dsh-xiaomi-tts 2.3.2 → 2.3.4
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.
- package/README.en.md +198 -120
- package/README.md +193 -185
- package/assets/menu.png +0 -0
- package/assets/preset.png +0 -0
- package/assets/social.png +0 -0
- package/assets/voice-avatars/baihua.webp +0 -0
- package/assets/voice-avatars/bingtang.webp +0 -0
- package/assets/voice-avatars/chloe.webp +0 -0
- package/assets/voice-avatars/dean.webp +0 -0
- package/assets/voice-avatars/mia.webp +0 -0
- package/assets/voice-avatars/milo.webp +0 -0
- package/assets/voice-avatars/moli.webp +0 -0
- package/assets/voice-avatars/suda.webp +0 -0
- package/assets/voice-presets/energetic-girl.webp +0 -0
- package/assets/voice-presets/liang-wenfeng.webp +0 -0
- package/cordis.patch.yml +4 -1
- package/lib/client/built-in-voice-picker.d.ts +10 -0
- package/lib/client/built-in-voice-picker.d.ts.map +1 -0
- package/lib/client/conversation.d.ts.map +1 -1
- package/lib/client/live-speech-controller.d.ts +6 -0
- package/lib/client/live-speech-controller.d.ts.map +1 -1
- package/lib/client/localization.d.ts +15 -4
- package/lib/client/localization.d.ts.map +1 -1
- package/lib/client/pcm-audio-queue.d.ts +7 -0
- package/lib/client/pcm-audio-queue.d.ts.map +1 -1
- package/lib/client/playback-controller.d.ts +4 -0
- package/lib/client/playback-controller.d.ts.map +1 -1
- package/lib/client/playback-types.d.ts +1 -1
- package/lib/client/playback-types.d.ts.map +1 -1
- package/lib/client/settings-card.d.ts.map +1 -1
- package/lib/client/settings-scope.d.ts.map +1 -1
- package/lib/client/styles.d.ts +1 -1
- package/lib/client/styles.d.ts.map +1 -1
- package/lib/client/voice-design-picker.d.ts.map +1 -1
- package/lib/client.js +880 -242
- package/lib/client.js.map +1 -1
- package/lib/index.d.ts +10 -2
- package/lib/index.d.ts.map +1 -1
- package/lib/index.js +122 -33
- package/lib/{shared-9qFFYsWe.js → shared-CIx9nBxX.js} +226 -44
- package/lib/shared.d.ts +100 -29
- package/lib/shared.d.ts.map +1 -1
- package/lib/shared.js +2 -2
- package/package.json +104 -103
package/lib/index.d.ts
CHANGED
|
@@ -16,11 +16,15 @@ 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
|
+
voiceDesignPlaybackMode: z<"complete" | "segmented", "complete" | "segmented">;
|
|
20
21
|
autoPlay: z<boolean, boolean>;
|
|
21
22
|
instruction: z<string, string>;
|
|
22
23
|
maxTextLength: z<number, number>;
|
|
23
24
|
requestTimeoutMs: z<number, number>;
|
|
25
|
+
maxMp3AudioBytes: z<number, number>;
|
|
26
|
+
maxWavAudioBytes: z<number, number>;
|
|
27
|
+
maxPausedPcmBytes: z<number, number>;
|
|
24
28
|
}>, Schemastery.ObjectT<{
|
|
25
29
|
enabled: z<boolean, boolean>;
|
|
26
30
|
apiKey: z<string, string>;
|
|
@@ -30,11 +34,15 @@ export declare const Config: z<Schemastery.ObjectS<{
|
|
|
30
34
|
voiceDesignPrompt: z<string, string>;
|
|
31
35
|
voiceDesignCustomPrompt: z<string, string>;
|
|
32
36
|
presetStylePrompt: z<string, string>;
|
|
33
|
-
format: z<"mp3" | "wav", "mp3" | "wav">;
|
|
37
|
+
format: z<"pcm" | "mp3" | "wav", "pcm" | "mp3" | "wav">;
|
|
38
|
+
voiceDesignPlaybackMode: z<"complete" | "segmented", "complete" | "segmented">;
|
|
34
39
|
autoPlay: z<boolean, boolean>;
|
|
35
40
|
instruction: z<string, string>;
|
|
36
41
|
maxTextLength: z<number, number>;
|
|
37
42
|
requestTimeoutMs: z<number, number>;
|
|
43
|
+
maxMp3AudioBytes: z<number, number>;
|
|
44
|
+
maxWavAudioBytes: z<number, number>;
|
|
45
|
+
maxPausedPcmBytes: z<number, number>;
|
|
38
46
|
}>>;
|
|
39
47
|
export type Config = ReturnType<typeof Config>;
|
|
40
48
|
/** Register the TTS settings and same-origin synthesis route. */
|
package/lib/index.d.ts.map
CHANGED
|
@@ -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
|
|
