feat: 处理视频数字人说话状态，静止状态衔接问题

src/renderer/screens/ShowVideo.vue

 <!-- eslint-disable no-unused-vars -->
 <!-- eslint-disable camelcase -->
 <script setup lang="ts">
-import { onMounted, ref } from 'vue'
+import { nextTick, onMounted, ref } from 'vue'
 import { useRoute, useRouter } from 'vue-router'
-import type {
-  ServerMessagePartialResult,
-  ServerMessageResult,
-  Model
-} from '@/renderer/plugins/asr/index'
-import { audioAiTTS, localTTS } from '../plugins/tts'
 import useStore from '@/renderer/store'
+import { guid } from '@/renderer/utils/index'
 
 const router = useRouter()
 const route = useRoute()
@@ -22,15 +17,11 @@ const recordVolume = ref(0)
 const url = route.query.url as string
 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 inputContext: {
   mediaStream?: MediaStream
   audioContext?: AudioContext
   audioContext2?: AudioContext
   scriptProcessorNode?: ScriptProcessorNode
-  model?: Model
   ws?: WebSocket
   voskWs?: WebSocket
   asrPartial: string
@@ -43,8 +34,18 @@ const inputContext: {
   ttsAudios: []
 }
 
+const videoElements = {
+  silence: createVideo(role?.url as string),
+  say: createVideo(role?.say as string),
+ }
+const can = ref<HTMLCanvasElement | null>(null)
+let videoElement: HTMLVideoElement | null = null;
+
 onMounted(() => {
-  // init();
+  init().catch((error) => {
+    microphoneState.value = 'waitInput'
+    showError(`init：${error}`)
+  })
 })
 
 router.beforeEach((g) => {
@@ -58,6 +59,41 @@ const showError = (msg: string) => {
   errorMsg.value = msg
 }
 
+function drawFrame(
+  ctx: CanvasRenderingContext2D,
+  video: HTMLVideoElement,
+) {
+  ctx.canvas.width = video.videoWidth
+  ctx.canvas.height = video.videoHeight
+  ctx.clearRect(0, 0, video.videoWidth, video.videoHeight)
+  ctx.drawImage(video, 0, 0, video.videoWidth, video.videoHeight)
+}
+
+
+async function init() {
+  const ctx = can.value?.getContext('2d');
+  if (!ctx) return
+
+  videoElement = videoElements.silence.ele
+  await videoElements.silence.load
+  // ctx.canvas.width = videoElement.videoWidth
+  // ctx.canvas.height = videoElement.videoHeight
+  videoElement.play();
+
+  const fps = 1000 / 30
+  let lastTime = Date.now()
+  const updateFrame = () => {
+    if (Date.now() - lastTime > fps) {
+      videoElement && drawFrame(ctx, videoElement)
+      lastTime = Date.now()
+    }
+    requestAnimationFrame(updateFrame)
+  }
+  requestAnimationFrame(updateFrame)
+
+}
+
+
 function analyzeMicrophoneVolume(stream: MediaStream, callback: (number) => void) {
   const audioContext = new AudioContext()
   const analyser = audioContext.createAnalyser()
@@ -187,7 +223,7 @@ async function llmEnd() {
     })
   ).json()
 
-  console.log('---------------->', resp)
+  console.log('----------------> llmEnd: ', resp)
 }
 
 async function endAudioInput() {
@@ -198,31 +234,55 @@ async function endAudioInput() {
   inputContext.audioContext?.close()
   inputContext.audioContext2?.close()
   inputContext.scriptProcessorNode && (inputContext.scriptProcessorNode.onaudioprocess = null)
-  inputContext.model?.terminate()
   if (inputContext.voskWs) {
     inputContext.voskWs.send('{"eof" : 1}')
     inputContext.voskWs.close()
   }
   inputContext.ttsAudios.length = 0
   inputContext.playingAudio?.pause()
-  videos[1].value?.pause()
-  videos[0].value?.pause()
+  toggleVideo(videoElements.silence.ele)
   isPlayRunning = false
 }
 
