xf 维语转语音

src/main/java/com/newyetai/ai/resume/controller/AiResumeController.java

@@ -3,6 +3,7 @@ package com.newyetai.ai.resume.controller;
 import com.newyetai.ai.resume.dto.AiResumeReqDTO;
 import com.newyetai.ai.resume.service.AiResumeService;
 import com.newyetai.common.AlitaResult;
+import com.newyetai.sdk.xf.dto.TtsReqDTO;
 import com.newyetai.sdk.xf.util.XfUtils;
 
 import org.springframework.beans.factory.annotation.Autowired;
@@ -48,8 +49,8 @@ public class AiResumeController {
   
   @PostMapping("/tts")
   @ApiOperation("ai简历-语音合成")
-  public AlitaResult<String> tts(@RequestParam("text") String text) {
+  public AlitaResult<String> tts(@RequestBody @Validated TtsReqDTO ttsReq) {
-    return AlitaResult.success(XfUtils.tts(text));
+    return AlitaResult.success(XfUtils.tts(ttsReq));
   }
 
   @PostMapping("/analyseResume")

src/main/java/com/newyetai/sdk/xf/dto/RtasrReqDTO.java

@@ -16,6 +16,10 @@ public class RtasrReqDTO implements Serializable {
 
   @NotBlank(message = "语音文件不能为空")
   private String fileCode;
+  /**
+   * 中文：cn
+   * 维语：cn_uyghur
+   */
   @Default
   private String from = "cn_uyghur";
 

src/main/java/com/newyetai/sdk/xf/dto/TtsReqDTO.java

@@ -0,0 +1,40 @@
+package com.newyetai.sdk.xf.dto;
+
+import com.newyetai.sdk.xf.util.WebTtsWS;
+import java.io.Serializable;
+import javax.validation.constraints.NotBlank;
+import lombok.AllArgsConstructor;
+import lombok.Builder;
+import lombok.Builder.Default;
+import lombok.Data;
+import lombok.NoArgsConstructor;
+
+@Data
+@AllArgsConstructor
+@NoArgsConstructor
+@Builder
+public class TtsReqDTO implements Serializable {
+
+  @NotBlank(message = "合成文本不能为空")
+  private String text;
+
+  /**
+   * 发音人
+   * x4_yezi（中文）/x2_ugsp_dilare（维语）
+   */
+  @Default
+  private String vcn = "xiaoyan";
+
+  /**
+   * 文本编码格式
+   * GB2312
+   * GBK
+   * BIG5
+   * UNICODE(小语种)
+   * GB18030
+   * UTF8（小语种推荐使用UTF8编码）
+   * */
+  @Default
+  private String tte = WebTtsWS.TTL_UTF8;
+
+}

src/main/java/com/newyetai/sdk/xf/util/WebTtsWS.java

@@ -5,6 +5,7 @@ import java.io.FileOutputStream;
 import java.io.IOException;
 import java.io.InputStream;
 import java.io.OutputStream;
+import java.io.UnsupportedEncodingException;
 import java.net.URL;
 import java.nio.charset.Charset;
 import java.nio.charset.StandardCharsets;
@@ -22,6 +23,7 @@ import org.apache.tomcat.util.http.fileupload.ByteArrayOutputStream;
 import com.google.gson.Gson;
 import com.newyetai.common.AlitaResult;
 import com.newyetai.sdk.xf.config.XfProperties;
+import com.newyetai.sdk.xf.dto.TtsReqDTO;
 
 import lombok.extern.slf4j.Slf4j;
 import okhttp3.HttpUrl;
@@ -44,6 +46,9 @@ public class WebTtsWS extends WebSocketListener {
   public static final int StatusFirstFrame = 0;
   public static final int StatusContinueFrame = 1;
   public static final int StatusLastFrame = 2;
+  public static final String TTL_UTF8 = "UTF8";
+  public static final String TTL_UNICODE = "UNICODE";
+
   public static final Gson json = new Gson();
   // mp3格式
   private String encoding = "lame";
@@ -52,14 +57,16 @@ public class WebTtsWS extends WebSocketListener {
   private String text;
   private String appId;
   OutputStream outputStream;
-  // 合成文本编码格式
+  // 小语种必须使用UNICODE编码作为值
-  public static final String TTE = "UTF8"; // 小语种必须使用UNICODE编码作为值
+  private String tte;
   // 发音人参数。到控制台-我的应用-语音合成-添加试用或购买发音人，添加后即显示该发音人参数值，若试用未添加的发音人会报错11200
-  public static final String VCN = "xiaoyan";
+  private String vcn;
 
-  WebTtsWS(String text, String appId, OutputStream outputStream){
+  WebTtsWS(String text, String appId, String vcn, String tte, OutputStream outputStream){
     this.text = text;
     this.appId = appId;
+    this.vcn = vcn;
+    this.tte = tte;
     this.outputStream = outputStream;
   }
 