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;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAkBjB,CAAA;AAEF,MAAM,MAAM,MAAM,GAAG,UAAU,CAAC,OAAO,MAAM,CAAC,CAAA;AAmT9C,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,
|
|
1
|
+
import { A as isNewerTtsVersion, C as TTS_VOICE_DESIGN_ASSET_ROUTE, E as TTS_VOICE_PRESETS, N as prepareTtsText, P as resolveTtsBaseURL, R as strictBase64DecodedLength, S as TTS_VOICE_ASSET_ROUTE, T as TTS_VOICE_DESIGN_PRESETS, _ as TTS_STREAM_ROUTE, b as TTS_VERSION, d as TTS_API_KEY_STATUS_ROUTE, f as TTS_AUDIO_RESPONSE_JSON_OVERHEAD_BYTES, g as TTS_SETTINGS_NAMESPACE, h as TTS_ROUTE, j as isSupportedTtsApiKey, m as TTS_MODELS, o as DEFAULT_TTS_SETTINGS, p as TTS_FORMATS, v as TTS_UNINSTALL_ROUTE, w as TTS_VOICE_DESIGN_PLAYBACK_MODES, y as TTS_UPDATE_ROUTE } from "./shared-CIx9nBxX.js";
|
|
2
2
|
import { createRequire } from "node:module";
|
|
3
3
|
import { spawn } from "node:child_process";
|
|
4
4
|
import { existsSync, readFileSync } from "node:fs";
|
|
@@ -30,10 +30,14 @@ const Config = z.object({
|
|
|
30
30
|
voiceDesignCustomPrompt: z.string().default(DEFAULT_TTS_SETTINGS.voiceDesignCustomPrompt),
|
|
31
31
|
presetStylePrompt: z.string().default(DEFAULT_TTS_SETTINGS.presetStylePrompt),
|
|
32
32
|
format: z.union(TTS_FORMATS).default(DEFAULT_TTS_SETTINGS.format),
|
|
33
|
+
voiceDesignPlaybackMode: z.union(TTS_VOICE_DESIGN_PLAYBACK_MODES).default(DEFAULT_TTS_SETTINGS.voiceDesignPlaybackMode),
|
|
33
34
|
autoPlay: z.boolean().default(DEFAULT_TTS_SETTINGS.autoPlay),
|
|
34
35
|
instruction: z.string().default(DEFAULT_TTS_SETTINGS.instruction),
|
|
35
36
|
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)
|
|
37
|
+
requestTimeoutMs: z.number().step(1).min(1e3).default(DEFAULT_TTS_SETTINGS.requestTimeoutMs),
|
|
38
|
+
maxMp3AudioBytes: z.number().step(1).min(1).default(DEFAULT_TTS_SETTINGS.maxMp3AudioBytes),
|
|
39
|
+
maxWavAudioBytes: z.number().step(1).min(1).default(DEFAULT_TTS_SETTINGS.maxWavAudioBytes),
|
|
40
|
+
maxPausedPcmBytes: z.number().step(1).min(1).default(DEFAULT_TTS_SETTINGS.maxPausedPcmBytes)
|
|
37
41
|
});
|
|
38
42
|
function requestMessages(options, text) {
|
|
39
43
|
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 +78,61 @@ function apiErrorMessage(status, parsed) {
|
|
|
74
78
|
const detail = typeof parsed?.error === "string" ? parsed.error : parsed?.error?.message ?? parsed?.message;
|
|
75
79
|
return detail && detail.length > 0 ? detail : `Xiaomi MiMo TTS request failed (HTTP ${status})`;
|
|
76
80
|
}
|
|
81
|
+
var CompleteAudioResponseError = class extends Error {
|
|
82
|
+
constructor(code) {
|
|
83
|
+
super(code);
|
|
84
|
+
this.code = code;
|
|
85
|
+
}
|
|
86
|
+
};
|
|
87
|
+
function completeAudioFormat(options) {
|
|
88
|
+
return options.format === "wav" ? "wav" : "mp3";
|
|
89
|
+
}
|
|
90
|
+
function completeAudioLimit(options, format) {
|
|
91
|
+
return format === "mp3" ? options.maxMp3AudioBytes : options.maxWavAudioBytes;
|
|
92
|
+
}
|
|
93
|
+
function completeAudioJsonLimit(audioLimit) {
|
|
94
|
+
return 4 * Math.ceil(audioLimit / 3) + TTS_AUDIO_RESPONSE_JSON_OVERHEAD_BYTES;
|
|
95
|
+
}
|
|
96
|
+
async function readLimitedXiaomiResponse(response, limit, controller) {
|
|
97
|
+
const contentLength = response.headers.get("content-length");
|
|
98
|
+
if (contentLength !== null && /^\d+$/u.test(contentLength) && Number(contentLength) > limit) {
|
|
99
|
+
controller.abort();
|
|
100
|
+
throw new CompleteAudioResponseError("xiaomi-response-too-large");
|
|
101
|
+
}
|
|
102
|
+
if (response.body === null) return void 0;
|
|
103
|
+
const reader = response.body.getReader();
|
|
104
|
+
const chunks = [];
|
|
105
|
+
let size = 0;
|
|
106
|
+
try {
|
|
107
|
+
while (true) {
|
|
108
|
+
const chunk = await reader.read();
|
|
109
