From bf701c21103474e81474c12afe60b80ea320c7b4 Mon Sep 17 00:00:00 2001
From: idanran <96647698+idanran@users.noreply.github.com>
Date: Sat, 23 Mar 2024 11:57:13 +0000
Subject: [PATCH] fix: audio encoding exception in some cases

---
 package-lock.json        |  8 ++++----
 package.json             |  2 +-
 src/common/utils/file.ts | 37 +++++++++++++++++++++++--------------
 3 files changed, 28 insertions(+), 19 deletions(-)

diff --git a/package-lock.json b/package-lock.json
index 38413f2..01a8cae 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -14,7 +14,7 @@
         "file-type": "^19.0.0",
         "fluent-ffmpeg": "^2.1.2",
         "level": "^8.0.1",
-        "silk-wasm": "^3.2.4",
+        "silk-wasm": "^3.3.2",
         "utf-8-validate": "^6.0.3",
         "uuid": "^9.0.1",
         "ws": "^8.16.0"
@@ -5895,9 +5895,9 @@
       }
     },
     "node_modules/silk-wasm": {
-      "version": "3.2.4",
-      "resolved": "https://registry.npmjs.org/silk-wasm/-/silk-wasm-3.2.4.tgz",
-      "integrity": "sha512-oBkXmdIRl7cyzpoXEeEVN7v1M2yCnH1/bN8oANoYTvCqbYa5lM/CGJP47DYbpUFVO9PUpm58KP/HZaVzt4J6jw=="
+      "version": "3.3.2",
+      "resolved": "https://registry.npmjs.org/silk-wasm/-/silk-wasm-3.3.2.tgz",
+      "integrity": "sha512-CXG/hikSInuI/A+EIPoq+z1BRv4SJdhIQKtscOiwZS3udSOr+4DBPYmqNtlV3AKldjEhJChIRCsBtWcWRe54ng=="
     },
     "node_modules/slash": {
       "version": "4.0.0",
diff --git a/package.json b/package.json
index e56cea1..80d9d20 100644
--- a/package.json
+++ b/package.json
@@ -19,7 +19,7 @@
     "file-type": "^19.0.0",
     "fluent-ffmpeg": "^2.1.2",
     "level": "^8.0.1",
-    "silk-wasm": "^3.2.4",
+    "silk-wasm": "^3.3.2",
     "utf-8-validate": "^6.0.3",
     "uuid": "^9.0.1",
     "ws": "^8.16.0"
diff --git a/src/common/utils/file.ts b/src/common/utils/file.ts
index 0da7968..956374b 100644
--- a/src/common/utils/file.ts
+++ b/src/common/utils/file.ts
@@ -3,7 +3,7 @@ import fsPromise from "fs/promises";
 import crypto from "crypto";
 import ffmpeg from "fluent-ffmpeg";
 import util from "util";
-import {encode, getDuration, isWav} from "silk-wasm";
+import {encode, getDuration, isWav, getWavFileInfo} from "silk-wasm";
 import path from "node:path";
 import {v4 as uuidv4} from "uuid";
 import {checkFfmpeg, DATA_DIR, log, TEMP_DIR} from "./index";
@@ -91,7 +91,7 @@ export async function encodeSilk(filePath: string) {
         return isWav(fs.readFileSync(filePath));
     }
 
-    async function guessDuration(pttPath: string){
+    async function guessDuration(pttPath: string) {
         const pttFileInfo = await fsPromise.stat(pttPath)
         let duration = pttFileInfo.size / 1024 / 3  // 3kb/s
         duration = Math.floor(duration)
@@ -100,9 +100,9 @@ export async function encodeSilk(filePath: string) {
         return duration
     }
 
-    function verifyDuration(oriDuration: number, guessDuration: number){
+    function verifyDuration(oriDuration: number, guessDuration: number) {
         // 单位都是秒
-        if (oriDuration - guessDuration > 10){
+        if (oriDuration - guessDuration > 10) {
             return guessDuration
         }
         oriDuration = Math.max(1, oriDuration)
@@ -126,10 +126,8 @@ export async function encodeSilk(filePath: string) {
             log(`语音文件${filePath}需要转换成silk`)
             const _isWav = await isWavFile(filePath);
             const wavPath = pttPath + ".wav"
-            if (!_isWav) {
-                log(`语音文件${filePath}正在转换成wav`)
-                // let voiceData = await fsp.readFile(filePath)
-                await new Promise((resolve, reject) => {
+            const convert = async () => {
+                return await new Promise((resolve, reject) => {
                     const ffmpegPath = getConfigUtil().getConfig().ffmpeg;
                     if (ffmpegPath) {
                         ffmpeg.setFfmpegPath(ffmpegPath);
@@ -148,10 +146,21 @@ export async function encodeSilk(filePath: string) {
                         });
                 })
             }
-            // const sampleRate = await getAudioSampleRate(filePath) || 0;
-            // log("音频采样率", sampleRate)
-            const pcm = fs.readFileSync(filePath);
-            const silk = await encode(pcm, 0);
+            let wav: Buffer
+            if (!_isWav) {
+                log(`语音文件${filePath}正在转换成wav`)
+                await convert()
+            } else {
+                wav = fs.readFileSync(filePath)
+                const allowSampleRate = [8000, 12000, 16000, 24000, 32000, 44100, 48000]
+                const { fmt } = getWavFileInfo(wav)
+                if (!allowSampleRate.includes(fmt.sampleRate)) {
+                    wav = undefined
+                    await convert()
+                }
+            }
+            wav ||= fs.readFileSync(filePath);
+            const silk = await encode(wav, 0);
             fs.writeFileSync(pttPath, silk.data);
             fs.unlink(wavPath, (err) => {
             });
@@ -286,9 +295,9 @@ export async function uri2local(uri: string, fileName: string = null): Promise<U
     } else if (url.protocol == "http:" || url.protocol == "https:") {
         // 下载文件
         let buffer: Buffer = null;
-        try{
+        try {
             buffer = await httpDownload(uri);
-        }catch (e) {
+        } catch (e) {
             res.errMsg = `${url}下载失败,` + e.toString()
             return res
         }