sionrui/frontend/app/web-gold/src/views/kling/hooks/useVoiceGeneration.ts

/**
 * @fileoverview useVoiceGeneration Hook - 语音生成逻辑封装
 * @author Claude Code
 */

import { ref, computed } from 'vue'
import { message } from 'ant-design-vue'
import type {
  UseVoiceGeneration,
  AudioState,
  VoiceMeta,
  AudioData,
} from '../types/identify-face'
// @ts-ignore
import { VoiceService } from '@/api/voice'
import { DEFAULT_VOICE_PROVIDER } from '@/config/voiceConfig'

/**
 * 语音生成 Hook
 * 独立管理所有状态，不依赖外部状态
 */
export function useVoiceGeneration(): UseVoiceGeneration {
  // ==================== 响应式状态 ====================

  const ttsText = ref<string>('')
  const speechRate = ref<number>(1.0)
  const selectedVoiceMeta = ref<VoiceMeta | null>(null)
  const audioState = ref<AudioState>({
    generated: null,
    durationMs: 0,
    validationPassed: false,
    generating: false,
  })

  // ==================== 计算属性 ====================

  /**
   * 是否可以生成配音
   */
  const canGenerateAudio = computed(() => {
    const hasText = ttsText.value.trim()
    const hasVoice = selectedVoiceMeta.value
    const hasVideo = true // 语音生成不依赖视频状态
    return !!(hasText && hasVoice && hasVideo && !audioState.value.generating)
  })

  /**
   * 建议的最大字符数（需要从外部传入）
   */
  const suggestedMaxChars = computed(() => {
    // 默认为 4000，需要从外部设置
    return 4000
  })

  // ==================== 核心方法 ====================

  /**
   * 生成配音
   */
  const generateAudio = async (): Promise<void> => {
    const voice = selectedVoiceMeta.value
    if (!voice) {
      message.warning('请选择音色')
      return
    }

    if (!ttsText.value.trim()) {
      message.warning('请输入文案内容')
      return
    }

    audioState.value.generating = true

    try {
      const params = {
        inputText: ttsText.value,
        voiceConfigId: voice.rawId || extractIdFromString(voice.id),
        speechRate: speechRate.value || 1.0,
        audioFormat: 'mp3' as const,
        providerType: DEFAULT_VOICE_PROVIDER,
      }

      const res = await VoiceService.synthesize(params)

      if (res.code === 0) {
        const audioData = res.data as AudioData

        if (!audioData.audioBase64) {
          throw new Error('未收到音频数据，无法进行时长解析')
        }

        audioState.value.generated = audioData

        try {
          // 解析音频时长
          audioState.value.durationMs = await parseAudioDuration(audioData.audioBase64)

          // 验证音频时长
          validateAudioDuration()

          message.success('配音生成成功！')
        } catch (error) {
          message.error('音频解析失败，请重新生成配音')
          audioState.value.durationMs = 0
          audioState.value.generated = null
          audioState.value.validationPassed = false
        }
      } else {
        throw new Error(res.msg || '配音生成失败')
      }
    } catch (error: any) {
      message.error(error.message || '配音生成失败')
    } finally {
      audioState.value.generating = false
    }
  }

  /**
   * 解析音频时长
   */
  const parseAudioDuration = async (base64Data: string): Promise<number> => {
    return new Promise((resolve, reject) => {
      try {
        const base64 = base64Data.includes(',') ? base64Data.split(',')[1] : base64Data

        const binaryString = window.atob(base64)
        const bytes = new Uint8Array(binaryString.length)
        for (let i = 0; i < binaryString.length; i++) {
          bytes[i] = binaryString.charCodeAt(i)
        }

        const blob = new Blob([bytes], { type: 'audio/mp3' })
        const audio = new Audio()
        const objectUrl = URL.createObjectURL(blob)

        audio.addEventListener('loadedmetadata', () => {
          URL.revokeObjectURL(objectUrl)
          const durationMs = Math.round(audio.duration * 1000)
          resolve(durationMs)
        })

        audio.addEventListener('error', (error) => {
          URL.revokeObjectURL(objectUrl)
          reject(error)
        })

        audio.src = objectUrl
        audio.load()
      } catch (error) {
        reject(error)
      }
    })
  }

  /**
   * 验证音频与人脸区间的重合时长（外部调用时传入校验参数）
   */
  const validateAudioDuration = (
    faceStartTime: number = 0,
    faceEndTime: number = 0,
    minOverlapMs: number = 2000
  ): boolean => {
    if (faceStartTime <= 0 || faceEndTime <= 0) {
      audioState.value.validationPassed = false
      return false
    }

    const faceDurationMs = faceEndTime - faceStartTime
    const audioDuration = audioState.value.durationMs

    const overlapStart = faceStartTime
    const overlapEnd = Math.min(faceEndTime, faceStartTime + audioDuration)
    const overlapDuration = Math.max(0, overlapEnd - overlapStart)

    const isValid = overlapDuration >= minOverlapMs

    audioState.value.validationPassed = isValid

    if (!isValid) {
      const overlapSec = (overlapDuration / 1000).toFixed(1)
      message.warning(
        `音频时长（${(audioDuration/1000).toFixed(1)}秒）与人脸区间（${(faceDurationMs/1000).toFixed(1)}秒）不匹配，重合部分仅${overlapSec}秒，至少需要${(minOverlapMs/1000)}秒`
      )
    } else {
      message.success('时长校验通过！')
    }

    return isValid
  }

  /**
   * 重置音频状态
   */
  const resetAudioState = (): void => {
    audioState.value.generated = null
    audioState.value.durationMs = 0
    audioState.value.validationPassed = false
    audioState.value.generating = false
  }

  return {
    // 响应式状态
    ttsText,
    speechRate,
    selectedVoiceMeta,
    audioState,

    // 计算属性
    canGenerateAudio,
    suggestedMaxChars,

    // 方法
    generateAudio,
    parseAudioDuration,
    validateAudioDuration,
    resetAudioState,
  }
}

/**
 * 从字符串中提取ID
 */
function extractIdFromString(str: string): string {
  // 尝试从各种格式中提取ID
  const match = str.match(/[\w-]+$/)
  return match ? match[0] : str
}