zyt/TUICallKit-Vue3/tongji/composables/useSpeechToText.js

import { ref, onUnmounted } from 'vue'

const MIN_HOLD_MS = 280

/**
 * 语音转文字（长按说话）：微信小程序 WechatSI；H5 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

    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)) {
            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
    }

    function startWechatRecord(session) {
        // #ifdef MP-WEIXIN
        if (!wxRecordManager || session !== holdSessionId || !sttHolding.value) {
            return
        }
        try {
            startRequested = true
            wxRecordManager.start({
                duration: 60000,
                lang: 'zh_CN'
            })
        } catch (e) {
            startRequested = false
            sttHolding.value = false
            notify('无法开始录音')
        }
        // #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) {
            return
        }
        sttHolding.value = false
        if (sttListening.value || startRequested) {
            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 = () => {
            if (!sttHolding.value) {
                try {
                    wxRecordManager.stop()
                } catch (e) {}
                return
            }
            startRequested = false
            sttListening.value = true
        }
        wxRecordManager.onStop = (res) => {
            sttListening.value = false
            startRequested = false
            emitResult(res?.result)
        }
        wxRecordManager.onError = (res) => {
            sttListening.value = false
            startRequested = false
            sttHolding.value = false
            notify(res?.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
        stopListening()
    })

    return {
        sttListening,
        sttHolding,
        sttSupported,
        beginHoldSpeech,
        endHoldSpeech,
        stopSpeechToText: endHoldSpeech
    }
}