feat: convert wav by ffmpeg

.github/workflows/publish.yml

@@ -17,7 +17,7 @@ jobs:
         node-version: 18
 
     - name: install dependenies
-      run: npm install
+      run: export ELECTRON_SKIP_BINARY_DOWNLOAD=1 && npm install
 
     - name: build
       run: npm run build

package.json

@@ -5,7 +5,6 @@
   "main": "dist/main.js",
   "scripts": {
     "test": "echo \"Error: no test specified\" && exit 1",
-    "postinstall": "cross-env ELECTRON_SKIP_BINARY_DOWNLOAD=1 && npm install electron --no-save",
     "build": "npm run build-main && npm run build-preload && npm run build-renderer",
     "build-main": "webpack --config webpack.main.config.js",
     "build-preload": "webpack --config webpack.preload.config.js",
@@ -19,6 +18,7 @@
   "license": "ISC",
   "dependencies": {
     "express": "^4.18.2",
+    "fluent-ffmpeg": "^2.1.2",
     "json-bigint": "^1.0.0",
     "music-metadata": "^8.1.4",
     "silk-wasm": "^3.2.3",
@@ -28,12 +28,14 @@
   "devDependencies": {
     "@babel/preset-env": "^7.23.2",
     "@types/express": "^4.17.20",
+    "@types/fluent-ffmpeg": "^2.1.24",
     "@types/node": "^20.11.19",
     "@types/uuid": "^9.0.8",
     "@types/ws": "^8.5.10",
     "babel-loader": "^9.1.3",
     "copy-webpack-plugin": "^12.0.2",
     "cross-env": "^7.0.3",
+    "electron": "^29.0.1",
     "ts-loader": "^9.5.0",
     "typescript": "^5.2.2",
     "webpack": "^5.89.0",

src/common/utils.ts

@@ -5,6 +5,8 @@ import util from "util";
 import {encode, getDuration} from "silk-wasm";
 import fs from 'fs';
 import {v4 as uuidv4} from "uuid";
+import {exec} from "node:child_process";
+import ffmpeg from "fluent-ffmpeg"
 
 export const CONFIG_DIR = global.LiteLoader.plugins["LLOneBot"].path.data;
 
@@ -136,7 +138,22 @@ export function mergeNewProperties(newObj: any, oldObj: any) {
     });
 }
 
+export function checkFFMPEG(): boolean {
+    let ffmpegExist: boolean;
+    exec('ffmpeg -version', (error, stdout, stderr) => {
+        if (error) {
+            log('ffmpeg is not installed or not found in PATH:', error);
+            ffmpegExist = false
+        }
+        log('ffmpeg is installed. Version info:', stdout);
+        ffmpegExist = true;
+    });
+    return ffmpegExist;
+}
+
 export async function encodeSilk(filePath: string) {
+    const fsp = require("fs").promises
+
     function getFileHeader(filePath: string) {
         // 定义要读取的字节数
         const bytesToRead = 7;
@@ -154,6 +171,35 @@ export async function encodeSilk(filePath: string) {
         }
     }
 
