univoice 0.7.0 → 0.8.0

package/README.md

@@ -269,7 +269,7 @@ const response = await tts.synthesize({ text: '你好' });
 import { createTTS } from 'univoice';
 
 const tts = createTTS({
-  provider: 'doubao' | 'openai' | 'minimax' | 'qwen' | 'gemini',
+  provider: 'doubao' | 'openai' | 'minimax' | 'qwen' | 'qwen-realtime' | 'gemini' | 'glm' | 'xfyun',
   // 通用配置
   apiKey?: string,
   baseUrl?: string,
@@ -314,7 +314,7 @@ const tts = createTTS({
 import { createASR } from 'univoice';
 
 const asr = createASR({
-  provider: 'doubao' | 'openai' | 'minimax' | 'qwen' | 'gemini',
+  provider: 'doubao' | 'openai' | 'minimax' | 'qwen' | 'gemini' | 'glm' | 'xfyun',
   apiKey?: string,
   baseUrl?: string,
   model?: string,
@@ -355,6 +355,7 @@ const asr = createASR({
 | OpenAI | `openai` | 待实现 | 待实现 | 待实现 | 待实现 |
 | MiniMax | `minimax` | - | - | - | - |
 | Gemini | `gemini` | 待实现 | 待实现 | 待实现 | 待实现 |
+| 科大讯飞 | `xfyun` | ✅ | ✅ | ✅ | ✅ |
 
 #### TTS 能力矩阵
 
@@ -366,6 +367,7 @@ const asr = createASR({
 | OpenAI | `openai` | 待实现 | 待实现 | 待实现 | 待实现 |
 | MiniMax | `minimax` | ✅ | ✅ | ✅ | ✅ |
 | Gemini | `gemini` | 待实现 | 待实现 | 待实现 | 待实现 |
+| 科大讯飞 | `xfyun` | ✅ | ✅ | ✅ | ✅ |
 
