dsh-xiaomi-tts 2.3.2 → 2.3.3

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (40) hide show
  1. package/README.en.md +84 -6
  2. package/README.md +19 -10
  3. package/assets/menu.png +0 -0
  4. package/assets/preset.png +0 -0
  5. package/assets/social.png +0 -0
  6. package/assets/voice-avatars/baihua.webp +0 -0
  7. package/assets/voice-avatars/bingtang.webp +0 -0
  8. package/assets/voice-avatars/chloe.webp +0 -0
  9. package/assets/voice-avatars/dean.webp +0 -0
  10. package/assets/voice-avatars/mia.webp +0 -0
  11. package/assets/voice-avatars/milo.webp +0 -0
  12. package/assets/voice-avatars/moli.webp +0 -0
  13. package/assets/voice-avatars/suda.webp +0 -0
  14. package/assets/voice-presets/energetic-girl.webp +0 -0
  15. package/assets/voice-presets/liang-wenfeng.webp +0 -0
  16. package/cordis.patch.yml +4 -1
  17. package/lib/client/built-in-voice-picker.d.ts +10 -0
  18. package/lib/client/built-in-voice-picker.d.ts.map +1 -0
  19. package/lib/client/conversation.d.ts.map +1 -1
  20. package/lib/client/live-speech-controller.d.ts +6 -0
  21. package/lib/client/live-speech-controller.d.ts.map +1 -1
  22. package/lib/client/localization.d.ts +10 -4
  23. package/lib/client/localization.d.ts.map +1 -1
  24. package/lib/client/pcm-audio-queue.d.ts +7 -0
  25. package/lib/client/pcm-audio-queue.d.ts.map +1 -1
  26. package/lib/client/settings-card.d.ts.map +1 -1
  27. package/lib/client/settings-scope.d.ts.map +1 -1
  28. package/lib/client/styles.d.ts +1 -1
  29. package/lib/client/styles.d.ts.map +1 -1
  30. package/lib/client/voice-design-picker.d.ts.map +1 -1
  31. package/lib/client.js +559 -224
  32. package/lib/client.js.map +1 -1
  33. package/lib/index.d.ts +8 -2
  34. package/lib/index.d.ts.map +1 -1
  35. package/lib/index.js +121 -33
  36. package/lib/{shared-9qFFYsWe.js → shared-Cf9qZwUl.js} +104 -32
  37. package/lib/shared.d.ts +82 -26
  38. package/lib/shared.d.ts.map +1 -1
  39. package/lib/shared.js +2 -2
  40. package/package.json +1 -1
package/lib/index.d.ts CHANGED
@@ -16,11 +16,14 @@ export declare const Config: z<Schemastery.ObjectS<{
16
16
  voiceDesignPrompt: z<string, string>;
17
17
  voiceDesignCustomPrompt: z<string, string>;
18
18
  presetStylePrompt: z<string, string>;
19
- format: z<"mp3" | "wav", "mp3" | "wav">;
19
+ format: z<"pcm" | "mp3" | "wav", "pcm" | "mp3" | "wav">;
20
20
  autoPlay: z<boolean, boolean>;
21
21
  instruction: z<string, string>;
22
22
  maxTextLength: z<number, number>;
23
23
  requestTimeoutMs: z<number, number>;
24
+ maxMp3AudioBytes: z<number, number>;
25
+ maxWavAudioBytes: z<number, number>;
26
+ maxPausedPcmBytes: z<number, number>;
24
27
  }>, Schemastery.ObjectT<{
25
28
  enabled: z<boolean, boolean>;
26
29
  apiKey: z<string, string>;
@@ -30,11 +33,14 @@ export declare const Config: z<Schemastery.ObjectS<{
30
33
  voiceDesignPrompt: z<string, string>;
31
34
  voiceDesignCustomPrompt: z<string, string>;
32
35
  presetStylePrompt: z<string, string>;
33
- format: z<"mp3" | "wav", "mp3" | "wav">;
36
+ format: z<"pcm" | "mp3" | "wav", "pcm" | "mp3" | "wav">;
34
37
  autoPlay: z<boolean, boolean>;
35
38
  instruction: z<string, string>;
36
39
  maxTextLength: z<number, number>;
37
40
  requestTimeoutMs: z<number, number>;
41
+ maxMp3AudioBytes: z<number, number>;
42
+ maxWavAudioBytes: z<number, number>;
43
+ maxPausedPcmBytes: z<number, number>;
38
44
  }>>;
39
45
  export type Config = ReturnType<typeof Config>;
40
46
  /** Register the TTS settings and same-origin synthesis route. */
@@ -1 +1 @@
1
- {"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,OAAO,EAAE,MAAM,qBAAqB,CAAA;AAQlD,OAAO,CAAC,MAAM,0BAA0B,CAAA;AAYxC,gCAAgC;AAChC,eAAO,MAAM,IAAI,oBAAoB,CAAA;AAErC,6CAA6C;AAC7C,eAAO,MAAM,MAAM,UAAgB,CAAA;AAEnC,uDAAuD;AACvD,eAAO,MAAM,kCAAkC,uDAA4C,CAAA;AAE3F,sCAAsC;AACtC,eAAO,MAAM,MAAM;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAcjB,CAAA;AAEF,MAAM,MAAM,MAAM,GAAG,UAAU,CAAC,OAAO,MAAM,CAAC,CAAA;AA4O9C,iEAAiE;AACjE,wBAAgB,KAAK,CAAC,GAAG,EAAE,OAAO,EAAE,MAAM,EAAE,MAAM,GAAG,IAAI,CA4WxD"}
1
