Alibaba Cloud Model Studio:Real-time speech recognition - Fun-ASR/Paraformer

import com.alibaba.dashscope.audio.asr.recognition.Recognition;
import com.alibaba.dashscope.audio.asr.recognition.RecognitionParam;
import com.alibaba.dashscope.audio.asr.recognition.RecognitionResult;
import com.alibaba.dashscope.common.ResultCallback;
import com.alibaba.dashscope.utils.Constants;

import javax.sound.sampled.AudioFormat;
import javax.sound.sampled.AudioSystem;
import javax.sound.sampled.TargetDataLine;

import java.nio.ByteBuffer;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;

public class Main {
    public static void main(String[] args) throws InterruptedException {
        // The following URL is for the Singapore region. If you use models in the Beijing region, replace the URL with: wss://dashscope.aliyuncs.com/api-ws/v1/inference
        Constants.baseWebsocketApiUrl = "wss://dashscope-intl.aliyuncs.com/api-ws/v1/inference";
        ExecutorService executorService = Executors.newSingleThreadExecutor();
        executorService.submit(new RealtimeRecognitionTask());
        executorService.shutdown();
        executorService.awaitTermination(1, TimeUnit.MINUTES);
        System.exit(0);
    }
}

class RealtimeRecognitionTask implements Runnable {
    @Override
    public void run() {
        RecognitionParam param = RecognitionParam.builder()
                .model("fun-asr-realtime")
                // API keys differ between Singapore and Beijing regions. Get an API key: https://www.alibabacloud.com/help/zh/model-studio/get-api-key
                // If you have not configured an environment variable, replace the following line with your Model Studio API key: .apiKey("sk-xxx")
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .format("pcm")
                .sampleRate(16000)
                .build();
        Recognition recognizer = new Recognition();

        ResultCallback<RecognitionResult> callback = new ResultCallback<RecognitionResult>() {
            @Override
            public void onEvent(RecognitionResult result) {
                if (result.isSentenceEnd()) {
                    System.out.println("Final Result: " + result.getSentence().getText());
                } else {
                    System.out.println("Intermediate Result: " + result.getSentence().getText());
                }
            }

            @Override
            public void onComplete() {
                System.out.println("Recognition complete");
            }

            @Override
            public void onError(Exception e) {
                System.out.println("RecognitionCallback error: " + e.getMessage());
            }
        };
        try {
            recognizer.call(param, callback);
            // Create audio format
            AudioFormat audioFormat = new AudioFormat(16000, 16, 1, true, false);
            // Match the default recording device based on the format
            TargetDataLine targetDataLine =
                    AudioSystem.getTargetDataLine(audioFormat);
            targetDataLine.open(audioFormat);
            // Start recording
            targetDataLine.start();
            ByteBuffer buffer = ByteBuffer.allocate(1024);
            long start = System.currentTimeMillis();
            // Record for 50s and perform real-time transcription
            while (System.currentTimeMillis() - start < 50000) {
                int read = targetDataLine.read(buffer.array(), 0, buffer.capacity());
                if (read > 0) {
                    buffer.limit(read);
                    // Send recorded audio data to the streaming recognition service
                    recognizer.sendAudioFrame(buffer);
                    buffer = ByteBuffer.allocate(1024);
                    // Sleep briefly to prevent high CPU usage due to limited recording speed
                    Thread.sleep(20);
                }
            }
            recognizer.stop();
        } catch (Exception e) {
            e.printStackTrace();
        } finally {
            // Close the Websocket connection after the task ends
            recognizer.getDuplexApi().close(1000, "bye");
        }

        System.out.println(
                "[Metric] requestId: "
                        + recognizer.getLastRequestId()
                        + ", first package delay ms: "
                        + recognizer.getFirstPackageDelay()
                        + ", last package delay ms: "
                        + recognizer.getLastPackageDelay());
    }
}

import os
import signal  # for keyboard events handling (press "Ctrl+C" to terminate recording)
import sys

import dashscope
import pyaudio
from dashscope.audio.asr import *

mic = None
stream = None

# Set recording parameters
sample_rate = 16000  # sampling rate (Hz)
channels = 1  # mono channel
dtype = 'int16'  # data type
format_pcm = 'pcm'  # the format of the audio data
block_size = 3200  # number of frames per buffer