|
+
if (chunk.done) break;
|
|
110
|
+
size += chunk.value.byteLength;
|
|
111
|
+
if (size > limit) {
|
|
112
|
+
controller.abort();
|
|
113
|
+
throw new CompleteAudioResponseError("xiaomi-response-too-large");
|
|
114
|
+
}
|
|
115
|
+
chunks.push(Buffer.from(chunk.value));
|
|
116
|
+
}
|
|
117
|
+
} finally {
|
|
118
|
+
reader.releaseLock();
|
|
119
|
+
}
|
|
120
|
+
if (chunks.length === 0) return void 0;
|
|
121
|
+
try {
|
|
122
|
+
return JSON.parse(Buffer.concat(chunks, size).toString("utf8"));
|
|
123
|
+
} catch {
|
|
124
|
+
return;
|
|
125
|
+
}
|
|
126
|
+
}
|
|
127
|
+
function decodeCompleteAudio(value, format, limit) {
|
|
128
|
+
const expectedLength = strictBase64DecodedLength(value);
|
|
129
|
+
if (expectedLength === null) throw new CompleteAudioResponseError("invalid-audio-base64");
|
|
130
|
+
if (expectedLength > limit) throw new CompleteAudioResponseError("audio-too-large");
|
|
131
|
+
const audio = Buffer.from(value, "base64");
|
|
132
|
+
if (audio.byteLength !== expectedLength) throw new CompleteAudioResponseError("invalid-audio-base64");
|
|
133
|
+
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");
|
|
134
|
+
return audio;
|
|
135
|
+
}
|
|
77
136
|
function webProfileDirectory() {
|
|
78
137
|
return join(process.env.DSH_HOME?.trim() || join(homedir(), ".dsh"), "profiles", WEB_PROFILE_NAME);
|
|
79
138
|
}
|
|
@@ -272,6 +331,9 @@ function apply(ctx, config) {
|
|
|
272
331
|
const voicePresetAssets = new Map(TTS_VOICE_DESIGN_PRESETS.map((preset) => {
|
|
273
332
|
return [`${TTS_VOICE_DESIGN_ASSET_ROUTE}/${preset.id}.webp`, readFileSync(new URL(`../assets/voice-presets/${preset.id}.webp`, import.meta.url))];
|
|
274
333
|
}));
|
|
334
|
+
const voiceAssets = new Map(TTS_VOICE_PRESETS.map((preset) => {
|
|
335
|
+
return [`${TTS_VOICE_ASSET_ROUTE}/${preset.id}.webp`, readFileSync(new URL(`../assets/voice-avatars/${preset.id}.webp`, import.meta.url))];
|
|
336
|
+
}));
|
|
275
337
|
installSettingsSection(ctx, XIAOMI_MIMO_TTS_SETTINGS_NAMESPACE, Config, config, {
|
|
276
338
|
setSource(source) {
|
|
277
339
|
current = source;
|
|
@@ -281,7 +343,7 @@ function apply(ctx, config) {
|
|
|
281
343
|
const endpoint = `${normalizeBaseURL(value.baseURL)}/chat/completions`;
|
|
282
344
|
if (!URL.canParse(endpoint)) throw new Error("baseURL must be a valid absolute URL");
|
|
283
345
|
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 (
|
|
346
|
+
if (!TTS_FORMATS.includes(value.format)) throw new Error("format must be pcm, mp3, or wav");
|
|
285
347
|
}
|
|
286
348
|
});
|
|
287
349
|
ctx.effect(() => ctx.webServer.register({
|
|
@@ -388,6 +450,31 @@ function apply(ctx, config) {
|
|
|
388
450
|
res.end(req.method === "HEAD" ? void 0 : asset);
|
|
389
451
|
}
|
|
390
452
|
}), "xiaomi-mimo-tts: voice preset assets");
|
|
453
|
+
ctx.effect(() => ctx.webServer.register({
|
|
454
|
+
kind: "prefix",
|
|
455
|
+
path: TTS_VOICE_ASSET_ROUTE,
|
|
456
|
+
handler(req, res) {
|
|
457
|
+
if (req.method !== "GET" && req.method !== "HEAD") {
|
|
458
|
+
res.statusCode = 405;
|
|
459
|
+
res.setHeader("allow", "GET, HEAD");
|
|
460
|
+
res.end();
|
|
461
|
+
return;
|
|
462
|
+
}
|
|
463
|
+
const pathname = new URL(req.url ?? "/", "http://localhost").pathname;
|
|
464
|
+
const asset = voiceAssets.get(pathname);
|
|
465
|
+
if (asset === void 0) {
|
|
466
|
+
res.statusCode = 404;
|
|
467
|
+
res.end();
|
|
468
|
+
return;
|
|
469
|
+
}
|
|
470
|
+
res.statusCode = 200;
|
|
471
|
+
res.setHeader("content-type", "image/webp");
|
|
472
|
+
res.setHeader("content-length", String(asset.byteLength));
|
|
473
|
+
res.setHeader("cache-control", "public, max-age=31536000, immutable");
|
|
474
|
+
res.setHeader("x-content-type-options", "nosniff");
|
|
475
|
+
