Alibaba Cloud Model Studio:Praktik terbaik sintesis ucapan dengan konkurensi tinggi

Python SDK

Pool objek Qwen-Audio-TTS dan CosyVoice memerlukan DashScope Python SDK 1.25.2 atau yang lebih baru:

pip install "dashscope>=1.25.2"

Java SDK

• Qwen-Audio-TTS, CosyVoice: DashScope Java SDK 2.16.6 atau yang lebih baru.

Contoh Java juga memerlukan Apache Commons Pool 2. Konfigurasikan dependensi dengan Maven atau Gradle:

<dependencies>
    <dependency>
        <groupId>com.alibaba</groupId>
        <artifactId>dashscope-sdk-java</artifactId>
        <version>2.23.2</version>
    </dependency>
    <dependency>
        <groupId>org.apache.commons</groupId>
        <artifactId>commons-pool2</artifactId>
        <version>2.12.0</version>
    </dependency>
</dependencies>

dependencies {
    implementation 'com.alibaba:dashscope-sdk-java:2.23.2'
    implementation 'org.apache.commons:commons-pool2:2.12.0'
}

Pengaturan Java

Atur variabel lingkungan SDK sebelum proses Java dimulai. Tiga variabel pertama merupakan pengaturan SDK; variabel kapasitas berikutnya ditentukan oleh kode contoh:

Contoh Java menyetel maxIdle sama dengan maxTotal untuk mengurangi pembuatan ulang setelah objek idle dihancurkan. Tutup koneksi saat menghancurkan objek. Pada maxTotal, permintaan peminjaman baru menunggu objek yang tersedia.

Referensi konfigurasi server

Konfigurasi server Alibaba Cloud berikut merupakan referensi ketika hanya layanan sintesis yang ditunjukkan yang dijalankan. Konkurensi per server berarti tugas sintesis simultan. Sesuaikan konfigurasi dengan beban aktual.

Pengaturan pool koneksi pada tabel menjadi nilai referensi untuk ketiga variabel: DASHSCOPE_CONNECTION_POOL_SIZE, DASHSCOPE_MAXIMUM_ASYNC_REQUESTS, dan DASHSCOPE_MAXIMUM_ASYNC_REQUESTS_PER_HOST. Untuk konfigurasi referensi ini, atur ketiganya ke 2000. Atur kapasitas pool objek melalui variabel lingkungan pada contoh yang sesuai.

Python SDK

import os
import threading
from concurrent.futures import ThreadPoolExecutor
import dashscope
from dashscope.audio.tts_v2 import SpeechSynthesizerObjectPool, ResultCallback, AudioFormat

dashscope.api_key = os.environ["DASHSCOPE_API_KEY"]
dashscope.base_websocket_api_url = "wss://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api-ws/v1/inference"
pool = SpeechSynthesizerObjectPool(max_size=3)

def synthesize(index):
    done = threading.Event()
    errors = []
    with open(f"output_{index}.mp3", "wb") as audio:
        class Callback(ResultCallback):
            def on_data(self, data: bytes):
                audio.write(data)

            def on_complete(self):
                done.set()

            def on_error(self, message):
                errors.append(message)
                done.set()

        synthesizer = pool.borrow_synthesizer(
            model="qwen-audio-3.1-tts-flash",
            voice="longanhuan_v3.1",
            format=AudioFormat.MP3_24000HZ_MONO_256KBPS,
            callback=Callback(),
        )
        successful = False
        try:
            synthesizer.streaming_call("你好，欢迎使用阿里云百炼。")
            synthesizer.streaming_complete(60000)
            if not done.wait(60):
                raise TimeoutError("Synthesis timed out")
            if errors:
                raise RuntimeError(errors[0])
            successful = True
        finally:
            if successful:
                pool.return_synthesizer(synthesizer)
            else:
                synthesizer.close()

try:
    with ThreadPoolExecutor(max_workers=3) as executor:
        list(executor.map(synthesize, range(3)))
finally:
    pool.shutdown()

