Merge pull request #79 from ChetanXpro/fix/mislabeled-wav-handling

fix: handling of mislabeled WAV files with non-WAV content

Author: ChetanXpro

src/utils.ts

@@ -8,25 +8,66 @@ export const checkIfFileExists = (filePath: string) => {
 	}
 }
 
-export const convertToWavType = async (inputFilePath: string, verbose: boolean) => {
-	const fileExtension = path.extname(inputFilePath).toLowerCase()
+async function isValidWavHeader(filePath) {
+	return new Promise((resolve, reject) => {
+		const readable = fs.createReadStream(filePath, { end: 11 })
+		let data = ''
 
-	if (fileExtension === '.wav') {
-		return inputFilePath
-	}
+		readable.on('data', chunk => {
+			data += chunk.toString('binary')
+		})
 
-	const outputFilePath = path.join(path.dirname(inputFilePath), `${path.basename(inputFilePath, fileExtension)}.wav`)
+		readable.on('end', () => {
+			const isValid = data.startsWith('RIFF') || data.startsWith('RIFX')
+			resolve(isValid)
+		})
 
-	if (verbose) {
-		console.log(`[Nodejs-whisper] Converting audio to WAV file type...\n`)
-	}
+		readable.on('error', err => {
+			reject(err)
+		})
+	})
+}
 
-	const command = `ffmpeg -nostats -loglevel error -y -i "${inputFilePath}" -ar 16000 -ac 1 -c:a pcm_s16le "${outputFilePath}"`
-	const result = shell.exec(command, { silent: !verbose })
+export const convertToWavType = async (inputFilePath, verbose) => {
+	const fileExtension = path.extname(inputFilePath).toLowerCase()
 
-	if (result.code !== 0) {
-		throw new Error(`[Nodejs-whisper] Failed to convert audio file: ${result.stderr}`)
+	if (verbose) {
+		console.log(`[Nodejs-whisper] Checking if the file is a valid WAV: ${inputFilePath}`)
 	}
 
-	return outputFilePath
+	if (fileExtension === '.wav') {
+		const isWav = await isValidWavHeader(inputFilePath)
+		if (isWav) {
+			if (verbose) {
+				console.log(`[Nodejs-whisper] File is a valid WAV file.`)
+			}
+			return inputFilePath
+		} else {
+			if (verbose) {
+				console.log(`[Nodejs-whisper] File has a .wav extension but is not a valid WAV, overwriting...`)
+			}
+			// Overwrite the original WAV file
+			const command = `ffmpeg -nostats -loglevel error -y -i "${inputFilePath}" -ar 16000 -ac 1 -c:a pcm_s16le "${inputFilePath}"`
+			const result = shell.exec(command, { silent: !verbose })
+			if (result.code !== 0) {
+				throw new Error(`[Nodejs-whisper] Failed to convert audio file: ${result.stderr}`)
+			}
+			return inputFilePath
+		}
+	} else {
+		// Convert to a new WAV file
+		const outputFilePath = path.join(
+			path.dirname(inputFilePath),
+			`${path.basename(inputFilePath, fileExtension)}.wav`
+		)
+		if (verbose) {
+			console.log(`[Nodejs-whisper] Converting to a new WAV file: ${outputFilePath}`)
+		}
+		const command = `ffmpeg -nostats -loglevel error -y -i "${inputFilePath}" -ar 16000 -ac 1 -c:a pcm_s16le "${outputFilePath}"`
+		const result = shell.exec(command, { silent: !verbose })
+		if (result.code !== 0) {
+			throw new Error(`[Nodejs-whisper] Failed to convert audio file: ${result.stderr}`)
+		}
+		return outputFilePath
+	}
 }