res.end(req.method === "HEAD" ? void 0 : asset);
|
|
476
|
+
}
|
|
477
|
+
}), "xiaomi-mimo-tts: built-in voice assets");
|
|
391
478
|
ctx.effect(() => ctx.webServer.register({
|
|
392
479
|
kind: "exact",
|
|
393
480
|
path: TTS_ROUTE,
|
|
@@ -422,8 +509,21 @@ function apply(ctx, config) {
|
|
|
422
509
|
return;
|
|
423
510
|
}
|
|
424
511
|
const controller = new AbortController();
|
|
425
|
-
|
|
512
|
+
let abortReason = null;
|
|
513
|
+
const abort = (reason) => {
|
|
514
|
+
if (controller.signal.aborted) return;
|
|
515
|
+
abortReason = reason;
|
|
516
|
+
controller.abort();
|
|
517
|
+
};
|
|
518
|
+
const timeout = setTimeout(() => abort("timeout"), options.requestTimeoutMs);
|
|
519
|
+
const abortOnRequest = () => abort("client-disconnect");
|
|
520
|
+
const abortOnResponseClose = () => {
|
|
521
|
+
if (!res.writableEnded) abort("client-disconnect");
|
|
522
|
+
};
|
|
523
|
+
req.once("aborted", abortOnRequest);
|
|
524
|
+
res.once("close", abortOnResponseClose);
|
|
426
525
|
try {
|
|
526
|
+
const format = body.format === "wav" ? "wav" : completeAudioFormat(options);
|
|
427
527
|
const endpoint = `${normalizeBaseURL(resolveTtsBaseURL(options.apiKey, options.baseURL))}/chat/completions`;
|
|
428
528
|
const response = await fetch(endpoint, {
|
|
429
529
|
method: "POST",
|
|
@@ -437,20 +537,16 @@ function apply(ctx, config) {
|
|
|
437
537
|
body: JSON.stringify({
|
|
438
538
|
model: options.model,
|
|
439
539
|
messages: requestMessages(options, text),
|
|
440
|
-
audio: options.model === "mimo-v2.5-tts-voicedesign" ? { format
|
|
441
|
-
format
|
|
540
|
+
audio: options.model === "mimo-v2.5-tts-voicedesign" ? { format } : {
|
|
541
|
+
format,
|
|
442
542
|
voice: options.voice
|
|
443
543
|
},
|
|
444
544
|
stream: false
|
|
445
545
|
}),
|
|
446
546
|
signal: controller.signal
|
|
447
547
|
});
|
|
448
|
-
|
|
449
|
-
|
|
450
|
-
parsed = await response.json();
|
|
451
|
-
} catch {
|
|
452
|
-
parsed = void 0;
|
|
453
|
-
}
|
|
548
|
+
const audioLimit = completeAudioLimit(options, format);
|
|
549
|
+
const parsed = await readLimitedXiaomiResponse(response, completeAudioJsonLimit(audioLimit), controller);
|
|
454
550
|
if (!response.ok) {
|
|
455
551
|
json(res, response.status, {
|
|
456
552
|
error: "xiaomi-api-error",
|
|
@@ -466,36 +562,29 @@ function apply(ctx, config) {
|
|
|
466
562
|
});
|
|
467
563
|
return;
|
|
468
564
|
}
|
|
469
|
-
|
|
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
|
-
}
|
|
565
|
+
const audio = decodeCompleteAudio(audioBase64, format, audioLimit);
|
|
486
566
|
res.statusCode = 200;
|
|
487
|
-
res.setHeader("content-type",
|
|
567
|
+
res.setHeader("content-type", format === "mp3" ? "audio/mpeg" : "audio/wav");
|
|
488
568
|
res.setHeader("content-length", String(audio.byteLength));
|
|
489
569
|
res.setHeader("cache-control", "no-store");
|
|
490
570
|
res.end(audio);
|
|
491
571
|
} catch (error) {
|
|
492
|
-
|
|
493
|
-
|
|
494
|
-
|
|
572
|
+
if (abortReason === "client-disconnect" || res.destroyed || res.writableEnded) return;
|
|
573
|
+
if (error instanceof CompleteAudioResponseError) {
|
|
574
|
+
json(res, 502, {
|
|
575
|
+
error: error.code,
|
|
576
|
+
message: error.message
|
|
577
|
+
});
|
|
578
|
+
return;
|
|
579
|
+
}
|
|
580
|
+
json(res, abortReason === "timeout" ? 504 : 502, {
|
|
581
|
+
error: abortReason === "timeout" ? "xiaomi-timeout" : "xiaomi-request-failed",
|
|
495
582
|
message: error instanceof Error ? error.message : String(error)
|
|
496
583
|
});
|
|
497
584
|
} finally {
|
|
498
585
|
clearTimeout(timeout);
|
|
586
|
+
req.off("aborted", abortOnRequest);
|
|
587
|
+
res.off("close", abortOnResponseClose);
|
|
499
588
|
}
|
|
500
589
|
}
|
|
501
590
|
}), "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.