Java SDK

import com.alibaba.dashscope.audio.tts.SpeechSynthesisResult;
import com.alibaba.dashscope.audio.ttsv2.SpeechSynthesisAudioFormat;
import com.alibaba.dashscope.audio.ttsv2.SpeechSynthesisParam;
import com.alibaba.dashscope.audio.ttsv2.SpeechSynthesizer;
import com.alibaba.dashscope.common.ResultCallback;
import com.alibaba.dashscope.utils.Constants;
import org.apache.commons.pool2.BasePooledObjectFactory;
import org.apache.commons.pool2.PooledObject;
import org.apache.commons.pool2.impl.DefaultPooledObject;
import org.apache.commons.pool2.impl.GenericObjectPool;
import org.apache.commons.pool2.impl.GenericObjectPoolConfig;
import java.io.FileOutputStream;
import java.nio.ByteBuffer;
import java.time.Duration;
import java.util.concurrent.*;
import java.util.concurrent.atomic.AtomicReference;

public class PooledTts {
    static class Factory extends BasePooledObjectFactory<SpeechSynthesizer> {
        public SpeechSynthesizer create() { return new SpeechSynthesizer(); }
        public PooledObject<SpeechSynthesizer> wrap(SpeechSynthesizer value) {
            return new DefaultPooledObject<>(value);
        }
        public void destroyObject(PooledObject<SpeechSynthesizer> value) {
            value.getObject().getDuplexApi().close(1000, "done");
        }
    }

    static void synthesize(GenericObjectPool<SpeechSynthesizer> pool, int index) throws Exception {
        SpeechSynthesizer synthesizer = pool.borrowObject();
        boolean successful = false;
        AtomicReference<Exception> error = new AtomicReference<>();
        try (FileOutputStream audio = new FileOutputStream("output_" + index + ".mp3")) {
            SpeechSynthesisParam param = SpeechSynthesisParam.builder()
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .model("qwen-audio-3.1-tts-flash")
                .voice("longanhuan_v3.1")
                .format(SpeechSynthesisAudioFormat.MP3_24000HZ_MONO_256KBPS)
                .build();
            synthesizer.updateParamAndCallback(param, new ResultCallback<SpeechSynthesisResult>() {
                public void onEvent(SpeechSynthesisResult result) {
                    ByteBuffer buffer = result.getAudioFrame();
                    if (buffer != null) {
                        byte[] bytes = new byte[buffer.remaining()];
                        buffer.get(bytes);
                        try { audio.write(bytes); }
                        catch (Exception e) { error.set(e); }
                    }
                }
                public void onComplete() {}
                public void onError(Exception e) { error.set(e); }
            });
            synthesizer.streamingCall("你好，欢迎使用阿里云百炼。");
            synthesizer.streamingComplete(60000);
            if (error.get() != null) { throw error.get(); }
            successful = true;
        } finally {
            if (successful) { pool.returnObject(synthesizer); }
            else { pool.invalidateObject(synthesizer); }
        }
    }

    public static void main(String[] args) throws Exception {
        Constants.baseWebsocketApiUrl = "wss://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api-ws/v1/inference";
        GenericObjectPoolConfig<SpeechSynthesizer> config = new GenericObjectPoolConfig<>();
        int poolSize = Integer.parseInt(System.getenv().getOrDefault("COSYVOICE_OBJECTPOOL_SIZE", "500"));
        config.setMaxTotal(poolSize);
        config.setMaxIdle(poolSize);
        config.setMaxWait(Duration.ofSeconds(60));
        ExecutorService executor = Executors.newFixedThreadPool(3);
        try (GenericObjectPool<SpeechSynthesizer> pool = new GenericObjectPool<>(new Factory(), config)) {
            java.util.List<Future<?>> futures = new java.util.ArrayList<>();
            for (int i = 0; i < 3; i++) {
                final int index = i;
                futures.add(executor.submit(() -> {
                    try { synthesize(pool, index); }
                    catch (Exception e) { throw new RuntimeException(e); }
                }));
            }
            for (Future<?> future : futures) { future.get(); }
        } finally {
            executor.shutdownNow();
        }
    }
}