 #### 能力说明
 
@@ -474,6 +476,29 @@ const asr = createASR({
 });
 ```
 
+#### 科大讯飞
+
+```typescript
+const tts = createTTS({
+  provider: 'xfyun',
+  appId: process.env.XFYUN_APP_ID,
+  apiSecret: process.env.XFYUN_API_SECRET,
+  apiKey: process.env.XFYUN_API_KEY,
+  voice: 'x5_lingxiaoxuan_flow',
+  model: 'super-human-tts',
+  format: 'pcm',
+  sampleRate: 16000,
+});
+
+const asr = createASR({
+  provider: 'xfyun',
+  appId: process.env.XFYUN_APP_ID,
+  apiSecret: process.env.XFYUN_API_SECRET,
+  apiKey: process.env.XFYUN_API_KEY,
+  language: 'zh-CN',
+});
+```
+
 ---
 
 <!-- PERFORMANCE_TABLE_START -->
@@ -493,7 +518,7 @@ const asr = createASR({
 >
 > 如需评估服务商的真实性能，建议直接使用服务商官方 SDK 进行测试。
 
-> 生成时间: 2026/3/29 21:20:30
+> 生成时间: 2026/4/8 17:39:11
 
 > 环境: Node.js v24.14.0, darwin arm64
 
@@ -588,7 +613,7 @@ const asr = createASR({
 | 豆包 | seed-tts-1.0 | zh_male_lengkugege_emo_v2_mars_bigtts | ogg_opus | 16000 | 3 | 653 | 62 | 3254 | 3461 | 231 | 88.9 |
 | 豆包 | seed-tts-1.0 | zh_male_lengkugege_emo_v2_mars_bigtts | ogg_opus | 24000 | 3 | 670 | 102 | 3429 | 7919 | 2209 | 59.3 |
 | 豆包 | seed-tts-1.0 | zh_male_lengkugege_emo_v2_mars_bigtts | ogg_opus | 48000 | 3 | 698 | 65 | 3217 | 3603 | 200 | 85.8 |
-| 豆包 | seed-tts-2.0 | zh_female_vv_uranus_bigtts | pcm | 8000 | 3 | 491 | 65 | 3810 | 3859 | **29 🏆** | 74.6 |
+| 豆包 | seed-tts-2.0 | zh_female_vv_uranus_bigtts | pcm | 8000 | 3 | 491 | 65 | 3810 | 3859 | 29 | 74.6 |
 | 豆包 | seed-tts-2.0 | zh_female_vv_uranus_bigtts | pcm | 16000 | 3 | 500 | 66 | 3752 | 3832 | 43 | 75.5 |
 | 豆包 | seed-tts-2.0 | zh_female_vv_uranus_bigtts | pcm | 24000 | 3 | 510 | 65 | 3760 | 3783 | 50 | 76.3 |
 | 豆包 | seed-tts-2.0 | zh_female_vv_uranus_bigtts | pcm | 48000 | 3 | 560 | 64 | 3770 | 3835 | 43 | 75.4 |
@@ -645,6 +670,9 @@ const asr = createASR({
 | MiniMax | speech-01-turbo | male-qn-qingse | pcm | 32000 | 3 | 490 | 2 | 2101 | 2148 | 99 | 138.6 |
 | MiniMax | speech-01-turbo | male-qn-qingse | pcm | 44100 | 3 | 448 | **2 🏆** | 1920 | 2100 | 99 | 145.2 |
 | 智谱 GLM | glm-tts | tongtong | pcm | 24000 | 3 | 861 | *542* | 5037 | 5614 | 296 | 54.8 |
+| 科大讯飞 | super-human-tts | x5_lingxiaoxuan_flow | pcm | 8000 | 3 | 543 | 17 | 2704 | 2897 | 119 | 104.1 |
+| 科大讯飞 | super-human-tts | x5_lingxiaoxuan_flow | pcm | 16000 | 3 | 511 | 17 | 2629 | 2721 | 51 | 107.5 |
+| 科大讯飞 | super-human-tts | x5_lingxiaoxuan_flow | pcm | 24000 | 3 | 592 | 16 | 2690 | 2709 | **28 🏆** | 106.3 |
 
 ## ASR 性能指标
 
@@ -675,13 +703,14 @@ const asr = createASR({
 
 | 服务商 | 模型 | 语言 | 输入格式 | 采样率 (Hz) | 测试次数 | 首包延迟 (ms) | 平均间隔 (ms) | P50 (ms) | P95 (ms) | 标准差 (ms) | RTF |
 |--------|------|------|----------|-------------|----------|---------------|---------------|----------|----------|-------------|-----|
-| 通义千问 | paraformer-realtime-v2 | zh-CN | pcm | 16000 | 3 | *978* | *82* | 685 | *2085* | *666* | *1.32* |
-| 通义千问 | paraformer-realtime-v1 | zh-CN | pcm | 16000 | 3 | **439 🏆** | **29 🏆** | **498 🏆** | **509 🏆** | **10 🏆** | **0.57 🏆** |
-| 豆包 | bigmodel | zh-CN | pcm | 16000 | 3 | 513 | 69 | *904* | 960 | 107 | 0.99 |
+| 通义千问 | paraformer-realtime-v2 | zh-CN | pcm | 16000 | 3 | 978 | 82 | 685 | 2085 | *666* | *1.32* |
+| 通义千问 | paraformer-realtime-v1 | zh-CN | pcm | 16000 | 3 | **439 🏆** | **29 🏆** | **498 🏆** | **509 🏆** | **10 🏆** | 0.57 |
+| 豆包 | bigmodel | zh-CN | pcm | 16000 | 3 | 513 | 69 | 904 | 960 | 107 | 0.99 |
+| 科大讯飞 | iat | zh-CN | pcm | 16000 | 3 | *1551* | *927* | *2835* | *2948* | 587 | **0.12 🏆** |
 
 ---
 
-*数据更新于: 2026-03-29*
+*数据更新于: 2026-04-08*
 
 <!-- PERFORMANCE_TABLE_END -->
 
@@ -756,26 +785,58 @@ src/
 ├── tts/               # TTS 模块
 │   ├── base.ts        # BaseTTS 抽象类
 │   ├── factory.ts     # 工厂函数
+│   ├── protocols/     # 协议实现
+│   │   ├── volcengine.ts
+│   │   ├── dashscope.ts
+│   │   ├── dashscope-realtime.ts
+│   │   ├── minimax.ts
+│   │   └── xfyun.ts
 │   ├── utils/         # 工具函数
-│   │   ├── save.ts    # 保存音频
-│   │   ├── collect.ts # 收集音频
-│   │   ├── play.ts    # 播放音频
-│   │   └── tee.ts     # 同时保存和播放
+│   │   ├── save.ts
+│   │   ├── save-audio.ts
+│   │   ├── collect.ts
+│   │   ├── play.ts
+│   │   └── tee.ts
 │   └── providers/     # 提供商实现
 │       ├── doubao.ts
 │       ├── openai.ts
 │       ├── minimax.ts
 │       ├── qwen.ts
-│       └── gemini.ts
+│       ├── qwen-realtime.ts
+│       ├── gemini.ts
+│       ├── glm.ts
+│       └── xfyun.ts
 ├── asr/               # ASR 模块
 │   ├── base.ts        # BaseASR 抽象类
 │   ├── factory.ts     # 工厂函数
+│   ├── protocols/     # 协议实现
+│   │   ├── dashscope.ts
+│   │   ├── sauc.ts
+│   │   └── xfyun.ts
 │   ├── utils/         # 工具函数
+│   │   ├── audio.ts
+│   │   ├── collect.ts
+│   │   ├── save.ts
+│   │   ├── ogg-muxer.ts
+│   │   └── opus-decode.ts
 │   └── providers/     # 提供商实现
+│       ├── doubao.ts
+│       ├── openai.ts
+│       ├── minimax.ts
+│       ├── qwen.ts
+│       ├── gemini.ts
+│       ├── glm.ts
+│       └── xfyun.ts
 └── types/             # 类型定义
-    ├── tts.ts         # TTS 相关类型
-    ├── asr.ts         # ASR 相关类型
-    └── llm-stream.ts  # LLM 流式输出类型
+    ├── index.ts
+    ├── tts.ts
+    ├── asr.ts
+    ├── llm-stream.ts
+    └── voices/
+        ├── doubao.ts
+        ├── minimax.ts
+        ├── qwen.ts
+        └── glm.ts
 ```
 
