@mingxy/ocosay 1.0.0

package/src/providers/minimax.ts

@@ -0,0 +1,515 @@
+/**
+ * MiniMax TTS Provider
+ * 支持 T2A v2 (同步/流式HTTP)、T2A Async v2 (异步HTTP轮询)
+ * 端点可配置: baseURL (默认 https://api.minimaxi.com)
+ */
+
+import { BaseTTSProvider } from './base'
+import {
+  TTSError,
+  TTSErrorCode,
+  AudioResult,
+  Voice,
+  SpeakOptions
+} from '../core/types'
+import axios, { AxiosInstance } from 'axios'
+import { WebSocket } from 'ws'
+
+export interface MiniMaxConfig {
+  apiKey: string
+  baseURL?: string
+  voiceId?: string
+  model?: 'sync' | 'async' | 'stream'
+  ttsModel?: string
+  audioFormat?: 'mp3' | 'wav' | 'flac'
+  speed?: number
+  volume?: number
+  pitch?: number
+  groupId?: string
+}
+
+/**
+ * MiniMax TTS Provider
+ * 官网: https://www.minimax.io/
+ * API文档: https://www.minimaxi.com/document/T2A
+ */
+export class MiniMaxProvider extends BaseTTSProvider {
+  name = 'minimax'
+  capabilities = {
+    speak: true,
+    stream: true,
+    sync: true,
+    async: true,
+    voiceList: true,
+    voiceClone: true
+  } as const
+
+  private config: MiniMaxConfig
+  private httpClient: AxiosInstance
+  private wsConnection?: WebSocket
+  private currentAudioData: Buffer[] = []
+  private audioFormat: 'mp3' | 'wav' | 'flac' = 'mp3'
+
+  constructor(config: MiniMaxConfig) {
+    super()
+    this.config = config
+    this.apiKey = config.apiKey
+    this.defaultVoice = config.voiceId
+    this.defaultModel = config.model || 'stream'
+    this.audioFormat = config.audioFormat || 'mp3'
+
+    this.httpClient = axios.create({
+      baseURL: this.config.baseURL || 'https://api.minimaxi.com',
+      headers: {
+        'Authorization': `Bearer ${this.apiKey}`,
+        'Content-Type': 'application/json'
+      },
+      timeout: 30000
+    })
+  }
+
+  async initialize(): Promise<void> {
+    this.validateApiKey()
+  }
+
+  async destroy(): Promise<void> {
+    if (this.wsConnection) {
+      this.wsConnection.close()
+      this.wsConnection = undefined
+    }
+  }
+
+  protected async doSpeak(
+    text: string,
+    voice: string | undefined,
+    model: 'sync' | 'async' | 'stream',
+    options?: SpeakOptions
+  ): Promise<AudioResult> {
+    this.validateApiKey()
+
+    switch (model) {
+      case 'stream':
+        return this.streamingSpeak(text, voice, options)
+      case 'sync':
+        return this.syncSpeak(text, voice, options)
+      case 'async':
+        return this.asyncSpeak(text, voice, options)
+      default:
+        return this.streamingSpeak(text, voice, options)
+    }
+  }
+
+  /**
+   * 流式合成 (HTTP) - T2A v2 with stream: true
+   */
+  private async streamingSpeak(
+    text: string,
+    voice: string | undefined,
+    options?: SpeakOptions
+  ): Promise<AudioResult> {
+    try {
+      const voiceId = voice || this.defaultVoice || 'male-qn-qingse'
+      const speed = options?.speed || this.config.speed || 1.0
+      const vol = options?.volume !== undefined ? options.volume / 10 : (this.config.volume !== undefined ? this.config.volume / 10 : 1.0)
+      const pitch = options?.pitch !== undefined ? Math.round((options.pitch - 1) * 12) : (this.config.pitch !== undefined ? Math.round((this.config.pitch - 1) * 12) : 0)
+
+      const response = await this.httpClient.post('/v1/t2a_v2', {
+        model: this.config.ttsModel || 'speech-2.8-hd',
+        text,
+        stream: true,
+        voice_setting: {
+          voice_id: voiceId,
+          speed: speed,
+          vol: vol,
+          pitch: pitch
+        },
+        audio_setting: {
+          sample_rate: 32000,
+          bitrate: 128000,
+          format: this.audioFormat,
+          channel: 1
+        }
+      }, {
+        headers: {
+          'Authorization': `Bearer ${this.apiKey}`,
+          'Content-Type': 'application/json'
+        },
+        responseType: 'stream'