|
|
13
|
+
const TTS_VERSION = "2.3.4";
|
|
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 +57,134 @@ 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: "
|
|
65
|
-
summary: "
|
|
66
|
-
prompt: "年轻女性
|
|
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: "
|
|
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: "
|
|
77
|
-
summary: "
|
|
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: "
|
|
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: "
|
|
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: "
|
|
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 = [
|
|
169
|
+
const TTS_FORMATS = [
|
|
170
|
+
"pcm",
|
|
171
|
+
"mp3",
|
|
172
|
+
"wav"
|
|
173
|
+
];
|
|
174
|
+
/** Voice-design playback strategies. Segmented playback keeps each upstream request short. */
|
|
175
|
+
const TTS_VOICE_DESIGN_PLAYBACK_MODES = ["complete", "segmented"];
|
|
176
|
+
const DEFAULT_TTS_SEGMENT_CHARACTERS = 120;
|
|
177
|
+
const MAX_TTS_SEGMENT_CHARACTERS = 180;
|
|
178
|
+
const MIN_TTS_SEGMENT_CHARACTERS = 40;
|
|
119
179
|
/** Minimum spoken characters to accumulate before starting one PCM stream request. */
|
|
120
180
|
const MIN_TTS_STREAM_CHARACTERS = 20;
|
|
181
|
+
/** Maximum decoded complete MP3 payload accepted from Xiaomi MiMo. */
|
|
182
|
+
const DEFAULT_MAX_MP3_AUDIO_BYTES = 32 * 1024 * 1024;
|
|
183
|
+
/** Maximum decoded complete WAV payload accepted from Xiaomi MiMo. */
|
|
184
|
+
const DEFAULT_MAX_WAV_AUDIO_BYTES = 128 * 1024 * 1024;
|
|
185
|
+
const DEFAULT_MAX_PAUSED_PCM_BYTES = 32 * 1024 * 1024;
|
|
186
|
+
/** Extra JSON envelope allowance around a Base64 audio payload. */
|
|
187
|
+
const TTS_AUDIO_RESPONSE_JSON_OVERHEAD_BYTES = 1024 * 1024;
|
|
121
188
|
const TTS_PUNCTUATION = {
|
|
122
189
|
",": ",",
|
|
123
190
|
"。": ".",
|
|
@@ -132,21 +199,33 @@ const TTS_PUNCTUATION = {
|
|
|
132
199
|
/** Decorative punctuation can make TTS emit non-speech artifacts, so omit it entirely. */
|
|
133
200
|
const TTS_NON_SPEECH_PUNCTUATION = /[()\[\]【】[]〔〕〖〗{}{}「」『』《》〈〉“”‘’"'`<>:…—–-~]/gu;
|
|
134
201
|
const TTS_URL_PATTERN = /\b(?:https?|ftp):\/\/[^\s<>()]+|\bwww\.[^\s<>()]+|\b(?:[a-z0-9](?:[a-z0-9-]{0,61}[a-z0-9])?\.)+(?:com|cn|net|org|io|ai|dev|me|co|edu|gov|xyz|tech|info|app|site|link)(?:[/:?#][^\s<>()]*)?/giu;
|
|
135
|
-
const TTS_WINDOWS_PATH_PATTERN = /(?:\b[A-Za-z]:[\\/]|\\\\)(?:[A-Za-z0-9._ -]+[\\/])
|
|
136
|
-
const TTS_UNIX_PATH_PATTERN = /\/(?:[A-Za-z0-9._-]+\/)
|
|
202
|
+
const TTS_WINDOWS_PATH_PATTERN = /(?:\b[A-Za-z]:[\\/]|\\\\)(?:[A-Za-z0-9._ -]+[\\/])+(?:[A-Za-z0-9._ -]+)/gu;
|
|
203
|
+
const TTS_UNIX_PATH_PATTERN = /\/(?:[A-Za-z0-9._-]+\/)+(?:[A-Za-z0-9._-]+)/gu;
|
|
137
204
|
const TTS_RELATIVE_PATH_PATTERN = /(?:\.\.?[\\/])(?:[A-Za-z0-9._-]+[\\/])*(?:[A-Za-z0-9._-]+(?:\.[A-Za-z0-9_-]+)?)/gu;
|
|
138
205
|
const TTS_PROJECT_PATH_PATTERN = /\b(?:[A-Za-z0-9_-]+[\\/])+[A-Za-z0-9_-]+\.[A-Za-z0-9_-]+\b/gu;
|
|
206
|
+
/** Combined Markdown syntax and path removal in a single pass. */
|
|
207
|
+
const MARKDOWN_CODE_BLOCK_RE = /```[\s\S]*?```|~~~[\s\S]*?~~~/g;
|
|
208
|
+
const MARKDOWN_INDENTED_CODE_RE = /(^|\n)(?: {4}|\t)[^\n]*(?=\n|$)/g;
|
|
209
|
+
const MARKDOWN_IMAGE_RE = /!\[([^\]\r\n]*)\]\([^\)\r\n]*\)/g;
|
|
210
|
+
const MARKDOWN_LINK_RE = /\[([^\]\r\n]*)\]\([^\)\r\n]*\)/g;
|
|
211
|
+
const MARKDOWN_HTML_TAG_RE = /<[^>\r\n]*>/g;
|
|
212
|
+