 ---
@@ -801,3 +862,4 @@ src/
 - [阿里云通义千问](https://tongyi.aliyun.com/)
 - [Google Gemini](https://ai.google.dev/)
 - [智谱 AI](https://open.bigmodel.cn/)
+- [科大讯飞](https://www.xfyun.cn/)

package/dist/{base--a8Bo2f0.d.ts → base-Bw2Puefv.d.ts}

@@ -39,6 +39,24 @@ interface GlmASROptions extends BaseASROptions {
     hotwords?: string[];
     context?: string;
 }
+interface XfyunASROptions extends BaseASROptions {
+    appId?: string;
+    apiSecret?: string;
+    sampleRate?: number;
+    domain?: string;
+    accent?: string;
+    eos?: number;
+    dwa?: string;
+    ltc?: number;
+    dhw?: string;
+    ptt?: number;
+    rlang?: string;
+    vinfo?: number;
+    nunum?: number;
+    nbest?: number;
+    wbest?: number;
+    sendInterval?: number;
+}
 type ASROptions = ({
     provider: 'doubao';
 } & DoubaoASROptions) | ({
@@ -52,6 +70,8 @@ type ASROptions = ({
 } & BaseASROptions) | ({
     provider: 'gemini';
 } & BaseASROptions) | ({
+    provider: 'xfyun';
+} & XfyunASROptions) | ({
     provider: string;
 } & BaseASROptions);
 interface ListenInstanceOptions {
@@ -85,7 +105,7 @@ interface ASRProvider {
     name: string;
     listenStream(audio: AudioStream): AsyncIterable<ASRStreamChunk>;
 }
-type ASRProviderType = 'doubao' | 'minimax' | 'qwen' | 'openai' | 'gemini' | string;
+type ASRProviderType = 'doubao' | 'minimax' | 'qwen' | 'openai' | 'gemini' | 'xfyun' | string;
 type AudioStream = AsyncIterable<Buffer | Uint8Array>;
 type AudioStreamInput = AudioStream | Buffer | Uint8Array | string;
 type ASRConnectionState = 'connected' | 'closed' | 'error';
@@ -131,4 +151,4 @@ declare abstract class BaseASR {
     }): Promise<ASRResponse>;
 }
 
