zhiyin/backend/src/modules/tts/tts.service.ts

import * as crypto from 'crypto'
import * as fs from 'fs'
import * as path from 'path'
import { execSync } from 'child_process'
import { Injectable, Logger } from '@nestjs/common'

const CACHE_DIR = '/tmp/tts-cache'

interface TtsResult {
  hash: string
  filePath: string
  durationMs: number
  amplitudeData: number[]
}

const VALID_VOICES = new Set([
  'zh-CN-XiaoxiaoNeural', 'zh-CN-XiaoyiNeural', 'zh-CN-YunjianNeural',
  'zh-CN-YunxiNeural', 'zh-CN-YunxiaNeural', 'zh-CN-YunyangNeural',
  'zh-CN-liaoning-XiaobeiNeural', 'zh-CN-shaanxi-XiaoniNeural',
])

function validateVoice(voice: string): void {
  if (!VALID_VOICES.has(voice)) {
    throw new Error(`不支持的语音: ${voice}`)
  }
}

@Injectable()
export class TtsService {
  private readonly logger = new Logger(TtsService.name)

  constructor() {
    if (!fs.existsSync(CACHE_DIR)) {
      fs.mkdirSync(CACHE_DIR, { recursive: true })
    }
  }

  async synthesize(text: string, voice: string = 'zh-CN-XiaoxiaoNeural'): Promise<TtsResult> {
    validateVoice(voice)
    const hash = crypto.createHash('md5').update(text + voice).digest('hex')
    const filePath = path.join(CACHE_DIR, `${hash}.mp3`)

    if (fs.existsSync(filePath)) {
      const durationMs = await this.getDuration(filePath)
      const amplitudeData = this.loadAmplitudeData(hash)
      if (amplitudeData) {
        return { hash, filePath, durationMs, amplitudeData }
      }
    }

    try {
      execSync(
        `edge-tts --voice "${voice}" --text "${this.escapeText(text)}" --write-media "${filePath}"`,
        { timeout: 30000 },
      )
      const durationMs = await this.getDuration(filePath)
      const amplitudeData = this.extractAmplitude(filePath, hash)
      this.logger.log(`TTS generated: hash=${hash} text="${text.slice(0, 40)}..." duration=${durationMs}ms`)
      return { hash, filePath, durationMs, amplitudeData }
    } catch (e) {
      this.logger.error(`TTS failed: ${e.message}`)
      throw e
    }
  }

  getCachedPath(hash: string): string | null {
    const filePath = path.join(CACHE_DIR, `${hash}.mp3`)
    return fs.existsSync(filePath) ? filePath : null
  }

  private extractAmplitude(mp3Path: string, hash: string): number[] {
    try {
      const pcmPath = `/tmp/tts-cache/${hash}.pcm`
      execSync(
        `ffmpeg -y -i "${mp3Path}" -f s16le -acodec pcm_s16le -ar 16000 -ac 1 "${pcmPath}" 2>/dev/null`,
        { timeout: 10000 },
      )
      const pcmBuf = fs.readFileSync(pcmPath)
      const samples = new Int16Array(pcmBuf.buffer, pcmBuf.byteOffset, pcmBuf.byteLength / 2)
      const chunkSize = Math.floor(16000 * 0.05) // 50ms
      const amplitudes: number[] = []
      for (let i = 0; i < samples.length; i += chunkSize) {
        const end = Math.min(i + chunkSize, samples.length)
        let sumSq = 0
        for (let j = i; j < end; j++) {
          sumSq += samples[j] * samples[j]
        }
        const rms = Math.sqrt(sumSq / (end - i))
        amplitudes.push(Number((Math.min(1, rms / 16000)).toFixed(4)))
      }
      try { fs.unlinkSync(pcmPath) } catch {}
      const ampPath = `/tmp/tts-cache/${hash}.amp`
      fs.writeFileSync(ampPath, JSON.stringify(amplitudes))
      return amplitudes
    } catch (e) {
      this.logger.warn(`振幅提取失败: ${e.message}`)
      return []
    }
  }

  private loadAmplitudeData(hash: string): number[] | null {
    try {
      const ampPath = `/tmp/tts-cache/${hash}.amp`
      if (!fs.existsSync(ampPath)) return null
      return JSON.parse(fs.readFileSync(ampPath, 'utf8'))
    } catch {
      return null
    }
  }

  private escapeText(text: string): string {
    return text.replace(/"/g, '\\"').replace(/\n/g, ' ').replace(/\r/g, '')
  }

  private async getDuration(filePath: string): Promise<number> {
    try {
      // Estimate duration from file size (~16kbps for mp3 at 22050Hz)
      const stat = fs.statSync(filePath)
      const bytesPerMs = 16 * 1024 / 8 / 1000  // 16kbps → bytes per ms
      return Math.round(stat.size / bytesPerMs)
    } catch {
      return 3000
    }
  }
}