+      })
+
+      const stream = response.data
+      const audioChunks: Buffer[] = []
+
+      return new Promise((resolve, reject) => {
+        stream.on('data', (chunk: Buffer) => {
+          try {
+            const lines = chunk.toString().split('\n')
+            for (const line of lines) {
+              if (line.startsWith('data:')) {
+                const data = JSON.parse(line.slice(5))
+                if (data.data?.audio) {
+                  audioChunks.push(Buffer.from(data.data.audio, 'hex'))
+                }
+                if (data.data?.status === 2) {
+                  const fullAudio = Buffer.concat(audioChunks)
+                  resolve({
+                    audioData: fullAudio,
+                    format: this.audioFormat,
+                    isStream: true,
+                    duration: this.estimateDuration(fullAudio.length)
+                  })
+                }
+              }
+            }
+          } catch (e) {
+            audioChunks.push(chunk)
+          }
+        })
+
+        stream.on('error', (err: any) => {
+          reject(new TTSError(
+            'Stream error',
+            TTSErrorCode.NETWORK,
+            this.name,
+            err
+          ))
+        })
+
+        stream.on('end', () => {
+          if (audioChunks.length > 0) {
+            const fullAudio = Buffer.concat(audioChunks)
+            resolve({
+              audioData: fullAudio,
+              format: this.audioFormat,
+              isStream: true,
+              duration: this.estimateDuration(fullAudio.length)
+            })
+          }
+        })
+      })
+    } catch (error: any) {
+      if (error instanceof TTSError) throw error
+      throw this.mapError(error)
+    }
+  }
+
+  /**
+   * 同步合成 (HTTP) - T2A v2
+   * API: POST https://api.minimax.io/v1/t2a_v2
+   */
+  private async syncSpeak(
+    text: string,
+    voice: string | undefined,
+    options?: SpeakOptions
+  ): Promise<AudioResult> {
+    try {
+      const voiceId = voice || this.defaultVoice || 'male-qn-qingse'
+      const speed = options?.speed || this.config.speed || 1.0
+      const vol = options?.volume !== undefined ? options.volume / 10 : (this.config.volume !== undefined ? this.config.volume / 10 : 1.0)
+      const pitch = options?.pitch !== undefined ? Math.round((options.pitch - 1) * 12) : (this.config.pitch !== undefined ? Math.round((this.config.pitch - 1) * 12) : 0)
+
+      const response = await this.httpClient.post('/v1/t2a_v2', {
+        model: this.config.ttsModel || 'speech-2.8-hd',
+        text,
+        stream: false,
+        output_format: 'hex',
+        voice_setting: {
+          voice_id: voiceId,
+          speed: speed,
+          vol: vol,
+          pitch: pitch
+        },
+        audio_setting: {
+          sample_rate: 32000,
+          bitrate: 128000,
+          format: this.audioFormat,
+          channel: 1
+        }
+      }, {
+        headers: {
+          'Authorization': `Bearer ${this.apiKey}`,
+          'Content-Type': 'application/json'
+        }
+      })
+
+      if (response.data.base_resp?.status_code !== 0) {
+        throw new TTSError(
+          response.data.base_resp?.status_msg || 'API request failed',
+          TTSErrorCode.UNKNOWN,
+          this.name,
+          response.data.base_resp
+        )
+      }
+
+      const audioHex = response.data.data?.audio
+      if (!audioHex) {
+        throw new TTSError(
+          'No audio data in response',
+          TTSErrorCode.UNKNOWN,
+          this.name
+        )
+      }
+
+      const audioBuffer = Buffer.from(audioHex, 'hex')
+
+      return {
+        audioData: audioBuffer,
+        format: this.audioFormat,
+        isStream: false,
+        duration: response.data.extra_info?.audio_length 
+          ? response.data.extra_info.audio_length / 1000 