+ {"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,OAAO,EAAE,MAAM,qBAAqB,CAAA;AAQlD,OAAO,CAAC,MAAM,0BAA0B,CAAA;AAYxC,gCAAgC;AAChC,eAAO,MAAM,IAAI,oBAAoB,CAAA;AAErC,6CAA6C;AAC7C,eAAO,MAAM,MAAM,UAAgB,CAAA;AAEnC,uDAAuD;AACvD,eAAO,MAAM,kCAAkC,uDAA4C,CAAA;AAE3F,sCAAsC;AACtC,eAAO,MAAM,MAAM;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAiBjB,CAAA;AAEF,MAAM,MAAM,MAAM,GAAG,UAAU,CAAC,OAAO,MAAM,CAAC,CAAA;AAiT9C,iEAAiE;AACjE,wBAAgB,KAAK,CAAC,GAAG,EAAE,OAAO,EAAE,MAAM,EAAE,MAAM,GAAG,IAAI,CA2YxD"}
package/lib/index.js CHANGED
@@ -1,4 +1,4 @@
1
- import { C as prepareTtsText, a as TTS_API_KEY_STATUS_ROUTE, b as isNewerTtsVersion, c as TTS_ROUTE, d as TTS_UNINSTALL_ROUTE, f as TTS_UPDATE_ROUTE, g as TTS_VOICE_DESIGN_PRESETS, h as TTS_VOICE_DESIGN_ASSET_ROUTE, l as TTS_SETTINGS_NAMESPACE, n as DEFAULT_TTS_SETTINGS, o as TTS_FORMATS, p as TTS_VERSION, s as TTS_MODELS, u as TTS_STREAM_ROUTE, w as resolveTtsBaseURL, x as isSupportedTtsApiKey } from "./shared-9qFFYsWe.js";
1
+ import { A as resolveTtsBaseURL, D as isSupportedTtsApiKey, E as isNewerTtsVersion, N as strictBase64DecodedLength, S as TTS_VOICE_PRESETS, _ as TTS_VERSION, a as DEFAULT_TTS_SETTINGS, b as TTS_VOICE_DESIGN_ASSET_ROUTE, c as TTS_API_KEY_STATUS_ROUTE, d as TTS_MODELS, f as TTS_ROUTE, g as TTS_UPDATE_ROUTE, h as TTS_UNINSTALL_ROUTE, k as prepareTtsText, l as TTS_AUDIO_RESPONSE_JSON_OVERHEAD_BYTES, m as TTS_STREAM_ROUTE, p as TTS_SETTINGS_NAMESPACE, u as TTS_FORMATS, x as TTS_VOICE_DESIGN_PRESETS, y as TTS_VOICE_ASSET_ROUTE } from "./shared-Cf9qZwUl.js";
2
2
  import { createRequire } from "node:module";
3
3
  import { spawn } from "node:child_process";
4
4
  import { existsSync, readFileSync } from "node:fs";
@@ -33,7 +33,10 @@ const Config = z.object({
33
33
  autoPlay: z.boolean().default(DEFAULT_TTS_SETTINGS.autoPlay),
34
34
  instruction: z.string().default(DEFAULT_TTS_SETTINGS.instruction),
35
35
  maxTextLength: z.number().step(1).min(1).default(DEFAULT_TTS_SETTINGS.maxTextLength),
36
- requestTimeoutMs: z.number().step(1).min(1e3).default(DEFAULT_TTS_SETTINGS.requestTimeoutMs)
36
+ requestTimeoutMs: z.number().step(1).min(1e3).default(DEFAULT_TTS_SETTINGS.requestTimeoutMs),
37
+ maxMp3AudioBytes: z.number().step(1).min(1).default(DEFAULT_TTS_SETTINGS.maxMp3AudioBytes),
38
+ maxWavAudioBytes: z.number().step(1).min(1).default(DEFAULT_TTS_SETTINGS.maxWavAudioBytes),
39
+ maxPausedPcmBytes: z.number().step(1).min(1).default(DEFAULT_TTS_SETTINGS.maxPausedPcmBytes)
37
40
  });
38
41
  function requestMessages(options, text) {
39
42
  const context = options.model === "mimo-v2.5-tts-voicedesign" ? options.voiceDesignPrompt.trim() : [options.presetStylePrompt.trim(), options.instruction.trim()].filter((item) => item.length > 0).join("\n");
@@ -74,6 +77,61 @@ function apiErrorMessage(status, parsed) {
74
77
  const detail = typeof parsed?.error === "string" ? parsed.error : parsed?.error?.message ?? parsed?.message;
75
78
  return detail && detail.length > 0 ? detail : `Xiaomi MiMo TTS request failed (HTTP ${status})`;
76
79
  }
80
+ var CompleteAudioResponseError = class extends Error {
81
+ constructor(code) {
82
+ super(code);
83
+ this.code = code;
84
+ }
85
+ };
86
+ function completeAudioFormat(options) {
87
+ return options.format === "wav" ? "wav" : "mp3";
88
+ }
89
+ function completeAudioLimit(options, format) {
90
+ return format === "mp3" ? options.maxMp3AudioBytes : options.maxWavAudioBytes;
91
+ }
92
+ function completeAudioJsonLimit(audioLimit) {
93
+ return 4 * Math.ceil(audioLimit / 3) + TTS_AUDIO_RESPONSE_JSON_OVERHEAD_BYTES;
94
+ }
95
+ async function readLimitedXiaomiResponse(response, limit, controller) {
96
+ const contentLength = response.headers.get("content-length");
97
+ if (contentLength !== null && /^\d+$/u.test(contentLength) && Number(contentLength) > limit) {
98
+ controller.abort();
99
+ throw new CompleteAudioResponseError("xiaomi-response-too-large");
100
+ }
101
+ if (response.body === null) return void 0;
102
+ const reader = response.body.getReader();
103
+ const chunks = [];
104
+ let size = 0;
105
+ try {
106
+ while (true) {
107
+ const chunk = await reader.read();
108
+ if (chunk.done) break;
109
+ size += chunk.value.byteLength;
110
+ if (size > limit) {
111
+ controller.abort();
112
+ throw new CompleteAudioResponseError("xiaomi-response-too-large");
113
+ }
114
+ chunks.push(Buffer.from(chunk.value));
115
+ }
116
+ } finally {
117
+ reader.releaseLock();
118
+ }
119
+ if (chunks.length === 0) return void 0;
