aws_transcribe_streaming 0.1.0

AWS Transcribe Streaming client for producing real-time transcriptions for your media content using HTTP/2.

example/main.dart

import 'dart:async';
import 'dart:typed_data';

import 'package:aws_common/aws_common.dart';
import 'package:aws_transcribe_streaming/aws_transcribe_streaming.dart';

void main() async {
  // Create a client.
  final transcribeStreamingClient = TranscribeStreamingClient(
    region: 'eu-central-1',
    // Provide credentials with `transcribe:StartStreamTranscription` permission
    credentialsProvider: StaticCredentialsProvider(AWSCredentials(
      'ASIAIOEXAMPLEEXAMPLE', // accessKeyId
      'wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY', // secretAccessKey
      'AQoDYXdzEJr...', // sessionToken
      DateTime.now().add(const Duration(hours: 1)), // expiration
    )),
  );

  late final StartStreamTranscriptionResponse response;
  late final Stream<TranscriptEvent> transcriptEventStream;
  late final StreamSink<Uint8List> audioStreamSink;
  StreamSubscription<Uint8List>? audioStreamSubscription;

  try {
    // Start a stream transcription.
    (response, audioStreamSink, transcriptEventStream) =
        await transcribeStreamingClient.startStreamTranscription(
      const StartStreamTranscriptionRequest(
        languageCode: LanguageCode.enUs,
        mediaSampleRateHertz: 48000,
        mediaEncoding: MediaEncoding.pcm,
      ),
    );
  } on TranscribeStreamingException catch (e) {
    print('Error starting transcription: $e');
    return;
  }

  print('Session ID: ${response.sessionId}');

  final transcriptionCompleter = Completer<void>();

  // Listen to transcript events.
  // final transcriptSubscription =
  //     transcriptEventStream.listen((TranscriptEvent event) => print(event));
  // or use a custom strategy to decode transcript events.
  transcriptEventStream
      .transform(
          const TranscriptEventStreamDecoder(PlainTextTranscriptionStrategy()))
      .listen(
    (String message) {
      print('Transcription: $message');
    },
    onError: (Object error, StackTrace stackTrace) async {
      print('Transcription error: $error');
      await audioStreamSubscription?.cancel();
      await audioStreamSink.close();
    },
    onDone: () async {
      print('Transcription done');
      await audioStreamSubscription?.cancel();
      transcriptionCompleter.complete();
    },
  );

  // Instead use a real stream of audio data from the microphone
  // in PCM signed 16-bit little-endian audio format with 48kHz sample rate.
  final audioStream = Stream<Uint8List>.periodic(
      const Duration(milliseconds: 200), (count) => Uint8List(19200)).take(25);

  // Send audio data to the audio stream sink.
  audioStreamSubscription = audioStream.listen(
    audioStreamSink.add,
    onDone: audioStreamSink.close,
  );

  await transcriptionCompleter.future;

  print('Finished');
}