新增：加入高清语音，回复“#chatgpt开启高清语音”来开启

1.新增：加入高清语音，回复“#chatgpt开启高清语音”来开启
2.优化：语音模式下遇到超长文本将会使用图片模式

apps/chat.js

@@ -1113,7 +1113,7 @@ export class chatgpt extends plugin {
       for (let quote of quotemessage) {
         if (quote.imageLink) imgUrls.push(quote.imageLink)
       }
-      if (useTTS) {
+      if (useTTS && response.length <= Config.autoUsePictureThreshold {
         // 缓存数据
         this.cacheContent(e, use, response, prompt, quotemessage, mood, chatMessage.suggestedResponses, imgUrls)
         if (response === 'Thanks for this conversation! I\'ve reached my limit, will you hit “New topic,” please?') {

apps/management.js

@@ -184,6 +184,11 @@ export class ChatgptManagement extends plugin {
           fnc: 'setDefaultReplySetting',
           permission: 'master'
         },
+        {
+            reg: '^#chatgpt(开启|关闭)高清语音',
+            fnc: 'enableTtsHD',
+            permission: 'master'
+        },
         {
           /** 命令正则匹配 */
           reg: '^#(关闭|打开)群聊上下文$',
@@ -483,7 +488,14 @@ export class ChatgptManagement extends plugin {
     await this.reply('设置成功', e.isGroup)
     return false
   }
-
+  async enableTtsHD (e) {
+    Config.ttsHD = e.msg.indexOf('开启') > -1
+    if(Config.ttsHD) {
+        await this.reply('已开启高清语音，电脑端将无法播放语音', true)
+    } else {
+        await this.reply('已关闭高清语音', true)
+    }
+  }
   async enableGroupContext (e) {
     const reg = /(关闭|打开)/
     const match = e.msg.match(reg)

utils/config.js

@@ -114,6 +114,7 @@ const defaultConfig = {
   cloudMode: 'url',
   cloudDPR: 1,
   ttsMode: 'vits-uma-genshin-honkai', // or azure
+  ttsHD: false,
   azureTTSKey: '',
   azureTTSRegion: '',
   azureTTSSpeaker: 'zh-CN-XiaochenNeural',

utils/uploadRecord.js

@@ -27,7 +27,9 @@ if (module) {
   try {
     pcm2slk = (await import('node-silk')).pcm2slk
   } catch (e) {
-    if (Config.cloudTranscode) {
+    if (Config.ttsHD) {
+      logger.info('已开启高清语音，电脑端将无法播放语音')
+    } else if (Config.cloudTranscode) {
       logger.warn('未安装node-silk，将尝试使用云转码服务进行合成')
     } else {
       Config.debug && logger.error(e)
@@ -49,8 +51,10 @@ async function uploadRecord (recordUrl, ttsMode = 'vits-uma-genshin-honkai') {
     tmpFile = `data/chatgpt/tts/tmp/${crypto.randomUUID()}.wav`
   }
   let result
-  if (pcm2slk) {
+  if (Config.ttsHD) {
-    result = await getPttBuffer(recordUrl, Bot.config.ffmpeg_path)
+    result = await getPttBuffer(recordUrl, Bot.config.ffmpeg_path, false)
+  } else if (pcm2slk) {
+    result = await getPttBuffer(recordUrl, Bot.config.ffmpeg_path, true)
   } else if (Config.cloudTranscode) {
     logger.mark('使用云转码silk进行高清语音生成:"')
     try {
@@ -196,21 +200,30 @@ async function uploadRecord (recordUrl, ttsMode = 'vits-uma-genshin-honkai') {
 