const MARKDOWN_HEADING_RE = /(^|\n)\s{0,3}#{1,6}\s+/g;
|
|
213
|
+
const MARKDOWN_LIST_RE = /(^|\n)\s*(?:[-*+]|\d+[.)])\s+/g;
|
|
214
|
+
const MARKDOWN_BLOCKQUOTE_RE = /(^|\n)\s*>\s?/g;
|
|
215
|
+
const MARKDOWN_INLINE_FORMAT_RE = /`+|[*_~]{1,3}/g;
|
|
216
|
+
const MARKDOWN_TABLE_SEP_RE = /[|]/g;
|
|
139
217
|
function removeTtsPaths(value) {
|
|
140
218
|
return value.replace(TTS_WINDOWS_PATH_PATTERN, " ").replace(TTS_RELATIVE_PATH_PATTERN, " ").replace(TTS_PROJECT_PATH_PATTERN, " ").replace(TTS_UNIX_PATH_PATTERN, " ");
|
|
141
219
|
}
|
|
142
220
|
function removeTtsMarkup(value) {
|
|
143
|
-
let text = value.replace(
|
|
221
|
+
let text = value.replace(MARKDOWN_CODE_BLOCK_RE, " ").replace(MARKDOWN_INDENTED_CODE_RE, "$1").replace(MARKDOWN_IMAGE_RE, " ").replace(MARKDOWN_LINK_RE, "$1").replace(MARKDOWN_HTML_TAG_RE, " ").replace(MARKDOWN_HEADING_RE, "$1").replace(MARKDOWN_LIST_RE, "$1").replace(MARKDOWN_BLOCKQUOTE_RE, "$1").replace(MARKDOWN_INLINE_FORMAT_RE, "").replace(MARKDOWN_TABLE_SEP_RE, " ");
|
|
144
222
|
text = text.replace(TTS_URL_PATTERN, " ");
|
|
145
223
|
return removeTtsPaths(text);
|
|
146
224
|
}
|
|
147
225
|
function removeTtsSymbols(value) {
|
|
148
226
|
return value.replace(/[\u0000-\u0008\u000B\u000C\u000E-\u001F\u007F-\u009F\u200B-\u200F\u202A-\u202E\u2060-\u2064\u2066-\u206F\uFEFF]/g, " ").replace(/(?:\p{Extended_Pictographic}|\p{Emoji_Presentation}|\p{Regional_Indicator}|\p{Emoji_Modifier}|\p{So}|\uFE0E|\uFE0F|\u200D|\u20E3)/gu, " ");
|
|
149
227
|
}
|
|
228
|
+
/** Combined punctuation normalization: collapse adjacent punctuation, remove non-speech characters, and normalize Chinese punctuation. */
|
|
150
229
|
function normalizeTtsPunctuation(value) {
|
|
151
230
|
return value.replace(TTS_NON_SPEECH_PUNCTUATION, "").replace(/[,。!?;、() ]/gu, (character) => TTS_PUNCTUATION[character] ?? character).replace(/(^|\s)[,;:!?]+(?=\s|$)/g, "$1").replace(/([,;:])\s*([.!?])/g, "$2").replace(/\s+([,.;:!?])/g, "$1");
|
|
152
231
|
}
|
|
@@ -160,24 +239,123 @@ function prepareTtsText(value) {
|
|
|
160
239
|
if (value.length === 0) return "";
|
|
161
240
|
return normalizeTtsPunctuation(removeTtsSymbols(removeTtsMarkup(value).replace(/\\[rn]|\/n/gi, " ").replace(/\r\n?|\n/g, "."))).replace(/\s+/g, " ").trim();
|
|
162
241
|
}
|
|
163
|
-
/**
|
|
242
|
+
/** Locate fenced code blocks in raw streaming text so their content is never read aloud. */
|
|
243
|
+
function findFencedCodeRegions(value) {
|
|
244
|
+
const regions = [];
|
|
245
|
+
let pendingStart = null;
|
|
246
|
+
const fence = /(^|\n)[ \t]{0,3}(`{3,}|~{3,})[^\r\n]*/g;
|
|
247
|
+
let open = null;
|
|
248
|
+
for (const match of value.matchAll(fence)) {
|
|
249
|
+
const lineStart = match.index + (match[1] === "\n" ? 1 : 0);
|
|
250
|
+
const char = match[2][0];
|
|
251
|
+
if (open === null) open = {
|
|
252
|
+
start: lineStart,
|
|
253
|
+
char
|
|
254
|
+
};
|
|
255
|
+
else if (open.char === char) {
|
|
256
|
+
regions.push([open.start, match.index + match[0].length]);
|
|
257
|
+
open = null;
|
|
258
|
+
}
|
|
259
|
+
}
|
|
260
|
+
if (open !== null) pendingStart = open.start;
|
|
261
|
+
return {
|
|
262
|
+
regions,
|
|
263
|
+
pendingStart
|
|
264
|
+
};
|
|
265
|
+
}
|
|
266
|
+
/** Split an accumulated model delta at completed sentence-ending punctuation, skipping fenced code blocks whole. */
|
|
164
267
|
function splitCompletedTtsSentences(value) {