-export { type ASRConnectOptions as A, BaseASR as B, type DoubaoASROptions as D, type GlmASROptions as G, type ListenInstanceOptions as L, type QwenASROptions as Q, type ASRConnection as a, type ASRConnectionState as b, type ASROptions as c, type ASRProvider as d, type ASRProviderType as e, type ASRRequest as f, type ASRResponse as g, type ASRSegment as h, type ASRStreamChunk as i, type AudioCodecFormat as j, type AudioContainerFormat as k, type AudioFormat as l, type AudioStream as m, type AudioStreamInput as n, type BaseASROptions as o };
+export { type ASRConnectOptions as A, BaseASR as B, type DoubaoASROptions as D, type GlmASROptions as G, type ListenInstanceOptions as L, type QwenASROptions as Q, type XfyunASROptions as X, type ASRConnection as a, type ASRConnectionState as b, type ASROptions as c, type ASRProvider as d, type ASRProviderType as e, type ASRRequest as f, type ASRResponse as g, type ASRSegment as h, type ASRStreamChunk as i, type AudioCodecFormat as j, type AudioContainerFormat as k, type AudioFormat as l, type AudioStream as m, type AudioStreamInput as n, type BaseASROptions as o };

package/dist/{base-Bae_riVx.d.ts → base-DhiS1mCx.d.ts}

@@ -96,6 +96,19 @@ interface QwenRealtimeTTSOptions extends BaseTTSOptions {
 interface GlmTTSOptions extends BaseTTSOptions {
     voice?: AcceptAnyString<GlmVoice>;
 }
+interface XfyunTTSOptions extends BaseTTSOptions {
+    appId?: string;
+    apiSecret?: string;
+    sampleRate?: number;
+    oralLevel?: 'high' | 'mid' | 'low';
+    sparkAssist?: number;
+    stopSplit?: number;
+    remain?: number;
+    reg?: number;
+    rdn?: number;
+    rhy?: number;
+    bgs?: number;
+}
 type TTSOptions = ({
     provider: 'doubao';
 } & DoubaoTTSOptions) | ({
@@ -111,6 +124,8 @@ type TTSOptions = ({
 } & BaseTTSOptions) | ({
     provider: 'glm';
 } & GlmTTSOptions) | ({
+    provider: 'xfyun';
+} & XfyunTTSOptions) | ({
     provider: string;
 } & BaseTTSOptions);
 interface TTSRequest {
@@ -141,7 +156,7 @@ interface TTSVoice {
     language: string;
     gender?: 'male' | 'female' | 'neutral';
 }
-type TTSProviderType = 'doubao' | 'minimax' | 'qwen' | 'openai' | 'gemini' | string;
+type TTSProviderType = 'doubao' | 'minimax' | 'qwen' | 'openai' | 'gemini' | 'xfyun' | string;
 type TTSConnectionState = 'connected' | 'closed' | 'error';
 interface TTSConnectOptions {
     timeout?: number;
@@ -188,4 +203,4 @@ declare abstract class BaseTTS implements TTSProvider {
     };
 }
 