120
+ try {
121
+ return JSON.parse(Buffer.concat(chunks, size).toString("utf8"));
122
+ } catch {
123
+ return;
124
+ }
125
+ }
126
+ function decodeCompleteAudio(value, format, limit) {
127
+ const expectedLength = strictBase64DecodedLength(value);
128
+ if (expectedLength === null) throw new CompleteAudioResponseError("invalid-audio-base64");
129
+ if (expectedLength > limit) throw new CompleteAudioResponseError("audio-too-large");
130
+ const audio = Buffer.from(value, "base64");
131
+ if (audio.byteLength !== expectedLength) throw new CompleteAudioResponseError("invalid-audio-base64");
132
+ if (!(format === "wav" ? audio.byteLength >= 12 && audio.toString("ascii", 0, 4) === "RIFF" && audio.toString("ascii", 8, 12) === "WAVE" : audio.byteLength >= 2 && (audio.toString("ascii", 0, 3) === "ID3" || audio[0] === 255 && (audio[1] & 224) === 224))) throw new CompleteAudioResponseError("invalid-audio-format");
133
+ return audio;
134
+ }
77
135
  function webProfileDirectory() {
78
136
  return join(process.env.DSH_HOME?.trim() || join(homedir(), ".dsh"), "profiles", WEB_PROFILE_NAME);
79
137
  }
@@ -272,6 +330,9 @@ function apply(ctx, config) {
272
330
  const voicePresetAssets = new Map(TTS_VOICE_DESIGN_PRESETS.map((preset) => {
273
331
  return [`${TTS_VOICE_DESIGN_ASSET_ROUTE}/${preset.id}.webp`, readFileSync(new URL(`../assets/voice-presets/${preset.id}.webp`, import.meta.url))];
274
332
  }));
333
+ const voiceAssets = new Map(TTS_VOICE_PRESETS.map((preset) => {
334
+ return [`${TTS_VOICE_ASSET_ROUTE}/${preset.id}.webp`, readFileSync(new URL(`../assets/voice-avatars/${preset.id}.webp`, import.meta.url))];
335
+ }));
275
336
  installSettingsSection(ctx, XIAOMI_MIMO_TTS_SETTINGS_NAMESPACE, Config, config, {
276
337
  setSource(source) {
277
338
  current = source;
@@ -281,7 +342,7 @@ function apply(ctx, config) {
281
342
  const endpoint = `${normalizeBaseURL(value.baseURL)}/chat/completions`;
282
343
  if (!URL.canParse(endpoint)) throw new Error("baseURL must be a valid absolute URL");
283
344
  if (value.model === "mimo-v2.5-tts-voicedesign" && value.voiceDesignPrompt.trim().length === 0) throw new Error("voiceDesignPrompt is required when using mimo-v2.5-tts-voicedesign");
284
- if (value.format !== "mp3" && value.format !== "wav") throw new Error("format must be mp3 or wav");
345
+ if (!TTS_FORMATS.includes(value.format)) throw new Error("format must be pcm, mp3, or wav");
285
346
  }
286
347
  });
287
348
  ctx.effect(() => ctx.webServer.register({
@@ -388,6 +449,31 @@ function apply(ctx, config) {
388
449
  res.end(req.method === "HEAD" ? void 0 : asset);
389
450
  }
390
451
  }), "xiaomi-mimo-tts: voice preset assets");
452
+ ctx.effect(() => ctx.webServer.register({
453
+ kind: "prefix",
454
+ path: TTS_VOICE_ASSET_ROUTE,
455
+ handler(req, res) {
456
+ if (req.method !== "GET" && req.method !== "HEAD") {
457
+ res.statusCode = 405;
458
+ res.setHeader("allow", "GET, HEAD");
459
+ res.end();
460
+ return;
461
+ }
462
+ const pathname = new URL(req.url ?? "/", "http://localhost").pathname;
463
+ const asset = voiceAssets.get(pathname);
464
+ if (asset === void 0) {
465
+ res.statusCode = 404;
466
+ res.end();
467
+ return;
468
+ }
469
+ res.statusCode = 200;
470
+ res.setHeader("content-type", "image/webp");
471
+ res.setHeader("content-length", String(asset.byteLength));
472
+ res.setHeader("cache-control", "public, max-age=31536000, immutable");
473
+ res.setHeader("x-content-type-options", "nosniff");
474
+ res.end(req.method === "HEAD" ? void 0 : asset);
475
+ }
476
+ }), "xiaomi-mimo-tts: built-in voice assets");
391
477
  ctx.effect(() => ctx.webServer.register({
392
478
  kind: "exact",
393
479
  path: TTS_ROUTE,
@@ -422,8 +508,21 @@ function apply(ctx, config) {
422
508
  return;
423
509
  }
424
510
  const controller = new AbortController();
425
- const timeout = setTimeout(() => controller.abort(), options.requestTimeoutMs);
511
+ let abortReason = null;
512
+ const abort = (reason) => {
513
+ if (controller.signal.aborted) return;
514
+ abortReason = reason;
515
+ controller.abort();
516
+ };
517
+ const timeout = setTimeout(() => abort("timeout"), options.requestTimeoutMs);
518
+ const abortOnRequest = () => abort("client-disconnect");
519
+ const abortOnResponseClose = () => {
