Merge pull request #100 from guardian/support-translation

Support translation of input audio to english

packages/api/src/index.ts

@@ -125,11 +125,35 @@ const getApp = async () => {
 				body.data.fileName,
 				signedUrl,
 				body.data.languageCode,
+				false,
 			);
 			if (isSqsFailure(sendResult)) {
 				res.status(500).send(sendResult.errorMsg);
 				return;
 			}
+			if (body.data.translationRequested) {
+				const translationSendResult =
+					await generateOutputSignedUrlAndSendMessage(
+						s3Key,
+						sqsClient,
+						config.app.taskQueueUrl,
+						config.app.transcriptionOutputBucket,
+						config.aws.region,
+						userEmail,
+						body.data.fileName,
+						signedUrl,
+						body.data.languageCode,
+						true,
+					);
+				if (isSqsFailure(translationSendResult)) {
+					res
+						.status(500)
+						.send(
+							`Translation request failed: ${translationSendResult.errorMsg}`,
+						);
+					return;
+				}
+			}
 			logger.info('API successfully sent the message to SQS', {
 				id: s3Key,
 				filename: body.data.fileName,
@@ -209,7 +233,7 @@ const getApp = async () => {
 			}
 			const exportResult = await createTranscriptDocument(
 				config,
-				`${parsedItem.data.originalFilename} transcript`,
+				`${parsedItem.data.originalFilename} transcript${parsedItem.data.isTranslation ? ' (English translation)' : ''}`,
 				exportRequest.data.oAuthTokenResponse,
 				transcriptText.text,
 			);

packages/backend-common/src/dynamodb.ts

@@ -38,6 +38,7 @@ export const TranscriptionDynamoItem = z.object({
 	transcriptKeys: TranscriptKeys,
 	userEmail: z.string(),
 	completedAt: z.optional(z.string()), // dynamodb can't handle dates so we need to use an ISO date
+	isTranslation: z.boolean(),
 });
 
 export type TranscriptionDynamoItem = z.infer<typeof TranscriptionDynamoItem>;

packages/backend-common/src/sqs.ts

@@ -57,7 +57,7 @@ export const isSqsFailure = (
 ): result is SQSFailure => result.status === AWSStatus.Failure;
 
 export const generateOutputSignedUrlAndSendMessage = async (
-	id: string,
+	s3Key: string,
 	client: SQSClient,
 	queueUrl: string,
 	outputBucket: string,
@@ -66,27 +66,30 @@ export const generateOutputSignedUrlAndSendMessage = async (
 	originalFilename: string,
 	inputSignedUrl: string,
 	languageCode: LanguageCode,
+	translate: boolean,
 ): Promise<SendResult> => {
 	const signedUrls = await generateOutputSignedUrls(
-		id,
+		s3Key,
 		region,
 		outputBucket,
 		userEmail,
-		originalFilename,
 		7,
+		translate,
 	);
 
+	const jobId = translate ? `${s3Key}-translation` : s3Key;
 	const job: TranscriptionJob = {
-		id, // id of the source file
+		id: jobId, // id of the source file
 		inputSignedUrl,
 		sentTimestamp: new Date().toISOString(),
 		userEmail,
 		transcriptDestinationService: DestinationService.TranscriptionService,
 		originalFilename,
 		outputBucketUrls: signedUrls,
 		languageCode,
+		translate,
 	};
-	return await sendMessage(client, queueUrl, JSON.stringify(job), id);
+	return await sendMessage(client, queueUrl, JSON.stringify(job), s3Key);
 };
 
 const sendMessage = async (
@@ -278,13 +281,14 @@ const generateOutputSignedUrls = async (
 	region: string,
 	outputBucket: string,
 	userEmail: string,
-	originalFilename: string,
 	expiresInDays: number,
+	translate: boolean,
 ): Promise<OutputBucketUrls> => {
+	const fileName = `${id}${translate ? '-translation' : ''}`;
 	const expiresIn = expiresInDays * 24 * 60 * 60;
-	const srtKey = `srt/${id}.srt`;
-	const jsonKey = `json/${id}.json`;
-	const textKey = `text/${id}.txt`;
+	const srtKey = `srt/${fileName}.srt`;
+	const jsonKey = `json/${fileName}.json`;
+	const textKey = `text/${fileName}.txt`;
 	const srtSignedS3Url = await getSignedUploadUrl(
 		region,
 		outputBucket,

packages/client/src/components/UploadForm.tsx

@@ -9,14 +9,15 @@ import {
 	TranscribeFileRequestBody,
 } from '@guardian/transcription-service-common';
 import { AuthContext } from '@/app/template';
-import { FileInput, Label, Select } from 'flowbite-react';
+import { Checkbox, FileInput, Label, Select } from 'flowbite-react';
 import { RequestStatus } from '@/types';
 import { iconForStatus, InfoMessage } from '@/components/InfoMessage';
 
 const uploadFileAndTranscribe = async (
 	file: File,
 	token: string,
 	languageCode: LanguageCode,
+	translationRequested: boolean,
 ) => {
 	const blob = new Blob([file as BlobPart]);
 