-export { BaseTTS as B, type CosyVoiceV1Voice as C, type DoubaoTTSOptions as D, type GlmTTSOptions as G, type MinimaxTTSOptions as M, type OpenAIChatCompletionChunk as O, type QwenRealtimeOptions as Q, type SpeakInstanceOptions as S, type TTSConnectOptions as T, type BaseTTSOptions as a, type QwenRealtimeTTSOptions as b, type QwenTTSOptions as c, type TTSConnection as d, type TTSConnectionState as e, type TTSOptions as f, type TTSProvider as g, type TTSProviderType as h, type TTSRequest as i, type TTSResponse as j, type TTSStreamChunk as k, type TTSVoice as l, type TextStream as m, type CosyVoiceV2Voice as n, type CosyVoiceV3FlashVoice as o, type CosyVoiceV3PlusVoice as p, type CosyVoiceVoice as q, type DoubaoJupiterVoice as r, type DoubaoV1Voice as s, type DoubaoV2Voice as t, type DoubaoVoice as u, type MinimaxVoice as v, type OpenAIStream as w, type QwenRealtimeVoice as x, type QwenTTSModel as y };
+export { BaseTTS as B, type CosyVoiceV1Voice as C, type DoubaoTTSOptions as D, type GlmTTSOptions as G, type MinimaxTTSOptions as M, type OpenAIChatCompletionChunk as O, type QwenRealtimeOptions as Q, type SpeakInstanceOptions as S, type TTSConnectOptions as T, type XfyunTTSOptions as X, type BaseTTSOptions as a, type QwenRealtimeTTSOptions as b, type QwenTTSOptions as c, type TTSConnection as d, type TTSConnectionState as e, type TTSOptions as f, type TTSProvider as g, type TTSProviderType as h, type TTSRequest as i, type TTSResponse as j, type TTSStreamChunk as k, type TTSVoice as l, type TextStream as m, type CosyVoiceV2Voice as n, type CosyVoiceV3FlashVoice as o, type CosyVoiceV3PlusVoice as p, type CosyVoiceVoice as q, type DoubaoJupiterVoice as r, type DoubaoV1Voice as s, type DoubaoV2Voice as t, type DoubaoVoice as u, type MinimaxVoice as v, type OpenAIStream as w, type QwenRealtimeVoice as x, type QwenTTSModel as y };