520
+ if (!res.writableEnded) abort("client-disconnect");
521
+ };
522
+ req.once("aborted", abortOnRequest);
523
+ res.once("close", abortOnResponseClose);
426
524
  try {
525
+ const format = completeAudioFormat(options);
427
526
  const endpoint = `${normalizeBaseURL(resolveTtsBaseURL(options.apiKey, options.baseURL))}/chat/completions`;
428
527
  const response = await fetch(endpoint, {
429
528
  method: "POST",
@@ -437,20 +536,16 @@ function apply(ctx, config) {
437
536
  body: JSON.stringify({
438
537
  model: options.model,
439
538
  messages: requestMessages(options, text),
440
- audio: options.model === "mimo-v2.5-tts-voicedesign" ? { format: options.format } : {
441
- format: options.format,
539
+ audio: options.model === "mimo-v2.5-tts-voicedesign" ? { format } : {
540
+ format,
442
541
  voice: options.voice
443
542
  },
444
543
  stream: false
445
544
  }),
446
545
  signal: controller.signal
447
546
  });
448
- let parsed;
449
- try {
450
- parsed = await response.json();
451
- } catch {
452
- parsed = void 0;
453
- }
547
+ const audioLimit = completeAudioLimit(options, format);
548
+ const parsed = await readLimitedXiaomiResponse(response, completeAudioJsonLimit(audioLimit), controller);
454
549
  if (!response.ok) {
455
550
  json(res, response.status, {
456
551
  error: "xiaomi-api-error",
@@ -466,36 +561,29 @@ function apply(ctx, config) {
466
561
  });
467
562
  return;
468
563
  }
469
- let audio;
470
- try {
471
- audio = Buffer.from(audioBase64, "base64");
472
- } catch {
473
- json(res, 502, {
474
- error: "invalid-audio-base64",
475
- message: "Xiaomi MiMo returned invalid Base64 audio data"
476
- });
477
- return;
478
- }
479
- if (audio.byteLength === 0) {
480
- json(res, 502, {
481
- error: "empty-audio",
482
- message: "Xiaomi MiMo returned an empty audio payload"
483
- });
484
- return;
485
- }
564
+ const audio = decodeCompleteAudio(audioBase64, format, audioLimit);
486
565
  res.statusCode = 200;
487
- res.setHeader("content-type", options.format === "mp3" ? "audio/mpeg" : "audio/wav");
566
+ res.setHeader("content-type", format === "mp3" ? "audio/mpeg" : "audio/wav");
488
567
  res.setHeader("content-length", String(audio.byteLength));
489
568
  res.setHeader("cache-control", "no-store");
490
569
  res.end(audio);
491
570
  } catch (error) {
492
- const aborted = controller.signal.aborted;
493
- json(res, aborted ? 504 : 502, {
494
- error: aborted ? "xiaomi-timeout" : "xiaomi-request-failed",
571
+ if (abortReason === "client-disconnect" || res.destroyed || res.writableEnded) return;
572
+ if (error instanceof CompleteAudioResponseError) {
573
+ json(res, 502, {
574
+ error: error.code,
575
+ message: error.message
576
+ });
577
+ return;
578
+ }
579
+ json(res, abortReason === "timeout" ? 504 : 502, {
580
+ error: abortReason === "timeout" ? "xiaomi-timeout" : "xiaomi-request-failed",
495
581
  message: error instanceof Error ? error.message : String(error)
496
582
  });
497
583
  } finally {
498
584
  clearTimeout(timeout);
585
+ req.off("aborted", abortOnRequest);
586
+ res.off("close", abortOnResponseClose);
499
587
  }
500
588
  }
501
589
  }), "xiaomi-mimo-tts: synthesis route");
@@ -10,7 +10,7 @@ const TTS_UNINSTALL_ROUTE = "/plugins/xiaomi-mimo-tts/uninstall";
10
10
  /** Same-origin route that checks the published npm version. */
11
11
  const TTS_UPDATE_ROUTE = "/plugins/xiaomi-mimo-tts/update";
12
12
  /** Keep the UI version visible without making the browser bundle load package.json. */
13
- const TTS_VERSION = "2.3.2";
13
+ const TTS_VERSION = "2.3.3";
14
14
  const SEMVER = /^(\d+)\.(\d+)\.(\d+)(?:-([0-9A-Za-z.-]+))?$/;
15
15
  /** Return true only when `candidate` is a strictly newer semantic version. */
16
16
  function isNewerTtsVersion(candidate, current) {
@@ -42,6 +42,8 @@ function isNewerTtsVersion(candidate, current) {
42
42
  const TTS_API_KEY_STATUS_ROUTE = "/plugins/xiaomi-mimo-tts/api-key-status";
43
43
  /** Same-origin prefix used by the Web client to load voice-design preset icons. */
44
44
  const TTS_VOICE_DESIGN_ASSET_ROUTE = "/plugins/xiaomi-mimo-tts/voice-presets";
45
+ /** Same-origin prefix used by the Web client to load official built-in voice avatars. */
46
+ const TTS_VOICE_ASSET_ROUTE = "/plugins/xiaomi-mimo-tts/voice-avatars";