+    function isWavFile(filePath: string) {
+        return new Promise((resolve, reject) => {
+            fs.open(filePath, 'r', (err, fd) => {
+                if (err) {
+                    reject(err);
+                    return;
+                }
+                // 读取前12个字节
+                const buffer = Buffer.alloc(12);
+                fs.read(fd, buffer, 0, 12, 0, (err, bytesRead, buffer) => {
+                    if (err) {
+                        reject(err);
+                        return;
+                    }
+                    fs.close(fd, (err) => {
+                        if (err) {
+                            reject(err);
+                            return;
+                        }
+                        // 检查RIFF头和WAVE格式标识
+                        const isRIFF = buffer.toString('utf8', 0, 4) === 'RIFF';
+                        const isWAVE = buffer.toString('utf8', 8, 12) === 'WAVE';
+                        resolve(isRIFF && isWAVE);
+                    });
+                });
+            });
+        });
+    }
+
     async function getAudioSampleRate(filePath: string) {
         try {
             const mm = await import('music-metadata');
@@ -172,22 +218,38 @@ export async function encodeSilk(filePath: string) {
         const pttPath = path.join(CONFIG_DIR, uuidv4());
         if (getFileHeader(filePath) !== "02232153494c4b") {
             log(`语音文件${filePath}需要转换`)
-            const sampleRate = await getAudioSampleRate(filePath) || 44100;
-            const silk = await encode(pcm, sampleRate);
-            fs.writeFileSync(pttPath, silk.data);
-            log(`语音文件${filePath}转换成功!`)
-            return {
-                converted: true,
-                path: pttPath,
-                duration: silk.duration,
-            };
-        } else {
-            const duration = getDuration(pcm);
-            return {
-                converted: false,
-                path: filePath,
-                duration: duration,
-            };
+            const isWav = await isWavFile(filePath);
+            if (!isWav) {
+                log(`语音文件${filePath}正在转换成wav`)
+                let voiceData = await fsp.readFile(filePath)
+                const wavPath = pttPath + ".wav"
+                await new Promise((resolve, reject) => {
+                    ffmpeg(filePath).toFormat("wav").on('end', function () {
+                        // console.log('转换完成');
+                        filePath = wavPath
+                    })
+                        .on('error', function (err) {
+                            // console.log('转换出错: ' + err.message);
+                        })
+                        .save(wavPath);
+                })
+                const sampleRate = await getAudioSampleRate(filePath) || 44100;
+                const silk = await encode(pcm, sampleRate);
+                fs.writeFileSync(pttPath, silk.data);
+                log(`语音文件${filePath}转换成功!`, pttPath)
+                return {
+                    converted: true,
+                    path: pttPath,
+                    duration: silk.duration,
+                };
+            } else {
+                const duration = getDuration(pcm);
+                return {
+                    converted: false,
+                    path: filePath,
+                    duration: duration,
+                };
+            }
         }
     } catch (error) {
         log("convert silk failed", error.stack);

src/ntqqapi/constructor.ts

@@ -8,8 +8,7 @@ import {
     SendTextElement
 } from "./types";
 import {NTQQApi} from "./ntcall";
-import {encodeSilk, log} from "../common/utils";
-import fs from "fs";
+import {encodeSilk} from "../common/utils";
 
 
 export class SendMsgElementConstructor {
@@ -83,9 +82,9 @@ export class SendMsgElementConstructor {
     static async ptt(pttPath: string): Promise<SendPttElement> {
         const {converted, path: silkPath, duration} = await encodeSilk(pttPath);
         // log("生成语音", silkPath, duration);
-        const {md5, fileName, path, fileSize} = await NTQQApi.uploadFile(silkPath);
+        const {md5, fileName, path, fileSize} = await NTQQApi.uploadFile(silkPath, ElementType.PTT);
         if (converted){
-            fs.unlink(silkPath, ()=>{});
+            // fs.unlink(silkPath, ()=>{});
         }
         return {
             elementType: ElementType.PTT,

src/ntqqapi/ntcall.ts

@@ -3,7 +3,9 @@ import {hookApiCallbacks, ReceiveCmd, registerReceiveHook, removeReceiveHook} fr
 import {log} from "../common/utils";
 import {
     ChatType,
-    Friend, FriendRequest,
+    ElementType,
+    Friend,
+    FriendRequest,
     Group,
     GroupMember,
     GroupNotifies,
@@ -329,7 +331,7 @@ export class NTQQApi {
     }
 
     // 上传文件到QQ的文件夹
-    static async uploadFile(filePath: string) {
+    static async uploadFile(filePath: string, elementType: ElementType=ElementType.PIC) {
         const md5 = await NTQQApi.getFileMd5(filePath);
         let ext = (await NTQQApi.getFileType(filePath))?.ext
         if (ext) {
@@ -344,7 +346,7 @@ export class NTQQApi {
                 path_info: {
                     md5HexStr: md5,
                     fileName: fileName,
-                    elementType: 2,
+                    elementType: elementType,
                     elementSubType: 0,
                     thumbSize: 0,
                     needCreate: true,