package/dist/{chunk-RRXTYGBU.js → chunk-2MN6RZSS.js}

@@ -1,4 +1,4 @@
-import { __name } from './chunk-7QVYU63E.js';
+import { __name, buildAuthUrl } from './chunk-R66GH6Y5.js';
 import { Buffer } from 'buffer';
 import { randomUUID } from 'crypto';
 import WebSocket from 'ws';
@@ -3142,6 +3142,376 @@ var QwenRealtimeTTSConnection = class {
     return this.provider.synthesizeOnConnection(this.ws, textChunks.join(""));
   }
 };
+function buildTTSAuthUrl(apiKey, apiSecret) {
+  return buildAuthUrl("cbm01.cn-huabei-1.xf-yun.com", "/v1/private/mcd9m97e6", apiKey, apiSecret);
+}
+__name(buildTTSAuthUrl, "buildTTSAuthUrl");
+function mapAudioEncoding(format) {
+  const encodingMap = {
+    mp3: "lame",
+    pcm: "raw",
+    opus: "opus"
+  };
+  return encodingMap[format] || "lame";
+}
+__name(mapAudioEncoding, "mapAudioEncoding");
+function createRequestPayload(options, text, status, seq) {
+  const payload = {
+    header: {
+      app_id: options.appId,
+      status
+    },
+    parameter: {
+      tts: {
+        vcn: options.vcn,
+        speed: options.speed,
+        volume: options.volume,
+        pitch: options.pitch,
+        bgs: options.bgs ?? 0,
+        reg: options.reg ?? 0,
+        rdn: options.rdn ?? 0,
+        rhy: options.rhy ?? 0,
+        audio: {
+          encoding: options.encoding,
+          sample_rate: options.sampleRate,
+          channels: 1,
+          bit_depth: 16,
+          frame_size: 0
+        }
+      }
+    },
+    payload: {
+      text: {
+        encoding: "utf8",
+        compress: "raw",
+        format: "plain",
+        status,
+        seq,
+        text: Buffer.from(text).toString("base64")
+      }
+    }
+  };
+  if (options.oralLevel || options.sparkAssist != null || options.stopSplit != null || options.remain != null) {
+    payload.parameter.oral = {
+      ...options.oralLevel ? { oral_level: options.oralLevel } : {},
+      ...options.sparkAssist != null ? { spark_assist: options.sparkAssist } : {},
+      ...options.stopSplit != null ? { stop_split: options.stopSplit } : {},
+      ...options.remain != null ? { remain: options.remain } : {}
+    };
+  }
+  return JSON.stringify(payload);
+}
+__name(createRequestPayload, "createRequestPayload");
+function parseTTSResponse(data) {
+  let text;
+  if (Buffer.isBuffer(data)) {
+    text = data.toString("utf8");
+  } else if (data instanceof ArrayBuffer) {
+    text = new TextDecoder().decode(data);
+  } else if (Array.isArray(data)) {
+    text = Buffer.concat(data).toString("utf8");
+  } else {
+    text = String(data);
+  }
+  return JSON.parse(text);
+}
+__name(parseTTSResponse, "parseTTSResponse");
+function extractAudioFromResponse(response) {
+  return response.payload?.audio?.audio ?? null;
+}
+__name(extractAudioFromResponse, "extractAudioFromResponse");
+function isTTSSuccessResponse(response) {
+  return response.header.code === 0;
+}
+__name(isTTSSuccessResponse, "isTTSSuccessResponse");
+function isTTSFinishedResponse(response) {
+  return response.header.status === 2;
+}
+__name(isTTSFinishedResponse, "isTTSFinishedResponse");
+
+// src/tts/providers/xfyun.ts
+var XfyunTTS = class extends BaseTTS {
+  static {
+    __name(this, "XfyunTTS");
+  }
+  name = "xfyun";
+  /** 讯飞 AppID */
+  appId;
+  /** 讯飞 APISecret（用于 HMAC-SHA256 签名） */
+  apiSecret;
+  /** 音频采样率 */
+  sampleRate;
+  /** 口语化等级（仅 x4 系列发音人支持） */
+  oralLevel;
+  /** 是否通过大模型进行口语化（仅 x4 系列发音人支持） */
+  sparkAssist;
+  /** 是否关闭服务端拆句（仅 x4 系列发音人支持） */
+  stopSplit;
+  /** 是否保留原书面语（仅 x4 系列发音人支持） */
+  remain;
+  /** 英文发音方式 */
+  reg;
+  /** 数字发音方式 */
+  rdn;
+  /** 是否返回拼音标注 */
+  rhy;
+  /** 背景音 */
+  bgs;
+  constructor(options) {
+    super(options);
+    this.appId = options.appId || "";
+    this.apiSecret = options.apiSecret || "";
+    this.sampleRate = options.sampleRate ?? 24e3;
+    this.oralLevel = options.oralLevel;
+    this.sparkAssist = options.sparkAssist;
+    this.stopSplit = options.stopSplit;
+    this.remain = options.remain;
+    this.reg = options.reg;
