Feat use ffmpeg static

enjoy/src/main.ts

@@ -43,9 +43,12 @@ protocol.registerSchemesAsPrivileged([
 app.on("ready", async () => {
   protocol.handle("enjoy", (request) => {
     let url = request.url.replace("enjoy://", "");
-    if (url.startsWith("library")) {
+    if (url.match(/library\/(audios|videos|recordings)/g)) {
       url = url.replace("library/", "");
       url = path.join(settings.userDataPath(), url);
+    } else if (url.startsWith("library")) {
+      url = url.replace("library/", "");
+      url = path.join(settings.libraryPath(), url);
     }
 
     return net.fetch(`file:///${url}`);

enjoy/src/main/ffmpeg.ts

@@ -5,24 +5,18 @@ import Ffmpeg from "fluent-ffmpeg";
 import log from "electron-log/main";
 import path from "path";
 import fs from "fs-extra";
+import settings from "./settings";
+
+Ffmpeg.setFfmpegPath(ffmpegPath);
+Ffmpeg.setFfprobePath(ffprobePath);
 
 const logger = log.scope("ffmpeg");
 export default class FfmpegWrapper {
-  public ffmpeg: Ffmpeg.FfmpegCommand;
-
-  constructor() {
-    const ff = Ffmpeg();
-    logger.debug("Using ffmpeg path:", ffmpegPath);
-    logger.debug("Using ffprobe path:", ffprobePath);
-    ff.setFfmpegPath(ffmpegPath);
-    ff.setFfprobePath(ffprobePath);
-    this.ffmpeg = ff;
-  }
-
   checkCommand(): Promise<boolean> {
+    const ffmpeg = Ffmpeg();
     const sampleFile = path.join(__dirname, "samples", "jfk.wav");
     return new Promise((resolve, _reject) => {
-      this.ffmpeg.input(sampleFile).getAvailableFormats((err, _formats) => {
+      ffmpeg.input(sampleFile).getAvailableFormats((err, _formats) => {
         if (err) {
           logger.error("Command not valid:", err);
           resolve(false);
@@ -35,8 +29,9 @@ export default class FfmpegWrapper {
   }
 
   generateMetadata(input: string): Promise<Ffmpeg.FfprobeData> {
+    const ffmpeg = Ffmpeg();
     return new Promise((resolve, reject) => {
-      this.ffmpeg
+      ffmpeg
         .input(input)
         .on("start", (commandLine) => {
           logger.info("Spawned FFmpeg with command: " + commandLine);
@@ -57,8 +52,9 @@ export default class FfmpegWrapper {
   }
 
   generateCover(input: string, output: string): Promise<string> {
+    const ffmpeg = Ffmpeg();
     return new Promise((resolve, reject) => {
-      this.ffmpeg
+      ffmpeg
         .input(input)
         .thumbnail({
           count: 1,
@@ -91,8 +87,9 @@ export default class FfmpegWrapper {
       fs.removeSync(output);
     }
 
+    const ffmpeg = Ffmpeg();
     return new Promise((resolve, reject) => {
-      this.ffmpeg
+      ffmpeg
         .input(input)
         .outputOptions("-ar", `${sampleRate}`)
         .on("error", (err) => {
@@ -112,8 +109,9 @@ export default class FfmpegWrapper {
     output: string,
     options: string[] = []
   ): Promise<string> {
+    const ffmpeg = Ffmpeg();
     return new Promise((resolve, reject) => {
-      this.ffmpeg
+      ffmpeg
         .input(input)
         .outputOptions(
           "-ar",
@@ -135,7 +133,7 @@ export default class FfmpegWrapper {
           }
 
           if (stderr) {
-            logger.error(stderr);
+            logger.info(stderr);
           }
 
           if (fs.existsSync(output)) {
@@ -176,9 +174,79 @@ export default class FfmpegWrapper {
     return this.convertToWav(input, output);
   }
 
+  async transcode(
+    input: string,
+    output?: string,
+    options?: string[]
+  ): Promise<string> {
+    if (input.match(/enjoy:\/\/library\/(audios|videos|recordings)/g)) {
+      input = path.join(
+        settings.userDataPath(),
+        input.replace("enjoy://library/", "")
+      );
+    } else if (input.startsWith("enjoy://library/")) {
+      input = path.join(
+        settings.libraryPath(),
+        input.replace("enjoy://library/", "")
+      );
+    }
+
+    if (!output) {
+      output = path.join(settings.cachePath(), `${path.basename(input)}.wav`);
+    }
+
+    if (output.startsWith("enjoy://library/")) {
+      output = path.join(
+        settings.libraryPath(),
+        output.replace("enjoy://library/", "")
+      );
+    }
+
+    options = options || ["-ar", "16000", "-ac", "1", "-c:a", "pcm_s16le"];
+
+    const ffmpeg = Ffmpeg();
+    return new Promise((resolve, reject) => {
+      ffmpeg
+        .input(input)
+        .outputOptions(...options)
+        .on("start", (commandLine) => {
+          logger.debug(`Trying to convert ${input} to ${output}`);
+          logger.info("Spawned FFmpeg with command: " + commandLine);
+          fs.ensureDirSync(path.dirname(output));
+        })
+        .on("end", (stdout, stderr) => {
+          if (stdout) {
+            logger.debug(stdout);
+          }
+
+          if (stderr) {
+            logger.info(stderr);
+          }
+
+          if (fs.existsSync(output)) {
+            resolve(output);
+          } else {
+            reject(new Error("FFmpeg command failed"));
+          }
+        })
+        .on("error", (err: Error) => {
+          logger.error(err);
+          reject(err);
+        })
+        .save(output);
+    });
+  }
+
   registerIpcHandlers() {
     ipcMain.handle("ffmpeg-check-command", async (_event) => {
       return await this.checkCommand();
     });
+
+    ipcMain.handle(
+      "ffmpeg-transcode",
+      async (_event, input, output, options) => {
+        return await this.transcode(input, output, options);
+      }
+    );
   }
 }

