Flutter语音识别插件deepgram_speech_to_text的使用

Features

PRs are welcome for all work-in-progress 🚧 features

Getting Started

All you need is a Deepgram API key. You can get a free one by signing up on Deepgram.

Usage

Initialize Client

First, create the client with optional parameters:

String apiKey = 'your_api_key';

// Optional parameters can be passed in client's baseQueryParams or in every method's queryParams
Deepgram deepgram = Deepgram(apiKey, baseQueryParams: {
  'model': 'nova-2-general',
  'detect_language': true,
  'filler_words': false,
  'punctuation': true,
  // More options here : https://developers.deepgram.com/reference/listen-file
});

Speech to Text

Call the methods under the proper listen subclass:

// From file
DeepgramListenResult resFile = await deepgram.listen.file(File('audio.wav'));

// From URL
DeepgramListenResult resUrl = await deepgram.listen.url('https://somewhere/audio.wav');

// From bytes
DeepgramListenResult resBytes = await deepgram.listen.bytes(List.from([1, 2, 3, 4, 5]));

// Streaming from audio stream (e.g., microphone)
Stream<List<int>> micStream = await AudioRecorder().startStream(RecordConfig(
  encoder: AudioEncoder.pcm16bits,
  sampleRate: 16000,
  numChannels: 1,
));

final streamParams = {
  'language': 'en', // Must specify encoding and sample_rate according to the audio stream
  'encoding': 'linear16',
  'sample_rate': 16000,
};

Deepgram deepgramStreaming = Deepgram(apiKey, baseQueryParams: streamParams);

// Automatically managed stream
Stream<DeepgramListenResult> streamAuto = deepgramStreaming.listen.live(micStream);

// Manually managed stream
DeepgramLiveListener listener = deepgramStreaming.listen.liveListener(micStream);
listener.stream.listen((res) {
    print(res.transcript);
});
listener.start();
listener.pause(); 
listener.resume();
listener.close();

Text to Speech

Call the methods under the proper speak subclass:

// From text
DeepgramSpeakResult resText = await deepgram.speak.text('Hello world');

// Streaming from text stream
Stream<String> textStream = Stream.fromIterable(['Hello', 'World']);
Stream<DeepgramSpeakResult> streamTTS = deepgram.speak.live(textStream);

// Manually managed stream
DeepgramLiveSpeaker speaker = deepgram.speak.liveSpeaker(textStream);
speaker.stream.listen((res) {
    print(res);
    // Use the audio data if needed
});

speaker.start();
speaker.flush();
speaker.clear();
speaker.close();

Debugging Common Errors

• Ensure your API key is valid and has enough credits.
• Check parameters; some models do not support certain parameters (e.g., whisper model with live streaming).
• For empty transcripts/metadata only, verify that the encoding and sample rate match the audio stream.

Additional Information & Support

This package was created to fulfill specific needs since there were no Dart SDKs available for Deepgram. Contributions and feature requests are welcome on GitHub. If you find this project useful, consider supporting it here.

Complete Demo Example

Below is a complete demo example showcasing how to use the deepgram_speech_to_text plugin within a Flutter application:

import 'dart:io';
import 'package:flutter/material.dart';
import 'package:deepgram_speech_to_text/deepgram_speech_to_text.dart';
import 'package:record/record.dart';

void main() => runApp(MyApp());

class MyApp extends StatelessWidget {
  @override
  Widget build(BuildContext context) {
    return MaterialApp(
      home: Scaffold(
        appBar: AppBar(title: Text('Deepgram Speech To Text Demo')),
        body: Center(child: SpeechToTextDemo()),
      ),
    );
  }
}

class SpeechToTextDemo extends StatefulWidget {
  @override
  _SpeechToTextDemoState createState() => _SpeechToTextDemoState();
}

class _SpeechToTextDemoState extends State<SpeechToTextDemo> {
  final String apiKey = "YOUR_API_KEY";
  final TextEditingController _textController = TextEditingController();
  bool _isRecording = false;
  bool _isTranscribing = false;
  String _transcription = "";

  late Deepgram deepgram;

  @override
  void initState() {
    super.initState();
    deepgram = Deepgram(apiKey, baseQueryParams: {
      'model': 'nova-2-general',
      'detect_language': true,
      'filler_words': false,
      'punctuation': true,
    });
  }

  Future<void> startRecording() async {
    final record = Record();
    if (await record.hasPermission()) {
      await record.start(
        path: null, // Use null for byte array output
        encoder: AudioEncoder.PCM_16, // Default is AAC
        bitRate: 128000,
        samplingRate: 16000,
      );
      setState(() {
        _isRecording = true;
      });

      final bytes = await record.stop();
      final audioStream = Stream.value(bytes);

      final streamParams = {
        'language': 'en',
        'encoding': 'linear16',
        'sample_rate': 16000,
      };

      final deepgramStreaming = Deepgram(apiKey, baseQueryParams: streamParams);
      final listener = deepgramStreaming.listen.liveListener(audioStream);

      listener.stream.listen((res) {
        setState(() {
          _transcription = res.transcript ?? '';
        });
      });

      listener.start();
      setState(() {
        _isTranscribing = true;
      });

      await Future.delayed(Duration(seconds: 5)); // Simulate transcription time
      listener.close();
      setState(() {
        _isTranscribing = false;
        _isRecording = false;
      });
    }
  }

  @override
  Widget build(BuildContext context) {
    return Column(
      mainAxisAlignment: MainAxisAlignment.center,
      children: <Widget>[
        ElevatedButton(
          onPressed: !_isRecording && !_isTranscribing ? startRecording : null,
          child: Text(_isRecording ? 'Recording...' : 'Start Recording'),
        ),
        SizedBox(height: 20),
        Text(_transcription),
      ],
    );
  }
}

In this demo, we’ve integrated the deepgram_speech_to_text plugin into a simple Flutter app that records audio from the device’s microphone and transcribes it using Deepgram’s speech-to-text service. The transcription result is then displayed on the screen. Adjust the API key and any other configurations as necessary for your environment.