+    this.rdn = options.rdn;
+    this.rhy = options.rhy;
+    this.bgs = options.bgs;
+    this.voice = options.voice || "x5_lingxiaoxuan_flow";
+    this.format = options.format || "mp3";
+  }
+  /**
+   * 将 BaseTTS 的 speed/volume/pitch (0-2 范围) 映射为讯飞的 0-100 范围
+   * BaseTTS 默认 1.0 → xfyun 50
+   */
+  mapParam(value) {
+    return Math.round(value * 50);
+  }
+  /**
+   * 构建协议配置选项
+   */
+  buildProtocolOptions() {
+    return {
+      appId: this.appId,
+      vcn: this.voice,
+      speed: this.mapParam(this.speed),
+      volume: this.mapParam(this.volume),
+      pitch: this.mapParam(this.pitch),
+      encoding: mapAudioEncoding(this.format),
+      sampleRate: this.sampleRate,
+      oralLevel: this.oralLevel,
+      sparkAssist: this.sparkAssist,
+      stopSplit: this.stopSplit,
+      remain: this.remain,
+      reg: this.reg,
+      rdn: this.rdn,
+      rhy: this.rhy,
+      bgs: this.bgs
+    };
+  }
+  /**
+   * 合并多个 Uint8Array
+   */
+  concatArrays(arrays) {
+    const totalLength = arrays.reduce((sum, arr) => sum + arr.length, 0);
+    const result = new Uint8Array(totalLength);
+    let offset = 0;
+    for (const arr of arrays) {
+      result.set(arr, offset);
+      offset += arr.length;
+    }
+    return result;
+  }
+  /**
+   * 合成语音（非流式）
+   * 建立 WebSocket → 发送请求（status=2 一次性发送）→ 收集所有音频块 → 合并返回
+   */
+  async synthesize(request) {
+    if (!this.appId) {
+      throw new Error("appId is required for Xfyun TTS");
+    }
+    if (!this.apiKey) {
+      throw new Error("apiKey is required for Xfyun TTS");
+    }
+    if (!this.apiSecret) {
+      throw new Error("apiSecret is required for Xfyun TTS");
+    }
+    const protocolOptions = this.buildProtocolOptions();
+    const url = buildTTSAuthUrl(this.apiKey, this.apiSecret);
+    const ws = new WebSocket(url);
+    await new Promise((resolve, reject) => {
+      ws.on("open", resolve);
+      ws.on("error", reject);
+    });
+    try {
+      const payload = createRequestPayload(protocolOptions, request.text, 2, 0);
+      ws.send(payload);
+      const audioChunks = [];
+      await new Promise((resolve, reject) => {
+        ws.on("message", (data) => {
+          try {
+            const response = parseTTSResponse(data);
+            if (!isTTSSuccessResponse(response)) {
+              reject(
+                new Error(`Xfyun TTS error: ${response.header.code} - ${response.header.message}`)
+              );
+              return;
+            }
+            const audioBase64 = extractAudioFromResponse(response);
+            if (audioBase64) {
+              audioChunks.push(Buffer.from(audioBase64, "base64"));
+            }
+            if (isTTSFinishedResponse(response)) {
+              resolve();
+            }
+          } catch (err) {
+            reject(err instanceof Error ? err : new Error(String(err)));
+          }
+        });
+        ws.on("error", reject);
+        ws.on("close", () => resolve());
+      });
+      const audio = this.concatArrays(audioChunks);
+      if (audio.length === 0) {
+        throw new Error("No audio received from Xfyun TTS service");
+      }
+      return {
+        audio: Buffer.from(audio),
+        format: this.format
+      };
+    } finally {
+      ws.close();
+    }
+  }
+  /**
+   * 流式语音合成（内部实现方法）
+   * 支持双向流式：流式文本输入，流式音频输出
+   *
+   * @param input 文本输入，可以是字符串或文本流
+   * @returns 流式音频块
+   * @internal
+   */
+  async *speakStream(input) {
+    if (!this.appId) {
+      throw new Error("appId is required for Xfyun TTS");
+    }
+    if (!this.apiKey) {
+      throw new Error("apiKey is required for Xfyun TTS");
+    }
+    if (!this.apiSecret) {
+      throw new Error("apiSecret is required for Xfyun TTS");
+    }
+    const textStream = normalizeTextStream(input);
+    const protocolOptions = this.buildProtocolOptions();
+    const url = buildTTSAuthUrl(this.apiKey, this.apiSecret);
+    const queue = [];
+    const syncState = { resolveWait: null, finished: false };
+    const enqueue = /* @__PURE__ */ __name((item) => {