-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()!)
-}
+function createVideo(url: string) {
+  const video = document.createElement('video');
 
-function loadVideo(url: string) {
-  videos[1].value!.src = url
-  videos[1].value!.style.opacity = '0'
-  videos[1].value!.addEventListener('canplay', canplay)
+  if (url === role?.url || url === role?.say) {
+    video.loop = true;
+    video.muted = true;
+  } else {
+    video.loop = false;
+    video.muted = false;
+  }
+
+  video.style.display = 'none';
+
+  const load = new Promise<void>((resolve, reject) => {
+    video.oncanplay = () => {
+      video.oncanplay = null;
+      resolve()
+      video.currentTime = 2;
+      document.body.appendChild(video);
+    };
+    video.onerror = reject;
+  })
+  video.src = url;
+
+  return {
+    ele: video,
+    load
+  };
+}
+function toggleVideo(ele: HTMLVideoElement) {
+  videoElement?.pause();
+  videoElement && (videoElement.currentTime = 0.1);
+  if (videoElement && videoElement !== videoElements.silence.ele && videoElement !== videoElements.say.ele) {
+    document.body.removeChild(videoElement);
+  }
+  ele.currentTime = 0.1;
+  ele.pause();
+  videoElement = ele
+  videoElement.play()
 }
 
 async function qNLP(question: string) {
@@ -258,19 +318,21 @@ async function onQ(question: string) {
   try {
     const nlpUrl = await qNLP(question)
     if (nlpUrl) {
-      loadVideo(nlpUrl)
+      const { ele, load } = createVideo(nlpUrl)
       microphoneState.value = 'reply'
-      const videoEle = videos[1].value
-      videoEle!.loop = false
-      videoEle!.muted = false
-      videoEle!.onended = () => {
-        videoEle!.onended = null
+      await load;
+      // 防止切换视频渲染黑屏帧
+      await new Promise(resolve => setTimeout(resolve, 200))
+      toggleVideo(ele)
+      ele.onended = () => {
+        toggleVideo(videoElements.silence.ele)
         microphoneState.value = 'input'
         // TODO: 是否需要初始化
       }
       return
     }
   } catch (error) {
+    console.error(error);
     microphoneState.value = 'input'
     showError(`nlp：${error}`)
     return
@@ -287,15 +349,15 @@ async function onQ(question: string) {
 
 async function llmLoop(question: string) {
   if (!role) return
-  microphoneState.value = 'loading'
 
+  const sessionId = guid()
   const resp = await (
     await fetch(`${settings.llmUrl}/api/v1/generate`, {
       method: 'POST',
       headers: {
         'Content-Type': 'application/json'
       },
-      body: JSON.stringify({ question }),
+      body: JSON.stringify({ generator_id: sessionId, question }),
       mode: 'cors'
     })
   ).json()
@@ -313,6 +375,7 @@ async function llmLoop(question: string) {
       break
     }
 
+    // 轮询间隔时间
     await new Promise((resolve) => setTimeout(resolve, 100))
 
     const { results } = await (
@@ -322,18 +385,18 @@ async function llmLoop(question: string) {
           'Content-Type': 'application/json'
         },
         mode: 'cors',
-        body: JSON.stringify({ question })
+        body: JSON.stringify({ generator_id: sessionId, question })
       })
     ).json()
 
     const audioList = results[0].audio_list as string[]
     if (audioList.length === 0) continue
-    const isEnd = audioList.at(-1) === 'stream_end'
+    inputContext.llmEnd = audioList.at(-1) === 'stream_end';
 