# Real-time speech recognition callback
class Callback(RecognitionCallback):
    def on_open(self) -> None:
        global mic
        global stream
        print('RecognitionCallback open.')
        mic = pyaudio.PyAudio()
        stream = mic.open(format=pyaudio.paInt16,
                          channels=1,
                          rate=16000,
                          input=True)

    def on_close(self) -> None:
        global mic
        global stream
        print('RecognitionCallback close.')
        stream.stop_stream()
        stream.close()
        mic.terminate()
        stream = None
        mic = None

    def on_complete(self) -> None:
        print('RecognitionCallback completed.')  # recognition completed

    def on_error(self, message) -> None:
        print('RecognitionCallback task_id: ', message.request_id)
        print('RecognitionCallback error: ', message.message)
        # Stop and close the audio stream if it is running
        if 'stream' in globals() and stream.active:
            stream.stop()
            stream.close()
        # Forcefully exit the program
        sys.exit(1)

    def on_event(self, result: RecognitionResult) -> None:
        sentence = result.get_sentence()
        if 'text' in sentence:
            print('RecognitionCallback text: ', sentence['text'])
            if RecognitionResult.is_sentence_end(sentence):
                print(
                    'RecognitionCallback sentence end, request_id:%s, usage:%s'
                    % (result.get_request_id(), result.get_usage(sentence)))


def signal_handler(sig, frame):
    print('Ctrl+C pressed, stop recognition ...')
    # Stop recognition
    recognition.stop()
    print('Recognition stopped.')
    print(
        '[Metric] requestId: {}, first package delay ms: {}, last package delay ms: {}'
        .format(
            recognition.get_last_request_id(),
            recognition.get_first_package_delay(),
            recognition.get_last_package_delay(),
        ))
    # Forcefully exit the program
    sys.exit(0)


# main function
if __name__ == '__main__':
    # API keys differ between the Singapore and Beijing regions. Get an API key: https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    # If you have not set an environment variable, replace the next line with your Model Studio API key: dashscope.api_key = "sk-xxx"
    dashscope.api_key = os.environ.get('DASHSCOPE_API_KEY')

    # The following URL is for the Singapore region. To use models in the Beijing region, replace it with: wss://dashscope.aliyuncs.com/api-ws/v1/inference
    dashscope.base_websocket_api_url='wss://dashscope-intl.aliyuncs.com/api-ws/v1/inference'

    # Create the recognition callback
    callback = Callback()

    # Call recognition service by async mode, you can customize the recognition parameters, like model, format,
    # sample_rate
    recognition = Recognition(
        model='fun-asr-realtime',
        format=format_pcm,
        # 'pcm'、'wav'、'opus'、'speex'、'aac'、'amr', you can check the supported formats in the document
        sample_rate=sample_rate,
        # support 8000, 16000
        semantic_punctuation_enabled=False,
        callback=callback)

    # Start recognition
    recognition.start()

    signal.signal(signal.SIGINT, signal_handler)
    print("Press 'Ctrl+C' to stop recording and recognition...")
    # Create a keyboard listener until "Ctrl+C" is pressed

    while True:
        if stream:
            data = stream.read(3200, exception_on_overflow=False)
            recognition.send_audio_frame(data)
        else:
            break

    recognition.stop()

Audio used in the example: asr_example.wav.