@@ -42,6 +43,7 @@ const uploadFileAndTranscribe = async (
 		s3Key: body.data.s3Key,
 		fileName: file.name,
 		languageCode,
+		translationRequested,
 	};
 
 	const sendMessageResponse = await authFetch('/api/transcribe-file', token, {
@@ -83,6 +85,8 @@ export const UploadForm = () => {
 	const [languageCodeValid, setLanguageCodeValid] = useState<
 		boolean | undefined
 	>(undefined);
+	const [translationRequested, setTranslationRequested] =
+		useState<boolean>(false);
 	const { token } = useContext(AuthContext);
 
 	const reset = () => {
@@ -143,10 +147,24 @@ export const UploadForm = () => {
 							role="alert"
 						>
 							<span className="font-medium">Upload complete. </span>{' '}
-							Transcription in progress - check your email for the completed
-							transcript. The service can take a few minutes to start up, but
-							thereafter the transcription process is typically shorter than the
-							length of the media file.{' '}
+							<p>
+								Transcription in progress - check your email for the completed
+								transcript.{' '}
+							</p>
+							<div className="font-medium">
+								<p>
+									{' '}
+									The service can take a few minutes to start up, but thereafter
+									the transcription process is typically shorter than the length
+									of the media file.{' '}
+								</p>
+								<p>
+									If you have requested a translation, you will receive 2 emails
+									- one for the transcription in the original language, another
+									for the english translation. The emails will arrive at
+									different times
+								</p>
+							</div>
 							<button
 								onClick={() => reset()}
 								className="font-medium text-blue-600 underline dark:text-blue-500 hover:no-underline"
@@ -191,6 +209,7 @@ export const UploadForm = () => {
 				file,
 				token,
 				mediaFileLanguageCode,
+				translationRequested,
 			);
 			if (!result) {
 				setUploads((prev) =>
@@ -261,6 +280,30 @@ export const UploadForm = () => {
 						))}
 					</Select>
 				</div>
+				{mediaFileLanguageCode !== 'en' && (
+					<div className="mb-6">
+						<div className="flex gap-2">
+							<div className="flex h-5 items-center">
+								<Checkbox
+									id="translation"
+									checked={translationRequested}
+									onChange={() =>
+										setTranslationRequested(!translationRequested)
+									}
+								/>
+							</div>
+							<div className="flex flex-col">
+								<Label htmlFor="shipping">Request English translation</Label>
+								<div className="text-gray-500 dark:text-gray-300">
+									<span className="text-xs font-normal">
+										You will receive two documents - a transcript in the
+										original language and a translation in English.
+									</span>
+								</div>
+							</div>
+						</div>
+					</div>
+				)}
 				<button
 					type="submit"
 					className="text-white bg-blue-700 hover:bg-blue-800 focus:ring-4 focus:outline-none focus:ring-blue-300 font-medium rounded-lg text-sm w-full sm:w-auto px-5 py-2.5 text-center dark:bg-blue-600 dark:hover:bg-blue-700 dark:focus:ring-blue-800"

packages/common/src/types.ts

@@ -45,6 +45,7 @@ export const TranscriptionJob = z.object({
 	transcriptDestinationService: z.nativeEnum(DestinationService),
 	outputBucketUrls: OutputBucketUrls,
 	languageCode: zodLanguageCode,
+	translate: z.boolean(),
 });
 
 export type TranscriptionJob = z.infer<typeof TranscriptionJob>;
@@ -53,6 +54,7 @@ const TranscriptionOutputBase = z.object({
 	id: z.string(),
 	originalFilename: z.string(),
 	userEmail: z.string(),
+	isTranslation: z.boolean(),
 });
 
 export const TranscriptionOutputSuccess = TranscriptionOutputBase.extend({
@@ -140,6 +142,7 @@ export const transcribeFileRequestBody = z.object({
 	s3Key: z.string(),
 	fileName: z.string(),
 	languageCode: zodLanguageCode,
+	translationRequested: z.boolean(),
 });
 export type TranscribeFileRequestBody = z.infer<
 	typeof transcribeFileRequestBody

packages/output-handler/src/index.ts

@@ -27,18 +27,23 @@ const successMessageBody = (
 	transcriptId: string,
 	originalFilename: string,
 	rootUrl: string,
+	isTranslation: boolean,
 ): string => {
 	const exportUrl = `${rootUrl}/export?transcriptId=${transcriptId}`;
 	return `
-		<h1>Transcript for ${originalFilename} ready</h1>
+		<h1>${isTranslation ? 'English translation ' : 'Transcription'} for ${originalFilename} ready</h1>
 		<p>Click <a href="${exportUrl}">here</a> to export to a google doc.</p>
 		<p><b>Note:</b> transcripts will expire after 7 days. Export your transcript to a doc now if you want to keep it. </p>
 	`;
 };
 
