腾讯云录音文件快速识别实战教程

- 视频字幕自动生成
- 呼叫中心/质检实时转写
- 会议记录转写

<dependency>
    <groupId>com.tencentcloudapi</groupId>
    <artifactId>tencentcloud-speech-sdk-java</artifactId>
    <version>1.0.52</version>
</dependency>

@Component
@Slf4j
public class FlashRecognizerService {

    private static final String SECRET_KEY = "你的SecretKey";
    private static final String SECRET_ID = "你的SecretId";
    private static final String APP_ID = "你的AppId";

    private final Credential credential = Credential.builder()
        .secretId(SECRET_ID)
        .secretKey(SECRET_KEY)
        .build();

    public String recognizer(File file) {
        log.info("开始解析音频文件：{}", file);
        FlashRecognizer recognizer = SpeechClient.newFlashRecognizer(APP_ID, credential);

        byte[] data = ByteUtils.inputStream2ByteArray(file);

        // 初始化识别请求
        FlashRecognitionRequest recognitionRequest = FlashRecognitionRequest.initialize();
        recognitionRequest.setEngineType("16k_zh"); // 中文普通话
        recognitionRequest.setFirstChannelOnly(1);  // 仅识别第一声道
        recognitionRequest.setVoiceFormat(getFileExtension(file));
        recognitionRequest.setSpeakerDiarization(0); // 不做说话人分离
        recognitionRequest.setFilterDirty(0); // 不过滤脏词
        recognitionRequest.setFilterModal(0); // 不过滤语气词
        recognitionRequest.setFilterPunc(0);  // 保留标点
        recognitionRequest.setConvertNumMode(1); // 开启数字转换
        recognitionRequest.setWordInfo(1); // 返回词级信息

        // 同步识别
        FlashRecognitionResponse response = recognizer.recognize(recognitionRequest, data);
        log.info("音频文件解析完成：{} {}", file, JSONObject.toJSONString(response));

        return response.getFlashResult().get(0).getText();
    }

    /**
     * 获取文件扩展名
     */
    public static String getFileExtension(File file) {
        String name = file.getName();
        int lastIndex = name.lastIndexOf(".");
        if (lastIndex == -1 || lastIndex == name.length() - 1) {
            return "";
        }
        return name.substring(lastIndex + 1);
    }
}