import com.alibaba.dashscope.api.GeneralApi;
import com.alibaba.dashscope.audio.asr.recognition.Recognition;
import com.alibaba.dashscope.audio.asr.recognition.RecognitionParam;
import com.alibaba.dashscope.audio.asr.recognition.RecognitionResult;
import com.alibaba.dashscope.base.HalfDuplexParamBase;
import com.alibaba.dashscope.common.GeneralListParam;
import com.alibaba.dashscope.common.ResultCallback;
import com.alibaba.dashscope.protocol.GeneralServiceOption;
import com.alibaba.dashscope.protocol.HttpMethod;
import com.alibaba.dashscope.protocol.Protocol;
import com.alibaba.dashscope.protocol.StreamingMode;
import com.alibaba.dashscope.utils.Constants;

import java.io.FileInputStream;
import java.nio.ByteBuffer;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.time.LocalDateTime;
import java.time.format.DateTimeFormatter;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;

class TimeUtils {
    private static final DateTimeFormatter formatter =
            DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss.SSS");

    public static String getTimestamp() {
        return LocalDateTime.now().format(formatter);
    }
}

public class Main {
    public static void main(String[] args) throws InterruptedException {
        // The following URL is for the Singapore region. If you use models in the Beijing region, replace the URL with: wss://dashscope.aliyuncs.com/api-ws/v1/inference
        Constants.baseWebsocketApiUrl = "wss://dashscope-intl.aliyuncs.com/api-ws/v1/inference";
        // In actual applications, run this method only once at program startup.
        warmUp();

        ExecutorService executorService = Executors.newSingleThreadExecutor();
        executorService.submit(new RealtimeRecognitionTask(Paths.get(System.getProperty("user.dir"), "asr_example.wav")));
        executorService.shutdown();

        // wait for all tasks to complete
        executorService.awaitTermination(1, TimeUnit.MINUTES);
        System.exit(0);
    }

    public static void warmUp() {
        try {
            // Lightweight GET request to establish connection
            GeneralServiceOption warmupOption = GeneralServiceOption.builder()
                    .protocol(Protocol.HTTP)
                    .httpMethod(HttpMethod.GET)
                    .streamingMode(StreamingMode.OUT)
                    .path("assistants")
                    .build();

            warmupOption.setBaseHttpUrl(Constants.baseHttpApiUrl);
            GeneralApi<HalfDuplexParamBase> api = new GeneralApi<>();
            api.get(GeneralListParam.builder().limit(1L).build(), warmupOption);
        } catch (Exception e) {
            // Reset flag to allow retry if pre-warming failed
        }
    }
}

class RealtimeRecognitionTask implements Runnable {
    private Path filepath;

    public RealtimeRecognitionTask(Path filepath) {
        this.filepath = filepath;
    }

    @Override
    public void run() {
        RecognitionParam param = RecognitionParam.builder()
                .model("fun-asr-realtime")
                // API keys differ between Singapore and Beijing regions. Get an API key: https://www.alibabacloud.com/help/zh/model-studio/get-api-key
                // If you have not configured an environment variable, replace the following line with your Model Studio API key: .apiKey("sk-xxx")
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .format("wav")
                .sampleRate(16000)
                .build();
        Recognition recognizer = new Recognition();

        String threadName = Thread.currentThread().getName();

        ResultCallback<RecognitionResult> callback = new ResultCallback<RecognitionResult>() {
            @Override
            public void onEvent(RecognitionResult message) {
                if (message.isSentenceEnd()) {

                    System.out.println(TimeUtils.getTimestamp()+" "+
                            "[process " + threadName + "] Final Result:" + message.getSentence().getText());
                } else {
                    System.out.println(TimeUtils.getTimestamp()+" "+
                            "[process " + threadName + "] Intermediate Result: " + message.getSentence().getText());
                }
            }

            @Override
            public void onComplete() {
                System.out.println(TimeUtils.getTimestamp()+" "+"[" + threadName + "] Recognition complete");
            }

            @Override
            public void onError(Exception e) {
                System.out.println(TimeUtils.getTimestamp()+" "+
                        "[" + threadName + "] RecognitionCallback error: " + e.getMessage());
            }
        };

        try {
            recognizer.call(param, callback);
            // Please replace the path with your audio file path
            System.out.println(TimeUtils.getTimestamp()+" "+"[" + threadName + "] Input file_path is: " + this.filepath);
            // Read file and send audio by chunks
            FileInputStream fis = new FileInputStream(this.filepath.toFile());
            byte[] allData = new byte[fis.available()];
            int ret = fis.read(allData);
            fis.close();

            int sendFrameLength = 3200;
            for (int i = 0; i * sendFrameLength < allData.length; i ++) {
                int start = i * sendFrameLength;
                int end = Math.min(start + sendFrameLength, allData.length);
                ByteBuffer byteBuffer = ByteBuffer.wrap(allData, start, end - start);
                recognizer.sendAudioFrame(byteBuffer);
                Thread.sleep(100);
            }