+          : this.estimateDuration(audioBuffer.length)
+      }
+    } catch (error: any) {
+      if (error instanceof TTSError) throw error
+      throw this.mapError(error)
+    }
+  }
+
+  /**
+   * 异步合成 (轮询) - T2A Async v2
+   */
+  private async asyncSpeak(
+    text: string,
+    voice: string | undefined,
+    options?: SpeakOptions
+  ): Promise<AudioResult> {
+    try {
+      const voiceId = voice || this.defaultVoice || 'male-qn-qingse'
+      const speed = options?.speed || this.config.speed || 1.0
+      const vol = options?.volume !== undefined ? options.volume / 10 : (this.config.volume !== undefined ? this.config.volume / 10 : 1.0)
+      const pitch = options?.pitch !== undefined ? Math.round((options.pitch - 1) * 12) : (this.config.pitch !== undefined ? Math.round((this.config.pitch - 1) * 12) : 0)
+
+      const createResponse = await this.httpClient.post('/v1/t2a_async_v2', {
+        model: this.config.ttsModel || 'speech-2.8-hd',
+        text,
+        voice_setting: {
+          voice_id: voiceId,
+          speed: speed,
+          vol: vol,
+          pitch: pitch
+        },
+        audio_setting: {
+          sample_rate: 32000,
+          bitrate: 128000,
+          format: this.audioFormat,
+          channel: 1
+        }
+      }, {
+        headers: {
+          'Authorization': `Bearer ${this.apiKey}`,
+          'Content-Type': 'application/json'
+        }
+      })
+
+      const taskId = createResponse.data.task_id
+      if (!taskId) {
+        throw new TTSError(
+          'No task_id in async response',
+          TTSErrorCode.UNKNOWN,
+          this.name,
+          createResponse.data
+        )
+      }
+
+      let attempts = 0
+      const maxAttempts = 60
+
+      while (attempts < maxAttempts) {
+        await this.delay(2000)
+
+        const statusResponse = await this.httpClient.get(
+          `/v1/query/t2a_async_query_v2?task_id=${taskId}`,
+          {
+            headers: {
+              'Authorization': `Bearer ${this.apiKey}`
+            }
+          }
+        )
+
+        if (statusResponse.data.status === 'success') {
+          const fileId = statusResponse.data.file_id
+          if (!fileId) {
+            throw new TTSError(
+              'No file_id in async response',
+              TTSErrorCode.UNKNOWN,
+              this.name,
+              statusResponse.data
+            )
+          }
+
+          const downloadResponse = await this.httpClient.get(
+            `/v1/files/retrieve_content?file_id=${fileId}`,
+            { 
+              headers: {
+                'Authorization': `Bearer ${this.apiKey}`
+              },
+              responseType: 'arraybuffer'
+            }
+          )
+
+          return {
+            audioData: Buffer.from(downloadResponse.data),
+            format: this.audioFormat,
+            isStream: false,
+            duration: 0
+          }
+        }
+
+        if (statusResponse.data.status === 'failed') {
+          throw new TTSError(
+            'Async TTS task failed',
+            TTSErrorCode.UNKNOWN,
+            this.name,
+            statusResponse.data
+          )
+        }
+
+        attempts++
+      }
+
+      throw new TTSError(
+        'Async TTS task timeout',
+        TTSErrorCode.NETWORK,
+        this.name
+      )
+    } catch (error: any) {
+      if (error instanceof TTSError) throw error
+      throw this.mapError(error)
+    }
+  }
+
+  /**
+   * 音色克隆 - 使用参考音频克隆声音
+   */
+  async voiceClone(audioUrl: string, text: string, voice?: string): Promise<AudioResult> {
+    this.validateApiKey()
+
+    try {
+      const response = await this.httpClient.post('/v1/t2a_v2/voice_clone', {
+        audio_url: audioUrl,
+        text,
+        voice_id: voice || 'custom_clone'
+      }, {
+        headers: {