|
|
165
|
-
const
|
|
268
|
+
const { regions, pendingStart } = findFencedCodeRegions(value);
|
|
269
|
+
const limit = pendingStart ?? value.length;
|
|
166
270
|
const boundary = /[。!?!?;;\n]+(?:[”’)】》〕\]}'"]*\s*)/gu;
|
|
167
|
-
|
|
168
|
-
|
|
169
|
-
|
|
170
|
-
|
|
171
|
-
|
|
271
|
+
const sentences = [];
|
|
272
|
+
let sentenceStart = 0;
|
|
273
|
+
let consumed = 0;
|
|
274
|
+
const splitSegment = (from, to) => {
|
|
275
|
+
boundary.lastIndex = from;
|
|
276
|
+
let match;
|
|
277
|
+
while ((match = boundary.exec(value)) !== null) {
|
|
278
|
+
const end = match.index + match[0].length;
|
|
279
|
+
if (end > to) break;
|
|
280
|
+
const sentence = value.slice(sentenceStart, end);
|
|
281
|
+
if (sentence.trim().length > 0) sentences.push(sentence);
|
|
282
|
+
sentenceStart = end;
|
|
283
|
+
consumed = end;
|
|
284
|
+
}
|
|
285
|
+
};
|
|
286
|
+
let pos = 0;
|
|
287
|
+
for (const [start, end] of regions) {
|
|
288
|
+
if (start >= limit) break;
|
|
289
|
+
if (pos < start) splitSegment(pos, Math.min(start, limit));
|
|
290
|
+
sentenceStart = Math.max(sentenceStart, end);
|
|
291
|
+
consumed = Math.max(consumed, end);
|
|
292
|
+
pos = end;
|
|
172
293
|
}
|
|
294
|
+
if (pos < limit) splitSegment(pos, limit);
|
|
295
|
+
if (pendingStart !== null) return {
|
|
296
|
+
sentences,
|
|
297
|
+
remainder: value.slice(pendingStart),
|
|
298
|
+
consumed: pendingStart,
|
|
299
|
+
inCode: true
|
|
300
|
+
};
|
|
173
301
|
return {
|
|
174
302
|
sentences,
|
|
175
|
-
remainder: value.slice(
|
|
303
|
+
remainder: value.slice(consumed),
|
|
304
|
+
consumed,
|
|
305
|
+
inCode: false
|
|
176
306
|
};
|
|
177
307
|
}
|
|
178
308
|
/** Count text-bearing characters only, excluding punctuation and whitespace. */
|
|
179
309
|
function countTtsSpeechCharacters(value) {
|
|
180
|
-
return
|
|
310
|
+
return (value.match(/[\p{L}\p{N}]/gu) ?? []).length;
|
|
311
|
+
}
|
|
312
|
+
/** Split cleaned speech text on natural boundaries while bounding VoiceDesign requests. */
|
|
313
|
+
function splitTtsSegments(value, target = DEFAULT_TTS_SEGMENT_CHARACTERS, maximum = MAX_TTS_SEGMENT_CHARACTERS) {
|
|
314
|
+
const text = prepareTtsText(value);
|
|
315
|
+
if (text.length === 0) return [];
|
|
316
|
+
const preferred = Math.max(MIN_TTS_SEGMENT_CHARACTERS, Math.min(target, maximum));
|
|
317
|
+
const pieces = text.match(/[^.!?;]+[.!?;]*|[.!?;]+/gu) ?? [text];
|
|
318
|
+
const segments = [];
|
|
319
|
+
let pending = "";
|
|
320
|
+
const append = (piece) => {
|
|
321
|
+
const candidate = `${pending}${piece}`.trim();
|
|
322
|
+
if (candidate.length === 0) return;
|
|
323
|
+
if (countTtsSpeechCharacters(candidate) <= maximum) {
|
|
324
|
+
pending = candidate;
|
|
325
|
+
if (countTtsSpeechCharacters(pending) >= preferred) {
|
|
326
|
+
segments.push(pending);
|
|
327
|
+
pending = "";
|
|
328
|
+
}
|
|
329
|
+
return;
|
|
330
|
+
}
|
|
331
|
+
if (pending.length > 0) {
|
|
332
|
+
segments.push(pending);
|
|
333
|
+
pending = "";
|
|
334
|
+
}
|
|
335
|
+
let rest = piece.trim();
|
|
336
|
+
while (countTtsSpeechCharacters(rest) > maximum) {
|
|
337
|
+
let cut = Math.min(rest.length, maximum);
|
|
338
|
+
const boundary = Math.max(rest.lastIndexOf(",", cut), rest.lastIndexOf(" ", cut), rest.lastIndexOf(":", cut));
|
|
339
|
+
if (boundary >= MIN_TTS_SEGMENT_CHARACTERS) cut = boundary + 1;