-const failureMessageBody = (originalFilename: string, id: string): string => {
+const failureMessageBody = (
+	originalFilename: string,
+	id: string,
+	isTranslation: boolean,
+): string => {
 	return `
-		<h1>Transcription for ${originalFilename} has failed.</h1>
+		<h1>${isTranslation ? 'English translation ' : 'Transcription'}for ${originalFilename} has failed.</h1>
 		<p>Please make sure that the file is a valid audio or video file.</p>
 		<p>Contact the digital investigations team for support.</p>
 		<p>Transcription ID: ${id}</p>
@@ -61,6 +66,7 @@ const handleTranscriptionSuccess = async (
 		},
 		userEmail: transcriptionOutput.userEmail,
 		completedAt: new Date().toISOString(),
+		isTranslation: transcriptionOutput.isTranslation,
 	};
 
 	try {
@@ -74,11 +80,12 @@ const handleTranscriptionSuccess = async (
 			sesClient,
 			config.app.emailNotificationFromAddress,
 			transcriptionOutput.userEmail,
-			`Transcription complete for ${transcriptionOutput.originalFilename}`,
+			`${transcriptionOutput.isTranslation ? 'English translation' : 'Transcription'} complete for ${transcriptionOutput.originalFilename}`,
 			successMessageBody(
 				transcriptionOutput.id,
 				transcriptionOutput.originalFilename,
 				config.app.rootUrl,
+				transcriptionOutput.isTranslation,
 			),
 		);
 
@@ -107,10 +114,11 @@ const handleTranscriptionFailure = async (
 			sesClient,
 			config.app.emailNotificationFromAddress,
 			transcriptionOutput.userEmail,
-			`Transcription failed for ${transcriptionOutput.originalFilename}`,
+			`${transcriptionOutput.isTranslation ? 'English translation ' : 'Transcription'} failed for ${transcriptionOutput.originalFilename}`,
 			failureMessageBody(
 				transcriptionOutput.originalFilename,
 				transcriptionOutput.id,
+				transcriptionOutput.isTranslation,
 			),
 		);
 

packages/worker/src/index.ts

@@ -111,6 +111,7 @@ const publishTranscriptionOutputFailure = async (
 		status: 'FAILURE',
 		userEmail: job.userEmail,
 		originalFilename: job.originalFilename,
+		isTranslation: job.translate,
 	};
 	try {
 		await publishTranscriptionOutput(sqsClient, destination, failureMessage);
@@ -251,6 +252,7 @@ const pollTranscriptionQueue = async (
 			numberOfThreads,
 			config.app.stage === 'PROD' ? 'medium' : 'tiny',
 			job.languageCode,
+			job.translate,
 		);
 
 		// if we've received an interrupt signal we don't want to perform a half-finished transcript upload/publish as
@@ -286,6 +288,7 @@ const pollTranscriptionQueue = async (
 			userEmail: job.userEmail,
 			originalFilename: job.originalFilename,
 			outputBucketKeys,
+			isTranslation: job.translate,
 		};
 
 		await publishTranscriptionOutput(

packages/worker/src/transcribe.ts

@@ -168,6 +168,7 @@ export const getTranscriptionText = async (
 	numberOfThreads: number,
 	model: WhisperModel,
 	languageCode: LanguageCode,
+	translate: boolean,
 ): Promise<TranscriptionResult> => {
 	try {
 		const { fileName, metadata } = await transcribe(
@@ -176,6 +177,7 @@ export const getTranscriptionText = async (
 			numberOfThreads,
 			model,
 			languageCode,
+			translate,
 		);
 
 		const srtPath = path.resolve(path.parse(file).dir, `${fileName}.srt`);
@@ -190,7 +192,7 @@ export const getTranscriptionText = async (
 
 		return { transcripts, metadata };
 	} catch (error) {
-		logger.error(`Could not read the transcripts result`);
+		logger.error(`Could not read the transcript result`);
 		throw error;
 	}
 };
@@ -223,6 +225,7 @@ export const transcribe = async (
 	numberOfThreads: number,
 	model: WhisperModel,
 	languageCode: LanguageCode,
+	translate: boolean,
 ) => {
 	const fileName = path.parse(file).name;
 	const containerOutputFilePath = path.resolve(CONTAINER_FOLDER, fileName);
@@ -246,10 +249,14 @@ export const transcribe = async (
 			containerOutputFilePath,
 			'--language',
 			languageCode,
+			`${translate ? '--translate' : ''}`,
 		]);
 		const metadata = extractWhisperStderrData(result.stderr);
 		logger.info('Transcription finished successfully', metadata);
-		return { fileName, metadata };
+		return {
+			fileName: `${fileName}`,
+			metadata,
+		};
 	} catch (error) {
 		logger.error(`Transcription failed due to `, error);
 		throw error;