45
47
  /** Default Xiaomi endpoint for Token Plan API keys. */
46
48
  const TOKEN_PLAN_TTS_BASE_URL = "https://token-plan-cn.xiaomimimo.com/v1";
47
49
  /** Supported built-in Xiaomi MiMo voices. */
@@ -55,69 +57,129 @@ const TTS_VOICES = [
55
57
  "Milo",
56
58
  "Dean"
57
59
  ];
60
+ /** Built-in voice metadata mirrored from Xiaomi MiMo's official selector. */
61
+ const TTS_VOICE_PRESETS = [
62
+ {
63
+ id: "bingtang",
64
+ value: "冰糖",
65
+ summary: "活泼少女"
66
+ },
67
+ {
68
+ id: "mia",
69
+ value: "Mia",
70
+ summary: "Lively girl"
71
+ },
72
+ {
73
+ id: "moli",
74
+ value: "茉莉",
75
+ summary: "知性女声"
76
+ },
77
+ {
78
+ id: "chloe",
79
+ value: "Chloe",
80
+ summary: "Witty Grace"
81
+ },
82
+ {
83
+ id: "suda",
84
+ value: "苏打",
85
+ summary: "阳光少年"
86
+ },
87
+ {
88
+ id: "milo",
89
+ value: "Milo",
90
+ summary: "Sunny boy"
91
+ },
92
+ {
93
+ id: "baihua",
94
+ value: "白桦",
95
+ summary: "成熟男声"
96
+ },
97
+ {
98
+ id: "dean",
99
+ value: "Dean",
100
+ summary: "Steady Gentle"
101
+ }
102
+ ];
58
103
  /** TTS models supported by this plugin. */
59
104
  const TTS_MODELS = ["mimo-v2.5-tts", "mimo-v2.5-tts-voicedesign"];
60
105
  /** Voice-design descriptions adapted from the reference voice-definition page. */
61
106
  const TTS_VOICE_DESIGN_PRESETS = [
62
107
  {
63
108
  id: "energetic-girl",
64
- label: "林小满",
65
- summary: "女 · 16岁 · 元气少女,明亮高饱和声线",
66
- prompt: "年轻女性15-20岁,普通话,明亮高饱和声线,笑意自然外放,咬字灵巧跳跃,语速偏快,语调上扬有活力,情绪积极爽朗,活力播报风格"
109
+ label: "鲸鱼娘",
110
+ summary: "爱吃白米饭",
111
+ prompt: "年轻女性16-22岁,标准普通话,清透甜美的中高音,音色明亮而不尖锐,带一点轻盈柔软的空气感;吐字清楚、节奏灵动,语速中等偏快,语调自然上扬,情绪开朗亲切又略带俏皮,整体听感温柔、有陪伴感。"
112
+ },
113
+ {
114
+ id: "liang-wenfeng",
115
+ label: "梁文峰",
116
+ summary: "理性克制",
117
+ prompt: "成年男性 35–40 岁,普通话,温和克制的中低音,声线清晰自然,声音偏亮不沉闷,略带书卷气和理工感,语速中等偏慢,停顿审慎,表达理性简洁,情绪稳定,不夸张。"
67
118
  },
68
119
  {
69
120
  id: "asmr-whisper",
70
121
  label: "沈听澜",
71
- summary: "女 · 19岁 · ASMR低语,轻柔耳语带微弱气息",
122
+ summary: "ASMR低语",
72
123
  prompt: "女性18-20岁,轻柔耳语带微弱气息,普通话,声线细腻清晰,私密温柔感,安静平和带轻柔低语,语速缓慢音量很轻,私密低语场景。"
73
124
  },
125
+ {
126
+ id: "young-man",
127
+ label: "江予辰",
128
+ summary: "阳光少年",
129
+ prompt: "男性青年16-22岁,清亮干净的中高音带少年感,普通话标准无口音,轻快明亮的活力声线,气息轻盈吐字利索,语速偏快语调自然上扬,情绪积极阳光带朝气,广告旁白或轻松解说场景。"
130
+ },
74
131
  {
75
132
  id: "gentle-girlfriend",
76
- label: "张子莯",
77
- summary: "女 · 22岁 · 温柔女友,声线柔软细腻",
133
+ label: "张子苜",
134
+ summary: "温柔陪伴",
78
135
  prompt: "年轻女性16-22岁,声线柔软细腻,低饱和带微微暖意,标准普通话,温柔亲密的邻家风格,语速偏慢,语调轻柔连贯,气息自然流畅,安静私密陪伴场景。"
79
136
  },
137
+ {
138
+ id: "tech-explainer",
139
+ label: "周砚川",
140
+ summary: "科技解说",
141
+ prompt: "成年男性25-35岁,清晰利落中音、干净偏冷,标准普通话,精准干练的都市精英感,语速中等偏快、语调平稳,理性简洁、逻辑感强,现代资讯播报或商业讲解场景。"
142
+ },
80
143
  {
81
144
  id: "girl-next-door",
82
145
  label: "陈念安",
83
- summary: "女 · 25岁 · 邻家女孩,柔润清甜带撒娇感",
146
+ summary: "邻家甜声",
84
147
  prompt: "年轻女性20-25岁,柔润清甜带撒娇感,普通话,清澈明亮的少女音,轻松温柔的亲近感,轻松平缓带温柔,中等偏快语速中等音量,生活分享场景。"
85
148
  },
149
+ {
150
+ id: "documentary-narrator",
151
+ label: "陆远山",
152
+ summary: "纪录片旁白",
153
+ prompt: "男性40-50岁,低沉醇厚有胸腔共鸣,普通话,稳重可靠的叙事者风格,语速中等偏慢,语调沉稳克制带叙事纵深感,气息舒展停顿有留白,纪录片旁白或深度访谈场景。"
154
+ },
86
155
  {
87
156
  id: "news-anchor",
88
157
  label: "顾知微",
89
- summary: "女 · 35岁 · 新闻播报,声线中低音区饱满清晰",
158
+ summary: "沉稳播报",
90
159
  prompt: "专业新闻播音女主持,成年女性30-40岁,普通话标准无口音,声线中低音区饱满清晰,端庄知性沉稳,语气克制权威,语速从容均匀,音量适中稳定,新闻播报与专题解说场景。"
91
160
  },
92
- {
93
- id: "young-man",
94
- label: "江予辰",
95
- summary: "男 · 19岁 · 青年男声,清亮干净的中高音带少年感",
96
- prompt: "男性青年16-22岁,清亮干净的中高音带少年感,普通话标准无口音,轻快明亮的活力声线,气息轻盈吐字利索,语速偏快语调自然上扬,情绪积极阳光带朝气,广告旁白或轻松解说场景。"
97
- },
98
- {
99
- id: "tech-explainer",
100
- label: "周砚川",
101
- summary: "男 · 30岁 · 科技解说,清晰利落中音、干净偏冷",
102
- prompt: "成年男性25-35岁,清晰利落中音、干净偏冷,标准普通话,精准干练的都市精英感,语速中等偏快、语调平稳,理性简洁、逻辑感强,现代资讯播报或商业讲解场景。"
103
- },
104
161
  {
105
162
  id: "suspense-narrator",
106
163
  label: "裴沉舟",
107
- summary: "男 · 35岁 · 悬疑旁白,低沉沉稳带神秘磁性",
164
+ summary: "悬疑旁白",
108
165
  prompt: "男性中年30-40岁,低沉沉稳带神秘磁性,普通话,压抑克制的叙事风格,语速缓慢均匀,语调低沉平稳,情绪冷静悬疑,旁白解说场景。"
109
- },
110
- {
111
- id: "documentary-narrator",
112
- label: "陆远山",
113
- summary: "男 · 45岁 · 纪录片,低沉醇厚有胸腔共鸣",
114
- prompt: "男性40-50岁,低沉醇厚有胸腔共鸣,普通话,稳重可靠的叙事者风格,语速中等偏慢,语调沉稳克制带叙事纵深感,气息舒展停顿有留白,纪录片旁白或深度访谈场景。"
115
166
  }
116
167
  ];
117
168
  /** Supported audio formats. */
118
- const TTS_FORMATS = ["mp3", "wav"];
