feat: 更换 icon；修复视频数字人切换视频闪烁问题，回答内容分割不准确问题，优化回答内容麦克风状态联动；新增 llmToTTSSliceLength 配置控制回答内容分割字数

src/renderer/components/layout/HeaderLayout.vue

@@ -84,6 +84,11 @@ if (setting.asr.value === 'vosk_asr') {
 async function changeOpenDevTools() {
   await window.mainApi.send('openDevTools', setting.isOpenDevTools.value)
 }
+
+function clear() {
+  localStorage.clear()
+  location.reload()
+}
 </script>
 <template>
   <v-app-bar color="#d71b1b" density="compact" class="header">
@@ -181,6 +186,27 @@ async function changeOpenDevTools() {
                   :model-value="setting.llmUrl"
                 ></v-text-field>
 
+                <v-slider
+                  v-model="setting.llmToTTSSliceLength.value"
+                  label="TTS 分句长度"
+                  class="align-center"
+                  :max="100"
+                  :min="0"
+                  hide-details
+                  :step="1"
+                >
+                  <template #append>
+                    <v-text-field
+                      v-model="setting.llmToTTSSliceLength.value"
+                      hide-details
+                      single-line
+                      density="compact"
+                      type="number"
+                      style="width: 80px"
+                    ></v-text-field>
+                  </template>
+                </v-slider>
+
                 <v-switch
                   v-model="setting.isFullscreen.value"
                   hide-details
@@ -202,6 +228,7 @@ async function changeOpenDevTools() {
 
             <v-card-actions>
               <v-spacer></v-spacer>
+              <v-btn color="#d71b1b" text="清除缓存并刷新" @click="clear"></v-btn>
               <v-btn text="关闭" @click="isActive.value = false"></v-btn>
             </v-card-actions>
           </v-card>

src/renderer/screens/ShowPhoto.vue

@@ -252,7 +252,7 @@ async function startVoskWsAudioInput() {
     return
   }
 
-  await initVoskWS();
+  await initVoskWS()
   sampleRate = 8000
   const mediaStream = await navigator.mediaDevices.getUserMedia({
     audio: {
@@ -369,14 +369,18 @@ async function onAsr(question: string) {
       answer += text
       isTime && console.time('sliceAnswer')
       isTime = false
-      sliceAnswer += text
 
-      if (/[。，？！；,.?!;]/.test(text) && sliceAnswer.length >= 20) {
-        console.timeEnd('sliceAnswer')
-        answerArray.push(sliceAnswer)
-        runTTSTask(answerArray)
-        sliceAnswer = ''
-        isTime = true
+      const textArr = text.split('');
+      for (let i = 0; i < textArr.length; i++) {
+        const t = textArr[i];
+        sliceAnswer += t
+        if (/[。，？！；,.?!;]/.test(t) && sliceAnswer.length >= settings.llmToTTSSliceLength) {
+          console.timeEnd('sliceAnswer')
+          answerArray.push(sliceAnswer)
+          runTTSTask(answerArray)
+          sliceAnswer = ''
+          isTime = true
+        }
       }
     } catch (error) {
       console.log('返回答案错误 -----> ' + JSON.stringify(error))

src/renderer/screens/ShowVideo.vue

@@ -24,8 +24,7 @@ const role = useVideo.list.find((i) => i.url === url)
 const microphoneState = ref<'waitInput' | 'input' | 'loading' | 'disabled' | 'reply'>('waitInput')
 const videoElement = ref<HTMLVideoElement | null>(null)
 const videoElement2 = ref<HTMLVideoElement | null>(null)
-const videos = [videoElement, videoElement2];
-
+const videos = [videoElement, videoElement2]
 
 onMounted(() => {
   // init();
@@ -166,7 +165,7 @@ async function startVoskWsAudioInput() {
   }
 
   initVoskWS()
-  sampleRate = 8000
+  sampleRate = 16000
   const mediaStream = await navigator.mediaDevices.getUserMedia({
     audio: {
       echoCancellation: true,
@@ -184,11 +183,15 @@ async function startVoskWsAudioInput() {
   processor.connect(audioContext.destination)
 
   processor.onaudioprocess = (audioDataChunk) => {
-    if (microphoneState.value === 'loading' || microphoneState.value === 'disabled' || microphoneState.value === 'reply') {
-      return;
+    if (
+      microphoneState.value === 'loading' ||
+      microphoneState.value === 'disabled' ||
+      microphoneState.value === 'reply'
+    ) {
+      return
     }
 
-    postAudio(audioDataChunk);
+    postAudio(audioDataChunk)
   }
 
   await analyzeMicrophoneVolume(mediaStream, (val) => {
@@ -255,24 +258,24 @@ function endAudioInput() {
 }
 
 const canplay = () => {
-  videos[1].value!.style.opacity = '1';
-  videos[0].value!.style.opacity = '0';
-  videos[0].value!.pause();
-  videos[1].value!.play();
-  videos[1].value!.removeEventListener('canplay', canplay);
-  videos.unshift(videos.pop()!);
+  videos[1].value!.style.opacity = '1'
+  videos[0].value!.style.opacity = '0'
+  videos[0].value!.pause()
+  videos[1].value!.play()
+  videos[1].value!.removeEventListener('canplay', canplay)
+  videos.unshift(videos.pop()!)
 }
 
