feature: format txt

.gitignore

@@ -13,4 +13,5 @@ subtitle
 perf
 log
 *.exe
-error.log
+error.log
+heard.config.json

src/app.tsx

@@ -3,13 +3,13 @@ import { AveRenderer, Grid, Window, getAppContext, IIconResource, IWindowCompone
 import { App, ThemePredefined_Dark } from "ave-ui";
 import { containerLayout, controlLayout } from "./layout";
 import { iconResource } from "./resource";
-import { safe } from "./common";
+import { safe, sleep } from "./common";
 import axios from "axios";
 import { ProgressType, Whisper } from "./whisper";
 import { useDragDrop } from "./hooks";
 import path from "path";
-import fs from "fs";
 import { getModelList, useModel } from "./config";
+import { writeSubtitleFile } from "./utils.ts";
 
 function onInit(app: App) {
 	const context = getAppContext();
@@ -23,9 +23,10 @@ function initTheme() {
 	themeDark.SetStyle(themeImage, 0);
 }
 
-enum ButtonText {
+enum Text {
 	GenerateSubtitle = "生成字幕",
-	OpenFile = "选择文件"
+	OpenFile = "选择文件",
+	ProgressLabel = "进度",
 }
 
 const modelOptions = getModelList().map((each, index) => {
@@ -47,11 +48,18 @@ export function Heard() {
 	const [title, setTitle] = useState("Heard");
 	const [whisperReady, setwhisperReady] = useState(false);
 	const [src, setSrc] = useState<string>("");
+	const [srcDesc, setSrcDesc] = useState<string>("");
+	const [srcList, setSrcList] = useState<Array<string>>([""]);
 	const promptRef = useRef<string>("");
-	useDragDrop((path) => setSrc(path));
+	useDragDrop((pathList) => {
+		setSrcList(pathList);
+		setSrcDesc(pathList.length === 1 ? pathList[0] : `${pathList.length} files`);
+		setProgressLabelText(`${Text.ProgressLabel}: 0/${pathList.length}`);
+	});
 
 	const [progress, setProgress] = useState<ProgressType>(ProgressType.None);
 	const [progressText, setProgressText] = useState<string>("");
+	const [progressLabelText, setProgressLabelText] = useState<string>(Text.ProgressLabel);
 	const [defaultSelectedKey, setDefaultSelectedKey] = useState<string>("1");
 	const [hintText, setHintText] = useState<string>("初始化中...");
 
@@ -64,7 +72,9 @@ export function Heard() {
 			console.log(`open file: ${filePath}`);
 
 			if (filePath) {
-				setSrc(filePath);
+				setSrcList([filePath]);
+				setSrcDesc(filePath);
+				setProgressLabelText(Text.ProgressLabel);
 			}
 		}),
 		[]
@@ -75,19 +85,23 @@ export function Heard() {
 			if (progress !== ProgressType.None) {
 				return;
 			}
-			whisper.transcribe(src, promptRef.current).then(
-				safe((response) => {
+
+			for (let i = 0; i < srcList.length; ++i) {
+				try {
+					const src = srcList[i];
+					setSrc(src);
+					setProgressLabelText(`${Text.ProgressLabel}: ${i + 1}/${srcList.length}: ${src}`);
+					const response = await whisper.transcribe(src, promptRef.current);
 					const data = response.data;
-					const dirName = path.dirname(src);
-					const fileName = path.basename(src);
-					console.log(`save subtitle json`, { dirName, fileName });
-
-					const outPath = path.resolve(dirName, `${fileName}.subtitle.json`);
-					fs.writeFileSync(outPath, JSON.stringify(data, null, 4), "utf8");
-				})
-			);
+
+					writeSubtitleFile(data, src);
+					await sleep(1500);
+				} catch (error) {
+					console.error(error?.message);
+				}
+			}
 		}),
-		[src]
+		[srcList]
 	);
 