enjoy/src/main/whisper.ts

@@ -104,7 +104,7 @@ class Whipser {
           }
 
           if (stderr) {
-            logger.error("stderr", stderr);
+            logger.info("stderr", stderr);
           }
 
           if (stdout) {
@@ -199,7 +199,7 @@ class Whipser {
 
       command.stderr.on("data", (data) => {
         const output = data.toString();
-        logger.error(`stderr: ${output}`);
+        logger.info(`stderr: ${output}`);
         if (output.startsWith("whisper_print_progress_callback")) {
           const progress = parseInt(output.match(/\d+%/)?.[0] || "0");
           if (typeof progress === "number") onProgress(progress);

enjoy/src/preload.ts

@@ -389,6 +389,9 @@ contextBridge.exposeInMainWorld("__ENJOY_APP__", {
     check: () => {
       return ipcRenderer.invoke("ffmpeg-check-command");
     },
+    transcode: (input: string, output: string, options: string[]) => {
+      return ipcRenderer.invoke("ffmpeg-transcode", input, output, options);
+    },
   },
   download: {
     onState: (

enjoy/src/renderer/components/audios/audio-detail.tsx

@@ -173,6 +173,7 @@ export const AudioDetail = (props: { id?: string; md5?: string }) => {
   }, [audio]);
 
   useEffect(() => {
+    if (!initialized) return;
     if (!transcription) return;
 
     addDblistener(onTransactionUpdate);
@@ -192,7 +193,7 @@ export const AudioDetail = (props: { id?: string; md5?: string }) => {
       removeDbListener(onTransactionUpdate);
       EnjoyApp.whisper.removeProgressListeners();
     };
-  }, [md5, transcription]);
+  }, [md5, transcription, initialized]);
 
   if (!audio) {
     return <LoaderSpin />;
@@ -324,7 +325,7 @@ export const AudioDetail = (props: { id?: string; md5?: string }) => {
 
             {!transcription ? (
               <div className="flex items-center space-x-4">
-                <PingPoint colorClassName="bg-muted" />
+                <LoaderIcon className="w-4 h-4 animate-spin" />
                 <span>{t("loadingTranscription")}</span>
               </div>
             ) : transcription.result ? (

enjoy/src/renderer/components/videos/video-detail.tsx

@@ -179,6 +179,7 @@ export const VideoDetail = (props: { id?: string; md5?: string }) => {
   }, [video]);
 
   useEffect(() => {
+    if (!initialized) return;
     if (!transcription) return;
 
     addDblistener(onTransactionUpdate);
@@ -198,7 +199,7 @@ export const VideoDetail = (props: { id?: string; md5?: string }) => {
       removeDbListener(onTransactionUpdate);
       EnjoyApp.whisper.removeProgressListeners();
     };
-  }, [md5, transcription]);
+  }, [md5, transcription, initialized]);
 
   if (!video) {
     return <LoaderSpin />;
@@ -337,7 +338,7 @@ export const VideoDetail = (props: { id?: string; md5?: string }) => {
 
             {!transcription ? (
               <div className="flex items-center space-x-4">
-                <PingPoint colorClassName="bg-muted" />
+                <LoaderIcon className="w-4 h-4 animate-spin" />
                 <span>{t("loadingTranscription")}</span>
               </div>
             ) : transcription.result ? (