169
+ const TTS_FORMATS = [
170
+ "pcm",
171
+ "mp3",
172
+ "wav"
173
+ ];
119
174
  /** Minimum spoken characters to accumulate before starting one PCM stream request. */
120
175
  const MIN_TTS_STREAM_CHARACTERS = 20;
176
+ /** Maximum decoded complete MP3 payload accepted from Xiaomi MiMo. */
177
+ const DEFAULT_MAX_MP3_AUDIO_BYTES = 32 * 1024 * 1024;
178
+ /** Maximum decoded complete WAV payload accepted from Xiaomi MiMo. */
179
+ const DEFAULT_MAX_WAV_AUDIO_BYTES = 128 * 1024 * 1024;
180
+ const DEFAULT_MAX_PAUSED_PCM_BYTES = 32 * 1024 * 1024;
181
+ /** Extra JSON envelope allowance around a Base64 audio payload. */
182
+ const TTS_AUDIO_RESPONSE_JSON_OVERHEAD_BYTES = 1024 * 1024;
121
183
  const TTS_PUNCTUATION = {
122
184
  ",": ",",
123
185
  "。": ".",
@@ -179,6 +241,13 @@ function splitCompletedTtsSentences(value) {
179
241
  function countTtsSpeechCharacters(value) {
180
242
  return Array.from(value).filter((character) => /[\p{L}\p{N}]/u.test(character)).length;
181
243
  }
244
+ /** Return the exact decoded size of canonical padded Base64, or null when invalid. */
245
+ function strictBase64DecodedLength(value) {
246
+ if (value.length === 0 || value.length % 4 !== 0) return null;
247
+ if (!/^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/u.test(value)) return null;
248
+ const padding = value.endsWith("==") ? 2 : value.endsWith("=") ? 1 : 0;
249
+ return value.length / 4 * 3 - padding;
250
+ }
182
251
  /** Accumulate short sentences so a PCM stream has enough text to sound natural. */
183
252
  function batchTtsStreamText(pending, next, flush) {
184
253
  const combined = `${pending}${next}`;
@@ -267,11 +336,14 @@ const DEFAULT_TTS_SETTINGS = {
267
336
  voiceDesignPrompt: "青年女性,声线清亮、亲切自然,吐字清楚,语速适中,情绪温柔克制。",
268
337
  voiceDesignCustomPrompt: "青年女性,声线清亮、亲切自然,吐字清楚,语速适中,情绪温柔克制。",
269
338
  presetStylePrompt: "使用清晰、自然、准确的声音朗读,语速适中,停顿自然,语气平和克制,避免夸张表达。",
270
- format: "mp3",
339
+ format: "pcm",
271
340
  autoPlay: true,
272
341
  instruction: "请忠实朗读原文,根据文本语气自然表达,不添加或改写内容。",
273
342
  maxTextLength: 12e3,
274
- requestTimeoutMs: 12e4
343
+ requestTimeoutMs: 12e4,
344
+ maxMp3AudioBytes: DEFAULT_MAX_MP3_AUDIO_BYTES,
345
+ maxWavAudioBytes: DEFAULT_MAX_WAV_AUDIO_BYTES,
346
+ maxPausedPcmBytes: DEFAULT_MAX_PAUSED_PCM_BYTES
275
347
  };
276
348
  /** Select the Xiaomi endpoint from the API key while preserving custom settings for other keys. */
277
349
  function resolveTtsBaseURL(apiKey, configuredBaseURL) {
@@ -297,4 +369,4 @@ function resolveTtsSettings(value) {
297
369
  }
298
370
 
299
371
  //#endregion
300
- export { prepareTtsText as C, splitCompletedTtsSentences as E, parseSseRecords as S, resolveTtsSettings as T, batchTtsStreamText as _, TTS_API_KEY_STATUS_ROUTE as a, isNewerTtsVersion as b, TTS_ROUTE as c, TTS_UNINSTALL_ROUTE as d, TTS_UPDATE_ROUTE as f, TTS_VOICE_DESIGN_PRESETS as g, TTS_VOICE_DESIGN_ASSET_ROUTE as h, TOKEN_PLAN_TTS_BASE_URL as i, TTS_SETTINGS_NAMESPACE as l, TTS_VOICES as m, DEFAULT_TTS_SETTINGS as n, TTS_FORMATS as o, TTS_VERSION as p, MIN_TTS_STREAM_CHARACTERS as r, TTS_MODELS as s, AbortableSentenceQueue as t, TTS_STREAM_ROUTE as u, classifyLiveSpeechTransition as v, resolveTtsBaseURL as w, isSupportedTtsApiKey as x, countTtsSpeechCharacters as y };
372
+ export { resolveTtsBaseURL as A, batchTtsStreamText as C, isSupportedTtsApiKey as D, isNewerTtsVersion as E, splitCompletedTtsSentences as M, strictBase64DecodedLength as N, parseSseRecords as O, TTS_VOICE_PRESETS as S, countTtsSpeechCharacters as T, TTS_VERSION as _, DEFAULT_TTS_SETTINGS as a, TTS_VOICE_DESIGN_ASSET_ROUTE as b, TTS_API_KEY_STATUS_ROUTE as c, TTS_MODELS as d, TTS_ROUTE as f, TTS_UPDATE_ROUTE as g, TTS_UNINSTALL_ROUTE as h, DEFAULT_MAX_WAV_AUDIO_BYTES as i, resolveTtsSettings as j, prepareTtsText as k, TTS_AUDIO_RESPONSE_JSON_OVERHEAD_BYTES as l, TTS_STREAM_ROUTE as m, DEFAULT_MAX_MP3_AUDIO_BYTES as n, MIN_TTS_STREAM_CHARACTERS as o, TTS_SETTINGS_NAMESPACE as p, DEFAULT_MAX_PAUSED_PCM_BYTES as r, TOKEN_PLAN_TTS_BASE_URL as s, AbortableSentenceQueue as t, TTS_FORMATS as u, TTS_VOICES as v, classifyLiveSpeechTransition as w, TTS_VOICE_DESIGN_PRESETS as x, TTS_VOICE_ASSET_ROUTE as y };
package/lib/shared.d.ts CHANGED