 	const startWhisper = safe(() => {
@@ -174,10 +188,10 @@ export function Heard() {
 			<Grid style={{ layout: containerLayout }}>
 				<Grid style={{ area: containerLayout.areas.control, layout: controlLayout }}>
 					<Grid style={{ area: controlLayout.areas.openFile }}>
-						<Button text={ButtonText.OpenFile} langKey="OpenFile" iconInfo={{ name: "open-file" }} onClick={onOpenFile}></Button>
+						<Button text={Text.OpenFile} langKey="OpenFile" iconInfo={{ name: "open-file" }} onClick={onOpenFile}></Button>
 					</Grid>
 					<Grid style={{ area: controlLayout.areas.filePath, margin: "12dpx 0 0 0" }}>
-						<TextBox readonly border={false} text={src}></TextBox>
+						<TextBox readonly border={false} text={srcDesc}></TextBox>
 					</Grid>
 					{whisperReady ? (
 						<>
@@ -191,10 +205,10 @@ export function Heard() {
 								<TextBox ime onChange={onChangePrompt}></TextBox>
 							</Grid>
 							<Grid style={{ area: controlLayout.areas.generate, margin: "12dpx 0 0 0" }}>
-								<Button enable={progress === ProgressType.None} text={ButtonText.GenerateSubtitle} iconInfo={{ name: "cc", size: 16 }} onClick={onTranscribe}></Button>
+								<Button enable={progress === ProgressType.None} text={Text.GenerateSubtitle} iconInfo={{ name: "cc", size: 16 }} onClick={onTranscribe}></Button>
 							</Grid>
 							<Grid style={{ area: controlLayout.areas.progressLabel }}>
-								<TextBox readonly border={false} text="进度"></TextBox>
+								<TextBox readonly border={false} text={progressLabelText}></TextBox>
 							</Grid>
 							<Grid style={{ area: controlLayout.areas.progress }}>
 								<TextBox readonly border={false} text={progressText}></TextBox>

src/common/index.ts

@@ -5,14 +5,15 @@ export function assetsPath(name: string) {
 	return path.resolve(root, `./${name}`);
 }
 
-export function safe(callback: Function) {
-	return (...args: any[]) => {
+export function safe<T extends Function>(callback: T): T {
+	const f = (...args: any[]) => {
 		try {
 			return callback(...args);
 		} catch (error) {
 			console.error(error);
 		}
-	};
+	}
+	return f as unknown as T;
 }
 
 export async function sleep(time: number) {

src/config/index.ts

@@ -28,3 +28,37 @@ export function getModelList() {
 const globalConfig = {
 	model: "",
 };
+
+// --- user config
+export function getSubtitleFormat() {
+	try {
+		const config = getConfig();
+		const format = config?.format ?? defaultConfig.format;
+		console.log(`current subtitle format: ${format}`);
+		return format;
+	} catch (error) {
+		console.error("get subtitle format failed", { error });
+		return defaultConfig.format;
+	}
+}
+
+const defaultConfig = {
+	format: "srt",
+};
+
+function getConfig() {
+	const configPath = path.resolve(process.cwd(), "./heard.config.json");
+	if (!fs.existsSync(configPath)) {
+		console.log(`config not exist at ${configPath}, create it!`);
+		fs.writeFileSync(configPath, JSON.stringify(defaultConfig, null, 4), "utf-8");
+	}
+
+	try {
+		const configJson = JSON.parse(fs.readFileSync(configPath, "utf-8"));
+		console.log(`parse config succeed, use it`);
+		return configJson;
+	} catch (error) {
+		console.log(`parse config failed, ${error?.message}, use default config`);
+		return defaultConfig;
+	}
+}

src/hooks/useDragDrop.ts

@@ -1,27 +1,32 @@
 import { getAppContext } from "ave-react";
 import { useEffect, useState } from "react";
 import { DragDropImage, DropBehavior } from "ave-ui";
+import { safe } from "../common";
 
-export function useDragDrop(onPathChange?: (path: string) => void) {
-	const [path, setPath] = useState("");
+export function useDragDrop(onPathChange?: (pathList: Array<string>) => void) {
+	const [pathList, setPathList] = useState([""]);
 
 	useEffect(() => {
 		const context = getAppContext();
 		const window = context.getWindow();
 