-    if (isEnd) audioList.pop()
+    if (inputContext.llmEnd) audioList.pop()
 
     const newList = audioList.slice(index)
-    if (newList.length === 0 && isEnd) break
+    if (newList.length === 0 && inputContext.llmEnd) return;
     if (newList.length === 0) continue
 
     for (let i = index; i < audioList.length; i++) {
@@ -358,18 +421,14 @@ async function llmLoop(question: string) {
     // TODO: test
     // inputContext.ttsAudios.push(
     //   ...newList.map((path) => {
-    //     const audio = new Audio(`http://192.168.1.57:6767/${path.split('\\').pop()}`)
+    //     const audio = new Audio(`http://10.90.120.45:6767/${path.split('\\').pop()}`)
     //     audio.load()
     //     return audio
     //   })
     // )
 
     runAudioPlay()
-
-    if (isEnd) break
   }
-
-  inputContext.llmEnd = true
 }
 
 let isPlayRunning = false
@@ -380,7 +439,7 @@ async function runAudioPlay() {
   const audio = inputContext.ttsAudios.shift()
   if (!audio) {
     isPlayRunning = false
-    videos[0].value!.pause()
+    toggleVideo(videoElements.silence.ele);
     inputContext.llmEnd && (microphoneState.value = 'input')
     return
   }
@@ -390,9 +449,12 @@ async function runAudioPlay() {
   }
   await audio.play()
   inputContext.playingAudio = audio
-  loadVideo(role!.playUrl)
-  videos[1].value!.loop = true
-  videos[1].value!.muted = true
+  videoElements.say.load.then(async () => {
+    if (videoElements.say.ele.paused) {
+      toggleVideo(videoElements.say.ele);
+      videoElements.say.ele.play()
+    }
+  });
 
   microphoneState.value = 'reply'
 }
@@ -416,8 +478,7 @@ async function down() {
     class="d-flex justify-center align-center"
     :style="{ background: '#000' }"
   >
-    <video id="videoElement" ref="videoElement" :src="url" class="video-ele active"></video>
-    <video id="videoElement2" ref="videoElement2" class="video-ele2"></video>
+    <canvas id="can" ref="can" style="width: 100%; height: 100%; aspect-ratio: 9/16; "></canvas>
   </div>
 
   <div class="voice">
@@ -465,7 +526,7 @@ async function down() {
     </v-chip>
   </div>
 
-  <v-snackbar v-model="errorSnackbar" multi-line :timeout="3000">
+  <v-snackbar v-model="errorSnackbar" multi-line :timeout="6000">
     {{ errorMsg }}
 
     <template #actions>
@@ -503,18 +564,6 @@ async function down() {
   border-radius: 36%;
 }
 
-.video-ele,
-.video-ele2 {
-  position: absolute;
-  width: 100%;
-  height: 100%;
-  opacity: 0;
-}
-.video-ele.active,
-.video-ele2.active {
-  opacity: 1;
-}
-
 .q-list {
   position: fixed;
   bottom: 0;

src/renderer/store/video.ts

@@ -2,7 +2,7 @@ import { defineStore } from 'pinia'
 
 type IVideo = {
   list: {
-    playUrl: string
+    say: string
     url: string
     poster: string
     name: string
@@ -17,9 +17,9 @@ const useVideoStore = defineStore('video', {
       list: [
         {
           url: new URL('/libai/wait.mp4', import.meta.url).href,
+          say: new URL('/libai/10.mp4', import.meta.url).href,
           poster: new URL('/libai/poster.jpg', import.meta.url).href,
           name: '李白',
-          playUrl: new URL('/libai/10.mp4', import.meta.url).href,
           qa: [
             {
               url: new URL('/libai/1.mp4', import.meta.url).href,
@@ -74,9 +74,9 @@ const useVideoStore = defineStore('video', {
           ]
         },
         {
-          url: new URL('/suhe/wait.mov', import.meta.url).href,
+          url: new URL('/suhe/silence.mov', import.meta.url).href,
+          say: new URL('/suhe/say.mov', import.meta.url).href,
           poster: new URL('/suhe/poster.jpg', import.meta.url).href,
-          playUrl: new URL('/suhe/5.mov', import.meta.url).href,
           name: '苏荷',
           qa: [
             {