 function loadVideo(url: string) {
   videos[1].value!.src = url
-  videos[1].value!.style.opacity = '0';
-  videos[1].value!.addEventListener('canplay', canplay);
+  videos[1].value!.style.opacity = '0'
+  videos[1].value!.addEventListener('canplay', canplay)
 }
 
 async function onAsr(question: string) {
   console.log('---------------->', question)
-  if (!role) return;
-  microphoneState.value = 'loading';
+  if (!role) return
+  microphoneState.value = 'loading'
 
   question = question.replace(/\s/g, '')
   for (let i = 0; i < role.qa.length; i++) {
@@ -280,13 +283,13 @@ async function onAsr(question: string) {
     console.log(question + ' : ' + q)
     if (q.includes(question)) {
       loadVideo(url)
-      microphoneState.value = 'reply';
+      microphoneState.value = 'reply'
       const videoEle = videos[1].value
       videoEle!.loop = false
       videoEle!.muted = false
       videoEle!.onended = () => {
-        videoEle!.onended = null;
-        microphoneState.value = 'input';
+        videoEle!.onended = null
+        microphoneState.value = 'input'
         // 是否需要初始化
       }
       return
@@ -324,18 +327,23 @@ async function onAsr(question: string) {
       answer += text
       isTime && console.time('sliceAnswer')
       isTime = false
-      sliceAnswer += text
 
-      if (/[。，？！；,.?!;]/.test(text) && sliceAnswer.length >= 10) {
-        console.timeEnd('sliceAnswer')
-        answerArray.push(sliceAnswer)
-        runTTSTask(answerArray)
-        sliceAnswer = ''
-        isTime = true
+      const textArr = text.split('');
+      for (let i = 0; i < textArr.length; i++) {
+        const t = textArr[i];
+        sliceAnswer += t
+        if (/[。，？！；,.?!;]/.test(t) && sliceAnswer.length >= settings.llmToTTSSliceLength) {
+          console.timeEnd('sliceAnswer')
+          answerArray.push(sliceAnswer)
+          runTTSTask(answerArray)
+          sliceAnswer = ''
+          isTime = true
+        }
       }
+
     } catch (error) {
       console.log('返回答案错误 -----> ' + JSON.stringify(error))
-      microphoneState.value = 'input';
+      microphoneState.value = 'input'
     }
   }
 
@@ -360,11 +368,11 @@ async function runTTSTask(tasks: string[]) {
     while (tasks.length) {
       const task = tasks.shift()
       if (!task) break
-      if (task.length < 1) continue
+      if (task.trim().length < 1) continue
 
       console.time(task + ' TTS: ')
 
-      microphoneState.value = 'loading';
+      microphoneState.value = 'loading'
       const res = await localTTS({
         url: settings.ttsHost,
         text: task,
@@ -395,21 +403,21 @@ async function runAudioPlay() {
 
   const audio = ttsAudios.shift()
   if (!audio) {
-    isPlayRunning = false;
-    videos[0].value!.pause();
-    !isTTSRunning && (microphoneState.value = 'input');
+    isPlayRunning = false
+    videos[0].value!.pause()
+    !isTTSRunning && (microphoneState.value = 'input')
     return
   }
   audio.onended = () => {
     isPlayRunning = false
     runAudioPlay()
   }
-  await audio.play();
+  await audio.play()
   loadVideo(new URL('/libai/10.mp4', import.meta.url).href)
   videos[1].value!.loop = true
   videos[1].value!.muted = true
 
-  microphoneState.value = 'reply';
+  microphoneState.value = 'reply'
 }
 
 // eslint-disable-next-line no-unused-vars
@@ -425,7 +433,6 @@ async function xfTTS(text: string) {
   })
   console.log('----------------> tts:', res)
 }
-
 </script>
 
 <template>
@@ -444,7 +451,11 @@ async function xfTTS(text: string) {
       color="#fff"
       variant="elevated"
       size="x-large"
-      :disabled="microphoneState === 'loading' || microphoneState === 'disabled' || microphoneState === 'reply'"
+      :disabled="
+        microphoneState === 'loading' ||
+        microphoneState === 'disabled' ||
+        microphoneState === 'reply'
+      "
       @pointerdown="startVoskWsAudioInput"
     >
       <v-icon v-if="microphoneState === 'waitInput'" icon="mdi-microphone"></v-icon>
@@ -512,13 +523,15 @@ async function xfTTS(text: string) {
   border-radius: 36%;
 }
 
-.video-ele, .video-ele2 {
+.video-ele,
+.video-ele2 {
   position: absolute;
   width: 100%;
   height: 100%;
   opacity: 0;
 }
-.video-ele.active, .video-ele2.active {
+.video-ele.active,
+.video-ele2.active {
   opacity: 1;
 }
 

src/renderer/store/settings.ts

@@ -25,6 +25,7 @@ export type ISettings = {
   isFullscreen: 'yes' | 'no'
   isOpenDevTools: boolean
   llmUrl: string
+  llmToTTSSliceLength: number
   voskWsLUrl: string
 }
 
@@ -58,7 +59,8 @@ const useSettingsStore = defineStore('settings', {
       selectSource: '',
       isFullscreen: 'no',
       isOpenDevTools: false,
-      llmUrl: 'ws://127.0.0.1:9001/api/v1/stream',
+      llmUrl: 'ws://127.0.0.1:9899/api/v1/stream',
+      llmToTTSSliceLength: 20,
       voskWsLUrl: 'ws://127.0.0.1:2700'
     }) as ISettings,
   getters: {},