+      queue.push(item);
+      syncState.resolveWait?.();
+      syncState.resolveWait = null;
+    }, "enqueue");
+    const ws = new WebSocket(url);
+    await new Promise((resolve, reject) => {
+      ws.on("open", resolve);
+      ws.on("error", reject);
+    });
+    const processPromise = (async () => {
+      try {
+        await Promise.all([
+          sendTextStream(ws, protocolOptions, textStream),
+          receiveAudioToQueue(ws, enqueue)
+        ]);
+      } catch (error) {
+        enqueue({
+          type: "error",
+          error: error instanceof Error ? error : new Error(String(error))
+        });
+      } finally {
+        syncState.finished = true;
+        syncState.resolveWait?.();
+        syncState.resolveWait = null;
+        ws.close();
+      }
+    })();
+    try {
+      while (!syncState.finished || queue.length > 0) {
+        while (queue.length === 0 && !syncState.finished) {
+          await new Promise((resolve) => {
+            syncState.resolveWait = resolve;
+          });
+        }
+        if (queue.length === 0) break;
+        const item = queue.shift();
+        if (!item) break;
+        switch (item.type) {
+          case "audio":
+            yield { audioChunk: item.chunk };
+            break;
+          case "error":
+            throw item.error;
+          case "end":
+            return;
+        }
+      }
+    } finally {
+      await processPromise.catch(() => {
+      });
+    }
+  }
+};
+async function sendTextStream(ws, protocolOptions, textStream) {
+  let seq = 0;
+  let isFirst = true;
+  for await (const chunk of textStream) {
+    if (!chunk) continue;
+    if (isFirst) {
+      const payload = createRequestPayload(protocolOptions, chunk, 0, seq);
+      ws.send(payload);
+      isFirst = false;
+    } else {
+      const payload = createRequestPayload(protocolOptions, chunk, 1, seq);
+      ws.send(payload);
+    }
+    seq++;
+  }
+  const endPayload = createRequestPayload(protocolOptions, "", 2, seq);
+  ws.send(endPayload);
+}
+__name(sendTextStream, "sendTextStream");
+async function receiveAudioToQueue(ws, enqueue) {
+  return new Promise((resolve, reject) => {
+    ws.on("message", (data) => {
+      try {
+        const response = parseTTSResponse(data);
+        if (!isTTSSuccessResponse(response)) {
+          enqueue({
+            type: "error",
+            error: new Error(
+              `Xfyun TTS error: ${response.header.code} - ${response.header.message}`
+            )
+          });
+          resolve();
+          return;
+        }
+        const audioBase64 = extractAudioFromResponse(response);
+        if (audioBase64) {
+          enqueue({ type: "audio", chunk: Buffer.from(audioBase64, "base64") });
+        }
+        if (isTTSFinishedResponse(response)) {
+          enqueue({ type: "end" });
+          resolve();
+        }
+      } catch (err) {
+        enqueue({
+          type: "error",
+          error: err instanceof Error ? err : new Error(String(err))
+        });
+        resolve();
+      }
+    });
+    ws.on("error", (err) => {
+      reject(err);
+    });
+    ws.on("close", () => {
+      resolve();
+    });
+  });
+}
+__name(receiveAudioToQueue, "receiveAudioToQueue");
 
 // src/tts/providers/index.ts
 registerTTSProvider("doubao", DoubaoTTS);
@@ -3151,7 +3521,8 @@ registerTTSProvider("qwen", QwenTTS);
 registerTTSProvider("qwen-realtime", QwenRealtimeTTS);
 registerTTSProvider("openai", TTS1);
 registerTTSProvider("gemini", GeminiTTS);
+registerTTSProvider("xfyun", XfyunTTS);
 
-export { BaseTTS, DoubaoTTS, GeminiTTS, GlmTTS, MinimaxTTS, QwenRealtimeTTS, QwenTTS, TTS1, createTTS, getTTSProviders, registerTTSProvider };
-//# sourceMappingURL=chunk-RRXTYGBU.js.map
-//# sourceMappingURL=chunk-RRXTYGBU.js.map
+export { BaseTTS, DoubaoTTS, GeminiTTS, GlmTTS, MinimaxTTS, QwenRealtimeTTS, QwenTTS, TTS1, XfyunTTS, createTTS, getTTSProviders, registerTTSProvider };
+//# sourceMappingURL=chunk-2MN6RZSS.js.map
+//# sourceMappingURL=chunk-2MN6RZSS.js.map