zyt/TUICallKit-Vue3/scripts/generate-voice.mjs

#!/usr/bin/env node
/* eslint-disable no-console */
/**
 * 用小米 MiMo TTS 生成训练语音素材
 *
 * 文档: https://platform.xiaomimimo.com/docs/en-US/usage-guide/speech-synthesis-v2.5
 * 接口: POST https://api.xiaomimimo.com/v1/chat/completions
 *
 * 准备:
 *   export MIMO_API_KEY=sk-your-key-here
 *   (或 export XIAOMI_API_KEY=...)
 *
 * 运行:
 *   cd uniapp
 *   node scripts/generate-voice.mjs              # 默认音色 冰糖
 *   node scripts/generate-voice.mjs --voice 茉莉
 *   node scripts/generate-voice.mjs --force       # 已存在的也重新生成
 *
 * 输出(分包):
 *   training/static/voice/numbers/{1..30}.mp3
 *   training/static/voice/prompts/{key}.mp3
 *
 * 需要 Node 18+（用内置 fetch）
 */

import { mkdir, writeFile, access, stat } from 'node:fs/promises'
import { dirname, join } from 'node:path'
import { fileURLToPath } from 'node:url'
import { Buffer } from 'node:buffer'

const __filename = fileURLToPath(import.meta.url)
const __dirname = dirname(__filename)

const ROOT = join(__dirname, '..')
const VOICE_DIR = join(ROOT, 'training/static/voice')
const NUMBERS_DIR = join(VOICE_DIR, 'numbers')
const PROMPTS_DIR = join(VOICE_DIR, 'prompts')

// ============ 参数 ============
const args = process.argv.slice(2)
const FORCE = args.includes('--force')
const VOICE = pickArg('--voice') || '冰糖'
const MODEL = pickArg('--model') || 'mimo-v2.5-tts'
const FORMAT = pickArg('--format') || 'mp3'

const API_KEY =
    process.env.MIMO_API_KEY ||
    process.env.XIAOMI_API_KEY ||
    process.env.XIAOMI_MIMO_API_KEY
const ENDPOINT =
    process.env.MIMO_ENDPOINT || 'https://api.xiaomimimo.com/v1/chat/completions'

if (!API_KEY) {
    console.error('❌ 未找到 API Key')
    console.error('   请先 export MIMO_API_KEY=your-key')
    process.exit(1)
}

function pickArg(name) {
    const idx = args.indexOf(name)
    if (idx === -1) return null
    return args[idx + 1]
}

// ============ 语音内容 ============

// 数字报数：简短的"教练点数"风格
const NUMBER_STYLE =
    '用简短有力的健身教练口吻报数，干净利落，节奏明快，每个数字独立清晰。'

// 口令：温柔有鼓励性的女教练
const PROMPT_STYLE =
    '用温柔但有力量的女教练口吻说话，语调亲切自然，节奏适中，像在带学员训练。'

const PROMPTS = {
    start: '开始',
    ready: '准备',
    rest: '休息一下',
    'next-set': '下一组开始',
    'last-rep': '最后一次',
    'keep-it-up': '继续坚持',
    'good-job': '很棒，训练完成',
    'breathe-in': '吸气',
    'breathe-out': '呼气',
}

// ============ 工具 ============

async function fileExists(p) {
    try {
        await access(p)
        return true
    } catch {
        return false
    }
}

async function ensureDir(d) {
    await mkdir(d, { recursive: true })
}

async function tts(text, outPath, styleInstruction) {
    if (!FORCE && (await fileExists(outPath))) {
        const s = await stat(outPath)
        if (s.size > 0) {
            console.log(`[skip] ${outPath}  (已存在)`)
            return
        }
    }

    const body = {
        model: MODEL,
        messages: [
            { role: 'user', content: styleInstruction },
            { role: 'assistant', content: text },
        ],
        audio: { format: FORMAT, voice: VOICE },
    }

    const res = await fetch(ENDPOINT, {
        method: 'POST',
        headers: {
            'api-key': API_KEY,
            'Content-Type': 'application/json',
        },
        body: JSON.stringify(body),
    })

    if (!res.ok) {
        const errText = await res.text().catch(() => '')
        throw new Error(`HTTP ${res.status} ${res.statusText}: ${errText.slice(0, 300)}`)
    }

    const json = await res.json()
    const audioBase64 = json?.choices?.[0]?.message?.audio?.data
    if (!audioBase64) {
        throw new Error(
            `响应里找不到 audio.data 字段。完整响应: ${JSON.stringify(json).slice(0, 500)}`,
        )
    }

    const buf = Buffer.from(audioBase64, 'base64')
    await writeFile(outPath, buf)
    console.log(`[ok]   ${outPath}  (${buf.length} bytes)  → "${text}"`)
}

async function sleep(ms) {
    return new Promise((r) => setTimeout(r, ms))
}

// ============ 主流程 ============

async function main() {
    console.log('====== 小米 MiMo TTS 语音生成 ======')
    console.log(`Endpoint : ${ENDPOINT}`)
    console.log(`Model    : ${MODEL}`)
    console.log(`Voice    : ${VOICE}`)
    console.log(`Format   : ${FORMAT}`)
    console.log(`Force    : ${FORCE}`)
    console.log(`Output   : ${VOICE_DIR}`)
    console.log('====================================\n')

    await ensureDir(NUMBERS_DIR)
    await ensureDir(PROMPTS_DIR)

    let successCount = 0
    let failCount = 0

    console.log('--- 数字 1-30 ---')
    for (let n = 1; n <= 30; n++) {
        try {
            await tts(String(n), join(NUMBERS_DIR, `${n}.${FORMAT}`), NUMBER_STYLE)
            successCount++
            await sleep(120)
        } catch (e) {
            failCount++
            console.error(`[fail] 数字 ${n}: ${e.message}`)
        }
    }

    console.log('\n--- 口令 ---')
    for (const [key, text] of Object.entries(PROMPTS)) {
        try {
            await tts(text, join(PROMPTS_DIR, `${key}.${FORMAT}`), PROMPT_STYLE)
            successCount++
            await sleep(120)
        } catch (e) {
            failCount++
            console.error(`[fail] ${key}: ${e.message}`)
        }
    }

    console.log('\n====================================')
    console.log(`成功 ${successCount} | 失败 ${failCount}`)
    console.log('====================================')

    if (FORMAT !== 'mp3') {
        console.log(
            `\n⚠️  当前生成格式是 ${FORMAT}，但 hooks/useVoiceCoach.ts 默认引用 .mp3。`,
        )
        console.log('   要么换 --format mp3 重跑，要么修改 hooks 里的 src 后缀。')
    }

    console.log('\n下一步：')
    console.log(' 1. 准备 click 音 → training/static/audio/click.mp3')
    console.log(' 2. 准备 BGM     → training/static/bgm/{train-light,rest-meditation}.mp3')
    console.log(' 3. 编译运行     → 进入 /training/pages/index')
}

main().catch((e) => {
    console.error('\n❌ 致命错误:', e)
    process.exit(1)
})