Downsizing sample for some language in speech to text

packages/predictions/__tests__/Providers/AWSAIConvertPredictionsProvider-unit-test.ts

@@ -217,6 +217,31 @@ describe('Predictions convert provider test', () => {
 				predictionsProvider.convert(validSpeechToTextInput)
 			).rejects.toMatch('region not configured for transcription');
 		});
+		test('Error languageCode not configured ', () => {
+			AmazonAIConvertPredictionsProvider.serializeDataFromTranscribe = jest.fn(
+				() => {
+					return 'Hello how are you';
+				}
+			);
+
+			const predictionsProvider = new AmazonAIConvertPredictionsProvider();
+			const speechGenOptions = {
+				transcription: {
+					region: 'us-west-2',
+					proxy: false,
+				},
+			};
+			predictionsProvider.configure(speechGenOptions);
+			jest.spyOn(Credentials, 'get').mockImplementationOnce(() => {
+				return Promise.resolve(credentials);
+			});
+
+			return expect(
+				predictionsProvider.convert(validSpeechToTextInput)
+			).rejects.toMatch(
+				'languageCode not configured or provided for transcription'
+			);
+		});
 		test('Happy case ', () => {
 			AmazonAIConvertPredictionsProvider.serializeDataFromTranscribe = jest.fn(
 				() => {
@@ -247,5 +272,35 @@ describe('Predictions convert provider test', () => {
 				},
 			} as SpeechToTextOutput);
 		});
+		test('Downsized Happy case ', async () => {
+			AmazonAIConvertPredictionsProvider.serializeDataFromTranscribe = jest.fn(
+				() => {
+					return 'Bonjour, comment vas tu?';
+				}
+			);
+			const downsampleBufferSpyon = jest.spyOn(
+				AmazonAIConvertPredictionsProvider.prototype as any,
+				'downsampleBuffer'
+			);
+
+			const predictionsProvider = new AmazonAIConvertPredictionsProvider();
+			const speechGenOptions = {
+				transcription: {
+					region: 'us-west-2',
+					proxy: false,
+					defaults: {
+						language: 'fr-FR',
+					},
+				},
+			};
+			predictionsProvider.configure(speechGenOptions);
+			jest.spyOn(Credentials, 'get').mockImplementationOnce(() => {
+				return Promise.resolve(credentials);
+			});
+
+			await predictionsProvider.convert(validSpeechToTextInput);
+			expect(downsampleBufferSpyon).toBeCalled();
+			downsampleBufferSpyon.mockClear();
+		});
 	});
 });

packages/predictions/src/Providers/AmazonAIConvertPredictionsProvider.ts

@@ -28,6 +28,8 @@ import { fromUtf8, toUtf8 } from '@aws-sdk/util-utf8-node';
 const logger = new Logger('AmazonAIConvertPredictionsProvider');
 const eventBuilder = new EventStreamMarshaller(toUtf8, fromUtf8);
 
+const LANGUAGES_CODE_IN_8KHZ = ['fr-FR', 'en-AU', 'en-GB', 'fr-CA'];
+
 export class AmazonAIConvertPredictionsProvider extends AbstractConvertPredictionsProvider {
 	private translateClient: TranslateClient;
 	private pollyClient: PollyClient;
@@ -178,10 +180,17 @@ export class AmazonAIConvertPredictionsProvider extends AbstractConvertPredictio
 					languageCode: language,
 				});
 
+				let sourceBytes = source.bytes;
+				if (LANGUAGES_CODE_IN_8KHZ.includes(languageCode)) {
+					sourceBytes = this.downsampleBuffer({
+						buffer: source.bytes,
+						outputSampleRate: 8000,
+					});
+				}
 				try {
 					const fullText = await this.sendDataToTranscribe({
 						connection,
-						raw: source.bytes,
+						raw: sourceBytes,
 					});
 					return {
 						transcription: {
@@ -206,9 +215,7 @@ export class AmazonAIConvertPredictionsProvider extends AbstractConvertPredictio
 		const transcribeMessage = eventBuilder.unmarshall(
 			Buffer.from(message.data)
 		);
-		const transcribeMessageJson = JSON.parse(
-			toUtf8(transcribeMessage.body)
-		);
+		const transcribeMessageJson = JSON.parse(toUtf8(transcribeMessage.body));
 		if (transcribeMessage.headers[':message-type'].value === 'exception') {
 			logger.debug(
 				'exception',
@@ -327,14 +334,13 @@ export class AmazonAIConvertPredictionsProvider extends AbstractConvertPredictio
 	}
 
 	private inputSampleRate = 44100;
-	private outputSampleRate = 16000;
 
-	private downsampleBuffer({ buffer }) {
-		if (this.outputSampleRate === this.inputSampleRate) {
+	private downsampleBuffer({ buffer, outputSampleRate = 16000 }) {
+		if (outputSampleRate === this.inputSampleRate) {
 			return buffer;
 		}
 
-		const sampleRateRatio = this.inputSampleRate / this.outputSampleRate;
+		const sampleRateRatio = this.inputSampleRate / outputSampleRate;
 		const newLength = Math.round(buffer.length / sampleRateRatio);
 		const result = new Float32Array(newLength);
 		let offsetResult = 0;
@@ -399,7 +405,9 @@ export class AmazonAIConvertPredictionsProvider extends AbstractConvertPredictio
 			`wss://transcribestreaming.${region}.amazonaws.com:8443`,
 			'/stream-transcription-websocket?',
 			`media-encoding=pcm&`,
-			`sample-rate=16000&`,
+			`sample-rate=${
+				LANGUAGES_CODE_IN_8KHZ.includes(languageCode) ? '8000' : '16000'
+			}&`,
 			`language-code=${languageCode}`,
 		].join('');