-		window.OnDragMove((sender, dc) => {
-			if (1 === dc.FileGetCount()) {
-				dc.SetDropTip(DragDropImage.Copy, "打开此文件");
+		window.OnDragMove(
+			safe((sender, dc) => {
+				const fileCount = dc.FileGetCount();
+				dc.SetDropTip(DragDropImage.Copy, fileCount === 1 ? "选择此文件" : `选择文件x${fileCount}`);
 				dc.SetDropBehavior(DropBehavior.Copy);
-			}
-		});
+			})
+		);
 
-		window.OnDragDrop((sender, dc) => {
-			const filePath = dc.FileGet()[0];
-			setPath(path);
-			onPathChange?.(filePath ?? "");
-		});
+		window.OnDragDrop(
+			safe((sender, dc) => {
+				const filePaths = dc.FileGet();
+				console.log(`on drap drop, file paths:`, { filePaths });
+				setPathList(filePaths);
+				onPathChange?.(filePaths ?? [""]);
+			})
+		);
 	}, []);
 
-	return { path };
+	return { pathList };
 }

src/layout/index.ts

@@ -27,7 +27,7 @@ export const controlLayout = {
 		selectModel: { row: 5, column: 0, columnSpan: 3 },
 		generate: { row: 5, column: 3, columnSpan: 5 },
 
-		progressLabel: { row: 7, column: 0, columnSpan: 4 },
+		progressLabel: { row: 7, column: 0, columnSpan: 8 },
 		progress: { row: 9, column: 0, columnSpan: 8 }
 	},
 };

src/utils.ts/index.ts

@@ -0,0 +1,54 @@
+import path from "path";
+import fs from "fs";
+import { getSubtitleFormat } from "../config";
+import { safe } from "../common";
+
+interface ISubtitleData {
+	result: {
+		text: string;
+		language: string;
+		segments: ISubtitleSegment[];
+	};
+}
+
+interface ISubtitleSegment {
+	id: number;
+	seek: number;
+	start: number;
+	end: number;
+	text: string;
+	tokens: number[];
+	temperature: number;
+	avg_logprob: number;
+	compression_ratio: number;
+	no_speech_prob: number;
+}
+
+export const writeSubtitleFile = safe((data: ISubtitleData, src: string) => {
+	const dirName = path.dirname(src);
+	const fileName = path.basename(src);
+	const outPath = path.resolve(dirName, `${fileName}.subtitle.json`);
+	fs.writeFileSync(outPath, JSON.stringify(data, null, 4), "utf8");
+
+	const parsed = path.parse(fileName); // en.wav
+	const baseName = parsed.name; // en
+	const extension = parsed.ext; // .wav
+	console.log(`save subtitle json`, { dirName, fileName, baseName, extension });
+
+	const format = getSubtitleFormat();
+	if (format === "txt") {
+		const txtPath = path.resolve(dirName, `${baseName}.txt`);
+		const txtContent = getTxtContent(data);
+		fs.writeFileSync(txtPath, txtContent, "utf8");
+		console.log(`write txt subtitle succeed`);
+
+		const srtPath = path.resolve(dirName, `${baseName}.srt`);
+		if (fs.existsSync(srtPath)) {
+			fs.unlinkSync(srtPath);
+		}
+	}
+});
+
+function getTxtContent(data: ISubtitleData) {
+	return data.result.segments.map((each) => each.text).join("\n");
+}

src/whisper/index.ts

@@ -56,7 +56,7 @@ export class Whisper {
 		const exePath = path.resolve(dir, "./Whisper-API.exe");
 		if (fs.existsSync(dir) && fs.existsSync(exePath)) {
 			return new Promise((resolve, reject) => {
-				console.log("asrDir exists, start asr server", dir);
+				console.log("whisper dir exists, start whisper server", dir);
 
 				const name = getModel();
 				console.log("start whisper server with model: ", name);