|
|
340
|
+
segments.push(rest.slice(0, cut).trim());
|
|
341
|
+
rest = rest.slice(cut).trim();
|
|
342
|
+
}
|
|
343
|
+
pending = rest;
|
|
344
|
+
};
|
|
345
|
+
for (const piece of pieces) append(piece);
|
|
346
|
+
if (pending.length > 0) {
|
|
347
|
+
const last = segments.at(-1);
|
|
348
|
+
if (last !== void 0 && countTtsSpeechCharacters(pending) < MIN_TTS_SEGMENT_CHARACTERS && countTtsSpeechCharacters(`${last}${pending}`) <= maximum) segments[segments.length - 1] = `${last}${pending}`;
|
|
349
|
+
else segments.push(pending);
|
|
350
|
+
}
|
|
351
|
+
return segments;
|
|
352
|
+
}
|
|
353
|
+
/** Return the exact decoded size of canonical padded Base64, or null when invalid. */
|
|
354
|
+
function strictBase64DecodedLength(value) {
|
|
355
|
+
if (value.length === 0 || value.length % 4 !== 0) return null;
|
|
356
|
+
if (!/^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/u.test(value)) return null;
|
|
357
|
+
const padding = value.endsWith("==") ? 2 : value.endsWith("=") ? 1 : 0;
|
|
358
|
+
return value.length / 4 * 3 - padding;
|
|
181
359
|
}
|
|
182
360
|
/** Accumulate short sentences so a PCM stream has enough text to sound natural. */
|
|
183
361
|
function batchTtsStreamText(pending, next, flush) {
|
|
@@ -267,11 +445,15 @@ const DEFAULT_TTS_SETTINGS = {
|
|
|
267
445
|
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
446
|
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
447
|
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: "
|
|
448
|
+
format: "pcm",
|
|
449
|
+
voiceDesignPlaybackMode: "complete",
|
|
271
450
|
autoPlay: true,
|
|
272
451
|
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
452
|
maxTextLength: 12e3,
|
|
274
|
-
requestTimeoutMs: 12e4
|
|
453
|
+
requestTimeoutMs: 12e4,
|
|
454
|
+
maxMp3AudioBytes: DEFAULT_MAX_MP3_AUDIO_BYTES,
|
|
455
|
+
maxWavAudioBytes: DEFAULT_MAX_WAV_AUDIO_BYTES,
|
|
456
|
+
maxPausedPcmBytes: DEFAULT_MAX_PAUSED_PCM_BYTES
|
|
275
457
|
};
|
|
276
458
|
/** Select the Xiaomi endpoint from the API key while preserving custom settings for other keys. */
|
|
277
459
|
function resolveTtsBaseURL(apiKey, configuredBaseURL) {
|
|
@@ -297,4 +479,4 @@ function resolveTtsSettings(value) {
|
|
|
297
479
|
}
|
|
298
480
|
|
|
299
481
|
//#endregion
|
|
300
|
-
export {
|
|
482
|
+
export { isNewerTtsVersion as A, TTS_VOICE_DESIGN_ASSET_ROUTE as C, batchTtsStreamText as D, TTS_VOICE_PRESETS as E, resolveTtsSettings as F, splitCompletedTtsSentences as I, splitTtsSegments as L, parseSseRecords as M, prepareTtsText as N, classifyLiveSpeechTransition as O, resolveTtsBaseURL as P, strictBase64DecodedLength as R, TTS_VOICE_ASSET_ROUTE as S, TTS_VOICE_DESIGN_PRESETS as T, TTS_STREAM_ROUTE as _, DEFAULT_TTS_SEGMENT_CHARACTERS as a, TTS_VERSION as b, MIN_TTS_SEGMENT_CHARACTERS as c, TTS_API_KEY_STATUS_ROUTE as d, TTS_AUDIO_RESPONSE_JSON_OVERHEAD_BYTES as f, TTS_SETTINGS_NAMESPACE as g, TTS_ROUTE as h, DEFAULT_MAX_WAV_AUDIO_BYTES as i, isSupportedTtsApiKey as j, countTtsSpeechCharacters as k, MIN_TTS_STREAM_CHARACTERS as l, TTS_MODELS as m, DEFAULT_MAX_MP3_AUDIO_BYTES as n, DEFAULT_TTS_SETTINGS as o, TTS_FORMATS as p, DEFAULT_MAX_PAUSED_PCM_BYTES as r, MAX_TTS_SEGMENT_CHARACTERS as s, AbortableSentenceQueue as t, TOKEN_PLAN_TTS_BASE_URL as u, TTS_UNINSTALL_ROUTE as v, TTS_VOICE_DESIGN_PLAYBACK_MODES as w, TTS_VOICES as x, TTS_UPDATE_ROUTE as y };
|