Files
zyt/TUICallKit-Vue3/tongji/composables/useSpeechToText.js
T
2026-06-09 11:42:35 +08:00

309 lines
9.3 KiB
JavaScript
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
import { ref, onUnmounted } from 'vue'
const MIN_HOLD_MS = 280
/**
* 语音转文字(长按说话):微信小程序 WechatSIH5 Web Speech API。
*
* @param {{ onResult?: (text: string) => void, showToast?: (msg: string) => void }} options
*/
export function useSpeechToText({ onResult, showToast } = {}) {
const sttListening = ref(false)
const sttHolding = ref(false)
const sttSupported = ref(false)
let wxRecordManager = null
let h5Recognition = null
let holdStartTs = 0
let holdSessionId = 0
let startRequested = false
let recordAuthorized = null
// WechatSI 录音管理器是全局单例:跨页面/快速点按可能残留会话,
// 用以下两个标记把 start/stop 串行化,避免 "please stop after start"
let managerRecording = false
let pendingStartSession = 0
function notify(msg) {
if (!msg) return
if (typeof showToast === 'function') {
showToast(msg)
return
}
uni.showToast({ title: msg, icon: 'none', duration: 2000 })
}
function emitResult(text) {
const t = String(text || '').trim()
const heldMs = Date.now() - holdStartTs
if (!t) {
if (heldMs < MIN_HOLD_MS) return
notify('没听清,请再试一次')
return
}
if (typeof onResult === 'function') {
onResult(t)
}
}
function stopListening() {
// #ifdef MP-WEIXIN
if (wxRecordManager && (sttListening.value || startRequested || managerRecording)) {
try {
wxRecordManager.stop()
} catch (e) {}
}
// #endif
// #ifdef H5
if (h5Recognition && sttListening.value) {
try {
h5Recognition.stop()
} catch (e) {}
}
// #endif
startRequested = false
sttListening.value = false
}
function ensureRecordAuth(onGranted) {
// #ifdef MP-WEIXIN
if (recordAuthorized === true) {
onGranted()
return
}
uni.authorize({
scope: 'scope.record',
success() {
recordAuthorized = true
onGranted()
},
fail() {
recordAuthorized = false
sttHolding.value = false
uni.showModal({
title: '需要麦克风权限',
content: '请在设置中允许使用麦克风,以便长按说话',
confirmText: '去设置',
success(res) {
if (res.confirm) {
uni.openSetting({})
}
}
})
}
})
// #endif
// #ifndef MP-WEIXIN
onGranted()
// #endif
}
// #ifdef MP-WEIXIN
function actuallyStartWechat(session) {
if (!wxRecordManager || session !== holdSessionId || !sttHolding.value) {
return
}
try {
startRequested = true
managerRecording = true
wxRecordManager.start({
duration: 60000,
lang: 'zh_CN'
})
} catch (e) {
startRequested = false
managerRecording = false
sttHolding.value = false
notify('无法开始录音')
}
}
// #endif
function startWechatRecord(session) {
// #ifdef MP-WEIXIN
if (!wxRecordManager || session !== holdSessionId || !sttHolding.value) {
return
}
// 上一段会话还没结束(含全局单例残留):先停止并排队,
// 待 onStop 回调后再真正开始,避免 "please stop after start"
if (managerRecording || startRequested) {
pendingStartSession = session
try {
wxRecordManager.stop()
} catch (e) {}
return
}
actuallyStartWechat(session)
// #endif
}
function startH5Record() {
// #ifdef H5
if (!h5Recognition) {
notify('当前浏览器不支持语音输入')
sttHolding.value = false
return
}
try {
startRequested = true
h5Recognition.start()
sttListening.value = true
} catch (e) {
startRequested = false
sttHolding.value = false
notify('无法开始语音识别,请检查麦克风权限')
}
// #endif
}
/** 长按开始:手指按下 */
function beginHoldSpeech() {
if (!sttSupported.value) {
notify('当前环境不支持语音输入')
return
}
if (sttHolding.value) return
const session = ++holdSessionId
holdStartTs = Date.now()
sttHolding.value = true
// #ifdef MP-WEIXIN
ensureRecordAuth(() => startWechatRecord(session))
// #endif
// #ifdef H5
startH5Record()
// #endif
// #ifndef MP-WEIXIN
// #ifndef H5
sttHolding.value = false
notify('当前端暂不支持语音输入')
// #endif
// #endif
}
/** 松手结束:手指抬起 */
function endHoldSpeech() {
if (!sttHolding.value && !sttListening.value && !startRequested && !managerRecording) {
return
}
sttHolding.value = false
// 松手后不再补开排队的录音
pendingStartSession = 0
if (sttListening.value || startRequested || managerRecording) {
stopListening()
return
}
holdSessionId += 1
}
// #ifdef MP-WEIXIN
try {
// eslint-disable-next-line no-undef
const plugin = requirePlugin('WechatSI')
wxRecordManager = plugin.getRecordRecognitionManager()
sttSupported.value = !!wxRecordManager
wxRecordManager.onStart = () => {
managerRecording = true
if (!sttHolding.value) {
try {
wxRecordManager.stop()
} catch (e) {}
return
}
startRequested = false
sttListening.value = true
}
wxRecordManager.onStop = (res) => {
managerRecording = false
sttListening.value = false
startRequested = false
// 有排队的开始请求(残留会话已停止 / 用户仍按住):现在再真正开始
if (pendingStartSession && pendingStartSession === holdSessionId && sttHolding.value) {
const s = pendingStartSession
pendingStartSession = 0
actuallyStartWechat(s)
return
}
pendingStartSession = 0
emitResult(res?.result)
}
wxRecordManager.onError = (res) => {
managerRecording = false
sttListening.value = false
startRequested = false
const msg = (res && res.msg) || ''
// 上一段未停止就再次 start 触发:停止后若仍按住则自动重试,不打扰用户
if (/please stop after start/i.test(msg)) {
try {
wxRecordManager.stop()
} catch (e) {}
pendingStartSession = sttHolding.value ? holdSessionId : 0
return
}
pendingStartSession = 0
sttHolding.value = false
notify(msg || '语音识别失败')
}
} catch (e) {
console.warn('WechatSI record recognition not available', e)
}
// #endif
// #ifdef H5
try {
if (typeof window !== 'undefined') {
const SR = window.SpeechRecognition || window.webkitSpeechRecognition
if (SR) {
h5Recognition = new SR()
h5Recognition.lang = 'zh-CN'
h5Recognition.interimResults = false
h5Recognition.continuous = false
h5Recognition.maxAlternatives = 1
sttSupported.value = true
h5Recognition.onresult = (event) => {
const item = event.results?.[0]?.[0]
emitResult(item?.transcript || '')
}
h5Recognition.onend = () => {
sttListening.value = false
startRequested = false
sttHolding.value = false
}
h5Recognition.onerror = (event) => {
sttListening.value = false
startRequested = false
sttHolding.value = false
if (event?.error === 'not-allowed') {
notify('请允许浏览器使用麦克风')
} else if (event?.error !== 'aborted') {
notify('语音识别失败')
}
}
}
}
} catch (e) {
console.warn('H5 SpeechRecognition not available', e)
}
// #endif
onUnmounted(() => {
sttHolding.value = false
holdSessionId += 1
pendingStartSession = 0
stopListening()
})
return {
sttListening,
sttHolding,
sttSupported,
beginHoldSpeech,
endHoldSpeech,
stopSpeechToText: endHoldSpeech
}
}