@@ -9,72 +9,120 @@ export declare const TTS_UNINSTALL_ROUTE = "/plugins/xiaomi-mimo-tts/uninstall";
9
9
  /** Same-origin route that checks the published npm version. */
10
10
  export declare const TTS_UPDATE_ROUTE = "/plugins/xiaomi-mimo-tts/update";
11
11
  /** Keep the UI version visible without making the browser bundle load package.json. */
12
- export declare const TTS_VERSION = "2.3.2";
12
+ export declare const TTS_VERSION = "2.3.3";
13
13
  /** Return true only when `candidate` is a strictly newer semantic version. */
14
14
  export declare function isNewerTtsVersion(candidate: string, current: string): boolean;
15
15
  /** Same-origin route that reports API key configuration status without exposing the key. */
16
16
  export declare const TTS_API_KEY_STATUS_ROUTE = "/plugins/xiaomi-mimo-tts/api-key-status";
17
17
  /** Same-origin prefix used by the Web client to load voice-design preset icons. */
18
18
  export declare const TTS_VOICE_DESIGN_ASSET_ROUTE = "/plugins/xiaomi-mimo-tts/voice-presets";
19
+ /** Same-origin prefix used by the Web client to load official built-in voice avatars. */
20
+ export declare const TTS_VOICE_ASSET_ROUTE = "/plugins/xiaomi-mimo-tts/voice-avatars";
19
21
  /** Default Xiaomi endpoint for Token Plan API keys. */
20
22
  export declare const TOKEN_PLAN_TTS_BASE_URL = "https://token-plan-cn.xiaomimimo.com/v1";
21
23
  /** Supported built-in Xiaomi MiMo voices. */
22
24
  export declare const TTS_VOICES: readonly ["冰糖", "茉莉", "苏打", "白桦", "Mia", "Chloe", "Milo", "Dean"];
25
+ /** Built-in voice metadata mirrored from Xiaomi MiMo's official selector. */
26
+ export declare const TTS_VOICE_PRESETS: readonly [{
27
+ readonly id: "bingtang";
28
+ readonly value: "冰糖";
29
+ readonly summary: "活泼少女";
30
+ }, {
31
+ readonly id: "mia";
32
+ readonly value: "Mia";
33
+ readonly summary: "Lively girl";
34
+ }, {
35
+ readonly id: "moli";
36
+ readonly value: "茉莉";
37
+ readonly summary: "知性女声";
38
+ }, {
39
+ readonly id: "chloe";
40
+ readonly value: "Chloe";
41
+ readonly summary: "Witty Grace";
42
+ }, {
43
+ readonly id: "suda";
44
+ readonly value: "苏打";
45
+ readonly summary: "阳光少年";
46
+ }, {
47
+ readonly id: "milo";
48
+ readonly value: "Milo";
49
+ readonly summary: "Sunny boy";
50
+ }, {
51
+ readonly id: "baihua";
52
+ readonly value: "白桦";
53
+ readonly summary: "成熟男声";
54
+ }, {
55
+ readonly id: "dean";
56
+ readonly value: "Dean";
57
+ readonly summary: "Steady Gentle";
58
+ }];
23
59
  /** TTS models supported by this plugin. */
24
60
  export declare const TTS_MODELS: readonly ["mimo-v2.5-tts", "mimo-v2.5-tts-voicedesign"];
25
61
  export type TtsModel = typeof TTS_MODELS[number];
26
62
  /** Voice-design descriptions adapted from the reference voice-definition page. */
27
63
  export declare const TTS_VOICE_DESIGN_PRESETS: readonly [{
28
64
  readonly id: "energetic-girl";
29
- readonly label: "林小满";
30
- readonly summary: "女 · 16岁 · 元气少女,明亮高饱和声线";
31
- readonly prompt: "年轻女性15-20岁,普通话,明亮高饱和声线,笑意自然外放,咬字灵巧跳跃,语速偏快,语调上扬有活力,情绪积极爽朗,活力播报风格";
65
+ readonly label: "鲸鱼娘";
66
+ readonly summary: "爱吃白米饭";
67
+ readonly prompt: "年轻女性16-22岁,标准普通话,清透甜美的中高音,音色明亮而不尖锐,带一点轻盈柔软的空气感;吐字清楚、节奏灵动,语速中等偏快,语调自然上扬,情绪开朗亲切又略带俏皮,整体听感温柔、有陪伴感。";
68
+ }, {
69
+ readonly id: "liang-wenfeng";
70
+ readonly label: "梁文峰";
71
+ readonly summary: "理性克制";
72
+ readonly prompt: "成年男性 35–40 岁,普通话,温和克制的中低音,声线清晰自然,声音偏亮不沉闷,略带书卷气和理工感,语速中等偏慢,停顿审慎,表达理性简洁,情绪稳定,不夸张。";
32
73
  }, {
33
74
  readonly id: "asmr-whisper";
34
75
  readonly label: "沈听澜";
35
- readonly summary: "女 · 19岁 · ASMR低语,轻柔耳语带微弱气息";
76
+ readonly summary: "ASMR低语";
36
77
  readonly prompt: "女性18-20岁,轻柔耳语带微弱气息,普通话,声线细腻清晰,私密温柔感,安静平和带轻柔低语,语速缓慢音量很轻,私密低语场景。";
78
+ }, {
79
+ readonly id: "young-man";
80
+ readonly label: "江予辰";
81
+ readonly summary: "阳光少年";
82
+ readonly prompt: "男性青年16-22岁,清亮干净的中高音带少年感,普通话标准无口音,轻快明亮的活力声线,气息轻盈吐字利索,语速偏快语调自然上扬,情绪积极阳光带朝气,广告旁白或轻松解说场景。";
37
83
  }, {
38
84
  readonly id: "gentle-girlfriend";
39
- readonly label: "张子莯";
40
- readonly summary: "女 · 22岁 · 温柔女友,声线柔软细腻";
85
+ readonly label: "张子苜";
86
+ readonly summary: "温柔陪伴";
41
87
  readonly prompt: "年轻女性16-22岁,声线柔软细腻,低饱和带微微暖意,标准普通话,温柔亲密的邻家风格,语速偏慢,语调轻柔连贯,气息自然流畅,安静私密陪伴场景。";
88
+ }, {
89
+ readonly id: "tech-explainer";
90
+ readonly label: "周砚川";
91
+ readonly summary: "科技解说";
92
