript);
+
+      try {
+        const result = await runCommand(powershell, [
+          "-NoProfile",
+          "-ExecutionPolicy",
+          "Bypass",
+          "-File",
+          scriptPath,
+          textPath,
+          outWav,
+          voice,
+        ]);
+
+        if (result.exitCode !== 0) {
+          throw new TtsError(
+            `PowerShell SAPI 退出码 ${result.exitCode}`,
+            "runtime",
+          );
+        }
+        if (!(await IOUtils.exists(outWav))) {
+          throw new TtsError("SAPI 未生成音频文件", "runtime");
+        }
+
+        return {
+          audioPath: outWav,
+          timestamps: [],
+          durationMs: await estimateWavDurationMs(outWav),
+          format: "wav",
+        };
+      } finally {
+        await removeIfExists(textPath);
+        await removeIfExists(scriptPath);
+      }
+    },
+  };
+}