@@ -73,6 +80,16 @@ public class WebTtsWS extends WebSocketListener {
     super.onOpen(webSocket, response);
     log.info("ws建立连接成功...");
     String requestJson;//请求参数json串
+    String encodedText ;
+    if (WebTtsWS.TTL_UNICODE.equals(tte)) {
+      try {
+        encodedText = Base64.getEncoder().encodeToString(text.getBytes("UTF-16LE"));
+      } catch (UnsupportedEncodingException e) {
+        throw new RuntimeException(e);
+      }
+    } else {
+      encodedText = Base64.getEncoder().encodeToString(text.getBytes(StandardCharsets.UTF_8));
+    }
     requestJson = "{\n" +
         "  \"common\": {\n" +
         "    \"app_id\": \"" + appId + "\"\n" +
@@ -80,21 +97,19 @@ public class WebTtsWS extends WebSocketListener {
         "  \"business\": {\n" +
         "    \"aue\": \"lame\",\n" +
         "    \"sfl\": 1,\n" +
-        "    \"tte\": \"" + TTE + "\",\n" +
+        "    \"tte\": \"" + tte + "\",\n" +
         "    \"ent\": \"intp65\",\n" +
-        "    \"vcn\": \"" + VCN + "\",\n" +
+        "    \"vcn\": \"" + vcn + "\",\n" +
         "    \"pitch\": 50,\n" +
         "    \"speed\": 50\n" +
         "  },\n" +
         "  \"data\": {\n" +
         "    \"status\": 2,\n" +
         "    \"text\": \""
-        + Base64.getEncoder().encodeToString(text.getBytes(StandardCharsets.UTF_8)) + "\"\n" +
+        + encodedText + "\"\n" +
-        //"    \"text\": \"" + Base64.getEncoder().encodeToString(TEXT.getBytes("UTF-16LE")) + "\"\n" +
         "  }\n" +
         "}";
     webSocket.send(requestJson);
-
   }
 
   @Override
@@ -145,7 +160,7 @@ public class WebTtsWS extends WebSocketListener {
     }
   }
 
-  public static AlitaResult<Void> execute(String text, XfProperties xfProperties, OutputStream outputStream) {
+  public static AlitaResult<Void> execute(TtsReqDTO ttsReq, XfProperties xfProperties, OutputStream outputStream) {
     try {
       // 构建鉴权url
       String authUrl = getAuthUrl(xfProperties.getTtsHostUrl(), xfProperties.getApiKey(),
@@ -154,7 +169,7 @@ public class WebTtsWS extends WebSocketListener {
       //将url中的 schema http://和https://分别替换为ws:// 和 wss://
       String url = authUrl.toString().replace("http://", "ws://").replace("https://", "wss://");
       Request request = new Request.Builder().url(url).build();
-      WebTtsWS webIATWS = new WebTtsWS(text, xfProperties.getAppId(), outputStream);
+      WebTtsWS webIATWS = new WebTtsWS(ttsReq.getText(), xfProperties.getAppId(), ttsReq.getVcn(), ttsReq.getTte(), outputStream);
       WebSocket webSocket = client.newWebSocket(request, webIATWS);
       while (webIATWS.result == null) {
         // 未拿到结果等待
@@ -165,7 +180,6 @@ public class WebTtsWS extends WebSocketListener {
       log.info("err ", e);
       return AlitaResult.error("语音合成失败：{}", e.getMessage());
     }
-
   }
 
   public static String getAuthUrl(String hostUrl, String apiKey, String apiSecret)

src/main/java/com/newyetai/sdk/xf/util/XfUtils.java

@@ -22,6 +22,7 @@ import com.newyetai.component.file.util.FileUtils;
 import com.newyetai.config.service.ConfigService;
 import com.newyetai.sdk.xf.config.XfProperties;
 import com.newyetai.sdk.xf.dto.InterviewContentDTO;
+import com.newyetai.sdk.xf.dto.TtsReqDTO;
 import com.newyetai.sdk.xf.model.ChatReq;
 import com.newyetai.sdk.xf.model.MessageDTO;
 import com.newyetai.sdk.xf.service.XfSparkProvider;
@@ -123,15 +124,14 @@ public class XfUtils {
   }
 
 
-
   /**
    * 语音合成
    * @return
    * @author sr on 2025-01-09
    */
-  public static String tts(String text) {
+  public static String tts(TtsReqDTO ttsReq) {
     try(ByteArrayOutputStream outputStream = new ByteArrayOutputStream()){
-      AlitaResult<Void> result = WebTtsWS.execute(text, xfProperties, outputStream);
+      AlitaResult<Void> result = WebTtsWS.execute(ttsReq, xfProperties, outputStream);
       Assert.isTrue(result.getSuccess(), result.getMsg());
       FileRecord fileRecord = fileService.upload("语音合成.mp3", outputStream.toByteArray());
       return fileRecord.getFileCode();