enjoy/src/renderer/context/app-settings-provider.tsx

@@ -16,7 +16,8 @@ type AppSettingsProviderState = {
   login?: (user: UserType) => void;
   logout?: () => void;
   setLibraryPath?: (path: string) => Promise<void>;
-  ffmpeg?: FFmpeg;
+  ffmpegWasm?: FFmpeg;
+  ffmpegValid?: boolean;
   EnjoyApp?: EnjoyAppType;
   language?: "en" | "zh-CN";
   switchLanguage?: (language: "en" | "zh-CN") => void;
@@ -44,7 +45,8 @@ export const AppSettingsProvider = ({
   const [webApi, setWebApi] = useState<Client>(null);
   const [user, setUser] = useState<UserType | null>(null);
   const [libraryPath, setLibraryPath] = useState("");
-  const [ffmpeg, setFfmpeg] = useState<FFmpeg>(null);
+  const [ffmpegWasm, setFfmpegWasm] = useState<FFmpeg>(null);
+  const [ffmpegValid, setFfmpegValid] = useState<boolean>(false);
   const [language, setLanguage] = useState<"en" | "zh-CN">();
   const [proxy, setProxy] = useState<ProxyConfigType>();
   const EnjoyApp = window.__ENJOY_APP__;
@@ -56,7 +58,7 @@ export const AppSettingsProvider = ({
     fetchUser();
     fetchLibraryPath();
     fetchLanguage();
-    loadFfmpegWASM();
+    prepareFfmpeg();
     fetchProxyConfig();
   }, []);
 
@@ -76,6 +78,14 @@ export const AppSettingsProvider = ({
     );
   }, [user, apiUrl, language]);
 
+  const prepareFfmpeg = async () => {
+    const valid = await EnjoyApp.ffmpeg.check();
+    setFfmpegValid(valid);
+    if (!valid) {
+      loadFfmpegWASM();
+    }
+  };
+
   const loadFfmpegWASM = async () => {
     const baseURL = "assets/libs";
     ffmpegRef.current.on("log", ({ message }) => {
@@ -101,7 +111,7 @@ export const AppSettingsProvider = ({
         wasmURL,
         workerURL,
       });
-      setFfmpeg(ffmpegRef.current);
+      setFfmpegWasm(ffmpegRef.current);
     } catch (err) {
       toast.error(err.message);
     }
@@ -195,7 +205,8 @@ export const AppSettingsProvider = ({
         logout,
         libraryPath,
         setLibraryPath: setLibraryPathHandler,
-        ffmpeg,
+        ffmpegValid,
+        ffmpegWasm,
         proxy,
         setProxy: setProxyConfigHandler,
         initialized,

enjoy/src/renderer/hooks/use-transcribe.tsx

@@ -12,26 +12,42 @@ import * as sdk from "microsoft-cognitiveservices-speech-sdk";
 import axios from "axios";
 import take from "lodash/take";
 import sortedUniqBy from "lodash/sortedUniqBy";
-import { groupTranscription, END_OF_WORD_REGEX, milisecondsToTimestamp } from "@/utils";
+import {
+  groupTranscription,
+  END_OF_WORD_REGEX,
+  milisecondsToTimestamp,
+} from "@/utils";
 
 export const useTranscribe = () => {
-  const { EnjoyApp, ffmpeg, user, webApi } = useContext(
+  const { EnjoyApp, ffmpegWasm, ffmpegValid, user, webApi } = useContext(
     AppSettingsProviderContext
   );
   const { whisperConfig, openai } = useContext(AISettingsProviderContext);
 
   const transcode = async (src: string, options?: string[]) => {
-    if (!ffmpeg?.loaded) return;
+    if (ffmpegValid) {
+      const output = `enjoy://library/cache/${src.split("/").pop()}.wav`;
+      const res = await EnjoyApp.ffmpeg.transcode(src, output, options);
+      console.log(res);
+      const data = await fetchFile(output);
+      return new Blob([data], { type: "audio/wav" });
+    } else {
+      return transcodeUsingWasm(src, options);
+    }
+  };
+
+  const transcodeUsingWasm = async (src: string, options?: string[]) => {
+    if (!ffmpegWasm?.loaded) return;
 
     options = options || ["-ar", "16000", "-ac", "1", "-c:a", "pcm_s16le"];
 
     try {
       const uri = new URL(src);
       const input = uri.pathname.split("/").pop();
       const output = input.replace(/\.[^/.]+$/, ".wav");
-      await ffmpeg.writeFile(input, await fetchFile(src));
-      await ffmpeg.exec(["-i", input, ...options, output]);
-      const data = await ffmpeg.readFile(output);
+      await ffmpegWasm.writeFile(input, await fetchFile(src));
+      await ffmpegWasm.exec(["-i", input, ...options, output]);
+      const data = await ffmpegWasm.readFile(output);
       return new Blob([data], { type: "audio/wav" });
     } catch (e) {
       toast.error(t("transcodeError"));

enjoy/src/types/enjoy-app.d.ts

@@ -228,6 +228,11 @@ type EnjoyAppType = {
   };
   ffmpeg: {
     check: () => Promise<boolean>;
+    transcode: (
+      input: string,
+      output: string,
+      options?: string[]
+    ) => Promise<string>;
   };
   download: {
     onState: (callback: (event, state) => void) => void;

enjoy/src/types/index.d.ts

@@ -90,16 +90,6 @@ type TransactionStateType = {
   record?: AudioType | UserType | RecordingType;
 };
 
-type FfmpegConfigType = {
-  os: string;
-  arch: string;
-  commandExists: boolean;
-  ffmpegPath?: string;
-  ffprobePath?: string;
-  scanDirs: string[];
-  ready: boolean;
-};
-
 type LookupType = {
   id: string;
   word: string;