+ readonly prompt: "成年男性25-35岁,清晰利落中音、干净偏冷,标准普通话,精准干练的都市精英感,语速中等偏快、语调平稳,理性简洁、逻辑感强,现代资讯播报或商业讲解场景。";
42
93
  }, {
43
94
  readonly id: "girl-next-door";
44
95
  readonly label: "陈念安";
45
- readonly summary: "女 · 25岁 · 邻家女孩,柔润清甜带撒娇感";
96
+ readonly summary: "邻家甜声";
46
97
  readonly prompt: "年轻女性20-25岁,柔润清甜带撒娇感,普通话,清澈明亮的少女音,轻松温柔的亲近感,轻松平缓带温柔,中等偏快语速中等音量,生活分享场景。";
98
+ }, {
99
+ readonly id: "documentary-narrator";
100
+ readonly label: "陆远山";
101
+ readonly summary: "纪录片旁白";
102
+ readonly prompt: "男性40-50岁,低沉醇厚有胸腔共鸣,普通话,稳重可靠的叙事者风格,语速中等偏慢,语调沉稳克制带叙事纵深感,气息舒展停顿有留白,纪录片旁白或深度访谈场景。";
47
103
  }, {
48
104
  readonly id: "news-anchor";
49
105
  readonly label: "顾知微";
50
- readonly summary: "女 · 35岁 · 新闻播报,声线中低音区饱满清晰";
106
+ readonly summary: "沉稳播报";
51
107
  readonly prompt: "专业新闻播音女主持,成年女性30-40岁,普通话标准无口音,声线中低音区饱满清晰,端庄知性沉稳,语气克制权威,语速从容均匀,音量适中稳定,新闻播报与专题解说场景。";
52
- }, {
53
- readonly id: "young-man";
54
- readonly label: "江予辰";
55
- readonly summary: "男 · 19岁 · 青年男声,清亮干净的中高音带少年感";
56
- readonly prompt: "男性青年16-22岁,清亮干净的中高音带少年感,普通话标准无口音,轻快明亮的活力声线,气息轻盈吐字利索,语速偏快语调自然上扬,情绪积极阳光带朝气,广告旁白或轻松解说场景。";
57
- }, {
58
- readonly id: "tech-explainer";
59
- readonly label: "周砚川";
60
- readonly summary: "男 · 30岁 · 科技解说,清晰利落中音、干净偏冷";
61
- readonly prompt: "成年男性25-35岁,清晰利落中音、干净偏冷,标准普通话,精准干练的都市精英感,语速中等偏快、语调平稳,理性简洁、逻辑感强,现代资讯播报或商业讲解场景。";
62
108
  }, {
63
109
  readonly id: "suspense-narrator";
64
110
  readonly label: "裴沉舟";
65
- readonly summary: "男 · 35岁 · 悬疑旁白,低沉沉稳带神秘磁性";
111
+ readonly summary: "悬疑旁白";
66
112
  readonly prompt: "男性中年30-40岁,低沉沉稳带神秘磁性,普通话,压抑克制的叙事风格,语速缓慢均匀,语调低沉平稳,情绪冷静悬疑,旁白解说场景。";
67
- }, {
68
- readonly id: "documentary-narrator";
69
- readonly label: "陆远山";
70
- readonly summary: "男 · 45岁 · 纪录片,低沉醇厚有胸腔共鸣";
71
- readonly prompt: "男性40-50岁,低沉醇厚有胸腔共鸣,普通话,稳重可靠的叙事者风格,语速中等偏慢,语调沉稳克制带叙事纵深感,气息舒展停顿有留白,纪录片旁白或深度访谈场景。";
72
113
  }];
73
114
  /** Supported audio formats. */
74
- export declare const TTS_FORMATS: readonly ["mp3", "wav"];
115
+ export declare const TTS_FORMATS: readonly ["pcm", "mp3", "wav"];
75
116
  export type TtsFormat = typeof TTS_FORMATS[number];
76
117
  /** Minimum spoken characters to accumulate before starting one PCM stream request. */
77
118
  export declare const MIN_TTS_STREAM_CHARACTERS = 20;
119
+ /** Maximum decoded complete MP3 payload accepted from Xiaomi MiMo. */
120
+ export declare const DEFAULT_MAX_MP3_AUDIO_BYTES: number;
121
+ /** Maximum decoded complete WAV payload accepted from Xiaomi MiMo. */
122
+ export declare const DEFAULT_MAX_WAV_AUDIO_BYTES: number;
123
+ export declare const DEFAULT_MAX_PAUSED_PCM_BYTES: number;
124
+ /** Extra JSON envelope allowance around a Base64 audio payload. */
125
+ export declare const TTS_AUDIO_RESPONSE_JSON_OVERHEAD_BYTES: number;
78
126
  /**
79
127
  * Prepare assistant text for speech synthesis without changing the chat transcript.
80
128
  *
@@ -89,6 +137,8 @@ export declare function splitCompletedTtsSentences(value: string): {
89
137
  };
90
138
  /** Count text-bearing characters only, excluding punctuation and whitespace. */
91
139
  export declare function countTtsSpeechCharacters(value: string): number;
140
+ /** Return the exact decoded size of canonical padded Base64, or null when invalid. */
141
+ export declare function strictBase64DecodedLength(value: string): number | null;
92
142
  /** Accumulate short sentences so a PCM stream has enough text to sound natural. */
93
143
  export declare function batchTtsStreamText(pending: string, next: string, flush: boolean): {
94
144
  pending: string;
@@ -139,6 +189,9 @@ export interface TtsSettings {
139
189
  instruction?: string;
140
190
  maxTextLength?: number;
141
191
  requestTimeoutMs?: number;
192
+ maxMp3AudioBytes?: number;
193
+ maxWavAudioBytes?: number;
194
+ maxPausedPcmBytes?: number;
142
195
  }
143
196
  export interface ResolvedTtsSettings {
144
197
  enabled: boolean;
@@ -154,6 +207,9 @@ export interface ResolvedTtsSettings {
154
207
  instruction: string;
155
208
  maxTextLength: number;
156
209
  requestTimeoutMs: number;
210
+ maxMp3AudioBytes: number;
211
+ maxWavAudioBytes: number;
212
+ maxPausedPcmBytes: number;
157
213
  }
158
214
  /** Defaults shared by the Schemastery config and the Web settings form. */
159
215
  export declare const DEFAULT_TTS_SETTINGS: ResolvedTtsSettings;