            System.out.println(TimeUtils.getTimestamp()+" "+LocalDateTime.now());
            recognizer.stop();
        } catch (Exception e) {
            e.printStackTrace();
        } finally {
            // Close the Websocket connection after the task ends
            recognizer.getDuplexApi().close(1000, "bye");
        }

        System.out.println(
                "["
                        + threadName
                        + "][Metric] requestId: "
                        + recognizer.getLastRequestId()
                        + ", first package delay ms: "
                        + recognizer.getFirstPackageDelay()
                        + ", last package delay ms: "
                        + recognizer.getLastPackageDelay());
    }
}

The audio file used in this example is asr_example.wav.

import os
import time
import dashscope
from dashscope.audio.asr import *

# API keys differ between the Singapore and Beijing regions. Get an API key: https://www.alibabacloud.com/help/zh/model-studio/get-api-key
# If you have not set an environment variable, replace the next line with your Model Studio API key: dashscope.api_key = "sk-xxx"
dashscope.api_key = os.environ.get('DASHSCOPE_API_KEY')

# The following URL is for the Singapore region. To use models in the Beijing region, replace it with: wss://dashscope.aliyuncs.com/api-ws/v1/inference
dashscope.base_websocket_api_url='wss://dashscope-intl.aliyuncs.com/api-ws/v1/inference'

from datetime import datetime


def get_timestamp():
    now = datetime.now()
    formatted_timestamp = now.strftime("[%Y-%m-%d %H:%M:%S.%f]")
    return formatted_timestamp


class Callback(RecognitionCallback):
    def on_complete(self) -> None:
        print(get_timestamp() + ' Recognition completed')  # recognition complete

    def on_error(self, result: RecognitionResult) -> None:
        print('Recognition task_id: ', result.request_id)
        print('Recognition error: ', result.message)
        exit(0)

    def on_event(self, result: RecognitionResult) -> None:
        sentence = result.get_sentence()
        if 'text' in sentence:
            print(get_timestamp() + ' RecognitionCallback text: ', sentence['text'])
        if RecognitionResult.is_sentence_end(sentence):
            print(get_timestamp() +
                  'RecognitionCallback sentence end, request_id:%s, usage:%s'
                  % (result.get_request_id(), result.get_usage(sentence)))


callback = Callback()

recognition = Recognition(model='fun-asr-realtime',
                          format='wav',
                          sample_rate=16000,
                          callback=callback)

try:
    audio_data: bytes = None
    f = open("asr_example.wav", 'rb')
    if os.path.getsize("asr_example.wav"):
        # Read all file data into buffer at once
        file_buffer = f.read()
        f.close()
        print("Start Recognition")
        recognition.start()

        # Send data from buffer in chunks of 3200 bytes
        buffer_size = len(file_buffer)
        offset = 0
        chunk_size = 3200

        while offset < buffer_size:
            # Calculate size of current chunk
            remaining_bytes = buffer_size - offset
            current_chunk_size = min(chunk_size, remaining_bytes)

            # Extract current chunk from buffer
            audio_data = file_buffer[offset:offset + current_chunk_size]

            # Send audio frame
            recognition.send_audio_frame(audio_data)
            # Update offset
            offset += current_chunk_size

            # Add delay to simulate real-time transmission
            time.sleep(0.1)

        recognition.stop()
    else:
        raise Exception(
            'The supplied file was empty (zero bytes long)')
except Exception as e:
    raise e

print(
    '[Metric] requestId: {}, first package delay ms: {}, last package delay ms: {}'
    .format(
        recognition.get_last_request_id(),
        recognition.get_first_package_delay(),
        recognition.get_last_package_delay(),
    ))