+          'Authorization': `Bearer ${this.apiKey}`,
+          'Content-Type': 'application/json'
+        },
+        responseType: 'arraybuffer'
+      })
+
+      return {
+        audioData: Buffer.from(response.data),
+        format: this.audioFormat,
+        isStream: false,
+        duration: this.estimateDuration(response.data.length)
+      }
+    } catch (error: any) {
+      throw this.mapError(error)
+    }
+  }
+
+  /**
+   * 获取音色列表
+   */
+  async listVoices(): Promise<Voice[]> {
+    this.validateApiKey()
+
+    try {
+      const response = await this.httpClient.get('/v1/t2a/voices', {
+        headers: {
+          'Authorization': `Bearer ${this.apiKey}`
+        }
+      })
+
+      return response.data.voices.map((v: any) => ({
+        id: v.voice_id,
+        name: v.name,
+        language: v.language,
+        gender: v.gender
+      }))
+    } catch (error: any) {
+      // 如果API调用失败，返回预定义音色列表
+      return MINIMAX_VOICES
+    }
+  }
+
+  /**
+   * 错误映射
+   */
+  private mapError(error: any): TTSError {
+    if (error.response) {
+      const status = error.response.status
+      const code =
+        status === 401 ? TTSErrorCode.AUTH :
+        status === 429 ? TTSErrorCode.QUOTA :
+        status >= 500 ? TTSErrorCode.NETWORK :
+        TTSErrorCode.UNKNOWN
+
+      return new TTSError(
+        error.response.data?.message || 'API request failed',
+        code,
+        this.name,
+        error.response.data
+      )
+    }
+
+    if (error.code === 'ECONNREFUSED' || error.code === 'ENOTFOUND') {
+      return new TTSError(
+        'Network error: Unable to connect to MiniMax API',
+        TTSErrorCode.NETWORK,
+        this.name,
+        error.message
+      )
+    }
+
+    return new TTSError(
+      error.message || 'Unknown error',
+      TTSErrorCode.UNKNOWN,
+      this.name,
+      error
+    )
+  }
+
+  /**
+   * 延迟辅助函数
+   */
+  private delay(ms: number): Promise<void> {
+    return new Promise(resolve => setTimeout(resolve, ms))
+  }
+
+  /**
+   * 估算音频时长
+   * 基于 32kbps MP3 估算
+   */
+  private estimateDuration(bytes: number): number {
+    return (bytes * 8) / (32000 * 60)
+  }
+}
+
+/**
+ * MiniMax 预定义音色列表
+ * 官方音色ID参考
+ */
+export const MINIMAX_VOICES: Voice[] = [
+  { id: 'male-qn-qingse', name: '青年清澈', language: 'zh-CN', gender: 'male' },
+  { id: 'male-qn-qingse_2', name: '青年清澈v2', language: 'zh-CN', gender: 'male' },
+  { id: 'female-shaonv', name: '少女', language: 'zh-CN', gender: 'female' },
+  { id: 'male-baiming', name: '成熟男声', language: 'zh-CN', gender: 'male' },
+  { id: 'female-tianmei', name: '甜美女声', language: 'zh-CN', gender: 'female' },
+  { id: 'male-zhongnan', name: '中年男声', language: 'zh-CN', gender: 'male' },
+  { id: 'female-yujie', name: '御姐音', language: 'zh-CN', gender: 'female' },
+  { id: 'male-qn-xiaoao', name: '青年豪爽', language: 'zh-CN', gender: 'male' },
+  { id: 'female-shandian', name: '甜心小娘', language: 'zh-CN', gender: 'female' },
+  { id: 'male-qn-buke', name: '青年低沉', language: 'zh-CN', gender: 'male' },
+  { id: 'male-qn-wenlv', name: '文绿青年', language: 'zh-CN', gender: 'male' },
+  { id: 'female-tianmei-2', name: '甜美女声v2', language: 'zh-CN', gender: 'female' },
+  { id: 'female-yujie-2', name: '御姐音v2', language: 'zh-CN', gender: 'female' },
+  { id: 'male-shaonian', name: '少年音', language: 'zh-CN', gender: 'male' },
+  { id: 'female-yunv', name: '温柔女声', language: 'zh-CN', gender: 'female' },
+  { id: 'male-qn-jingdian', name: '经典男声', language: 'zh-CN', gender: 'male' },
+  { id: 'male-qn-kuang野', name: '狂野青年', language: 'zh-CN', gender: 'male' },
+  { id: 'female-yujie-old', name: '优雅低沉', language: 'zh-CN', gender: 'female' },
+  { id: 'female-tianmei-old', name: '甜美女孩', language: 'zh-CN', gender: 'female' },
+  { id: 'male-qn-taohua', name: '桃花青年', language: 'zh-CN', gender: 'male' }
+]