 export default uploadRecord
 
-async function getPttBuffer (file, ffmpeg = 'ffmpeg') {
+async function getPttBuffer (file, ffmpeg = 'ffmpeg', transcoding = true) {
   let buffer
   let time
   if (file instanceof Buffer || file.startsWith('base64://')) {
     // Buffer或base64
     const buf = file instanceof Buffer ? file : Buffer.from(file.slice(9), 'base64')
     const head = buf.slice(0, 7).toString()
-    if (head.includes('SILK') || head.includes('AMR')) {
+    if (head.includes('SILK') || head.includes('AMR') || !transcoding) {
-      return buf
+      const tmpfile = path.join(TMP_DIR, (0, uuid)())
-    } else {
-      const tmpfile = TMP_DIR + '/' + (0, uuid)()
       await fs.promises.writeFile(tmpfile, buf)
-      return audioTrans(tmpfile, ffmpeg)
+      let result = await getAudioTime(tmpfile, ffmpeg)
+      if (result.code == 1) time = result.data
+      fs.unlink(tmpfile, NOOP)
+      buffer = buf
+    } else {
+      const tmpfile = path.join(TMP_DIR, (0, uuid)())
+      let result = await getAudioTime(tmpfile, ffmpeg)
+      if (result.code == 1) time = result.data
+      await fs.promises.writeFile(tmpfile, buf)
+      buffer = await audioTrans(tmpfile, ffmpeg)
     }
   } else if (file.startsWith('http://') || file.startsWith('https://')) {
+    // 网络文件
+    // const readable = (await axios.get(file, { responseType: "stream" })).data;
     try {
       const headers = {
         'User-Agent': 'Dalvik/2.1.0 (Linux; U; Android 12; MI 9 Build/SKQ1.211230.001)'
@@ -220,11 +233,14 @@ async function getPttBuffer (file, ffmpeg = 'ffmpeg') {
         headers
       })
       const buf = Buffer.from(await response.arrayBuffer())
-      const tmpfile = TMP_DIR + '/' + (0, uuid)()
+      const tmpfile = path.join(TMP_DIR, (0, uuid)())
       await fs.promises.writeFile(tmpfile, buf)
       // await (0, pipeline)(readable.pipe(new DownloadTransform), fs.createWriteStream(tmpfile));
       const head = await read7Bytes(tmpfile)
-      if (head.includes('SILK') || head.includes('AMR')) {
+      let result = await getAudioTime(tmpfile, ffmpeg)
+      if (result.code == 1) time = result.data
+      if (head.includes('SILK') || head.includes('AMR') || !transcoding) {
+        // const buf = await fs.promises.readFile(tmpfile);
         fs.unlink(tmpfile, NOOP)
         buffer = buf
       } else {
@@ -236,7 +252,9 @@ async function getPttBuffer (file, ffmpeg = 'ffmpeg') {
     file = String(file).replace(/^file:\/{2}/, '')
     IS_WIN && file.startsWith('/') && (file = file.slice(1))
     const head = await read7Bytes(file)
-    if (head.includes('SILK') || head.includes('AMR')) {
+    let result = await getAudioTime(file, ffmpeg)
+    if (result.code == 1) time = result.data
+    if (head.includes('SILK') || head.includes('AMR') || !transcoding) {
       buffer = await fs.promises.readFile(file)
     } else {
       buffer = await audioTrans(file, ffmpeg)
@@ -245,6 +263,34 @@ async function getPttBuffer (file, ffmpeg = 'ffmpeg') {
   return { buffer, time }
 }
 
+async function getAudioTime (file, ffmpeg = 'ffmpeg') {
+  return new Promise((resolve, _reject) => {
+    (0, child_process.exec)(`${ffmpeg} -i "${file}"`, async (_error, _stdout, stderr) => {
+      try {
+        let time = stderr.split('Duration:')[1]?.split(',')[0].trim()
+        let arr = time?.split(':')
+        arr.reverse()
+        let n = 1
+        let s = 0
+        for (let val of arr) {
+          if (parseInt(val) > 0) s += parseInt(val) * n
+          n *= 60
+        }
+        resolve({
+          code: 1,
+          data: {
+            time,
+            seconds: s,
+            exec_text: stderr
+          }
+        })
+      } catch {
+        resolve({ code: -1 })
+      }
+    })
+  })
+}
+
 async function audioTrans (file, ffmpeg = 'ffmpeg') {
   const tmpfile = path.join(TMP_DIR, uuid())
   const cmd = IS_WIN