feat: 照片数字人语音播放

src/renderer/plugins/tts/FetchTTS.ts

+/* eslint-disable camelcase */
 export async function audioAiTTS({
   host,
   text,
@@ -35,14 +36,23 @@ export async function audioAiTTS({
   return res.data
 }
 
-export async function localTTS({ url, text }: { url: string; text: string }) {
+export async function localTTS({
+  url,
+  text,
+  audio_path
+}: {
+  url: string
+  text: string
+  audio_path?: string
+}) {
   const resp = await fetch(url, {
     headers: {
       accept: 'application/json, text/plain, */*',
       'content-type': 'application/json'
     },
     body: JSON.stringify({
-      text
+      text,
+      audio_path
     }),
     method: 'POST',
     mode: 'cors'

src/renderer/screens/ShowPhoto.vue

@@ -247,6 +247,8 @@ function endAudioInput() {
 }
 
 async function onAsr(question: string) {
+  console.log('---------------->question: ', question);
+  endAudioInput();
   const ws = await initSocket()
   inputContext.ws = ws
 
@@ -264,6 +266,10 @@ async function onAsr(question: string) {
       }
 
       if (event === 'stream_end') {
+        answerArray.push(sliceAnswer)
+        runTTSTask(answerArray)
+        sliceAnswer = ''
+
         answerArray.push(sliceAnswer)
         sliceAnswer = ''
         inputContext.ws?.close()
@@ -301,7 +307,6 @@ function initSocket(): Promise<WebSocket> {
 }
 
 let isTTSRunning = false
-const ttsAudios: HTMLAudioElement[] = []
 async function runTTSTask(tasks: string[]) {
   if (isTTSRunning) return
   isTTSRunning = true
@@ -310,11 +315,16 @@ async function runTTSTask(tasks: string[]) {
     while (tasks.length) {
       const task = tasks.shift()
       if (!task) break
+      if (task.length < 1) continue;
+
       console.time(task + ' TTS: ')
-      const res = await localTTS({ url: settings.ttsHost, text: task })
-      console.log('----------------> TTS:', res)
+      const res = await localTTS({ url: settings.ttsHost, text: task, audio_path: settings.userData })
+      console.log('----------------> TTS:', res[0].text)
       console.timeEnd(task + ' TTS: ')
-      console.log('---------------->', ttsAudios)
+      const audio = new Audio(`file://${res[0].text}`);
+      audio.load();
+      ttsAudios.push(audio);
+      runAudioPlay();
     }
   } catch (error) {
     console.error(error)
@@ -323,6 +333,24 @@ async function runTTSTask(tasks: string[]) {
   isTTSRunning = false
 }
 
+const ttsAudios: HTMLAudioElement[] = []
+let isPlayRunning = false
+async function runAudioPlay() {
+  if (isPlayRunning) return
+  isPlayRunning = true
+
+  const audio = ttsAudios.shift();
+  if (!audio) {
+    isPlayRunning = false;
+    return;
+  }
+  audio.onended = () => {
+    isPlayRunning = false;
+    runAudioPlay()
+  };
+  await audio.play();
+}
+
 // eslint-disable-next-line no-unused-vars
 async function xfTTS(text: string) {
   const tone = settings.source.find(({ sourceId }) => settings.selectSource === sourceId)