TTS 返回语音优化

vetti-admin/src/main/java/com/vetti/socket/ChatWebSocketHandler.java

@@ -144,8 +144,10 @@ public class ChatWebSocketHandler {
                             log.info("3、开始进行AI回答时间:{}",System.currentTimeMillis()/1000);
                             //持续返回数据流给客户端
                             try {
+                                String resultOutPathUrl = RuoYiConfig.getProfile() + VOICE_STORAGE_RESULT_DIR + "110_"+resultFileName;
+                                handleVoice(resultPathUrl,resultOutPathUrl);
                                 //文件转换成文件流
-                                ByteBuffer outByteBuffer = convertFileToByteBuffer(resultPathUrl);
+                                ByteBuffer outByteBuffer = convertFileToByteBuffer(resultOutPathUrl);
                                 //发送文件流数据
                                 session.getBasicRemote().sendBinary(outByteBuffer);
                                 // 发送响应确认
@@ -404,6 +406,50 @@ public class ChatWebSocketHandler {
         }
     }
 
+    private void handleVoice(String inputPath,String outputPath){
+        double trimMs = 270; // 要去掉的尾部时长（毫秒）
+
+        try {
+            // 1. 解析音频格式和总长度
+            AudioInputStream audioIn = AudioSystem.getAudioInputStream(new File(inputPath));
+            AudioFormat format = audioIn.getFormat();
+            long totalBytes = audioIn.getFrameLength() * format.getFrameSize(); // 总字节数
+
+            // 2. 计算300毫秒对应的字节数
+            float sampleRate = format.getSampleRate(); // 采样率（Hz）
+            int frameSize = format.getFrameSize(); // 每帧字节数（位深/8 * 声道数）
+            double trimSeconds = trimMs / 1000.0; // 转换为秒
+            long trimBytes = (long) (sampleRate * trimSeconds * frameSize); // 要去掉的字节数
+
+            // 3. 计算需要保留的字节数（避免负数）
+            long keepBytes = Math.max(0, totalBytes - trimBytes);
+            if (keepBytes == 0) {
+                System.out.println("音频长度小于300毫秒，无法截断");
+                return;
+            }
+
+            // 4. 读取并保留前半部分（去掉最后300毫秒）
+            try (InputStream in = new FileInputStream(inputPath);
+                 OutputStream out = new FileOutputStream(outputPath)) {
+
+                byte[] buffer = new byte[4096];
+                long totalRead = 0;
+                int bytesRead;
+
+                while (totalRead < keepBytes && (bytesRead = in.read(buffer)) != -1) {
+                    long remaining = keepBytes - totalRead;
+                    int writeBytes = (remaining < bytesRead) ? (int) remaining : bytesRead;
+                    out.write(buffer, 0, writeBytes);
+                    totalRead += writeBytes;
+                }
+
+                System.out.println("处理完成，去掉了最后" + trimMs + "毫秒，保留了" + totalRead + "字节");
+            }
+
+        } catch (UnsupportedAudioFileException | IOException e) {
+            e.printStackTrace();
+        }
+    }
 
 }
 

vetti-admin/src/main/java/com/vetti/web/controller/ai/AiCommonController.java

@@ -38,7 +38,7 @@ public class AiCommonController extends BaseController
     @GetMapping("/handleTextToVice")
     public AjaxResult handleTextToVice()
     {
-        elevenLabsClient.handleTextToVoice("我只是测试的文本转换成语音","/Users/wangxiangshun/Desktop/临时文件/output1112.mp3");
+        elevenLabsClient.handleTextToVoice("Hello ! I can","/Users/wangxiangshun/Desktop/临时文件/output1112.wav");
         return success();
     }
 

vetti-admin/target/classes/application-druid.yml

@@ -145,7 +145,8 @@ verification:
 # 文本转语音
 elevenLabs:
     baseUrl: https://api.elevenlabs.io/v1
-    apiKey: sk_5240d8f56cb1eb5225fffcf903f62479884d1af5b3de6812
+#    apiKey: sk_5240d8f56cb1eb5225fffcf903f62479884d1af5b3de6812
+    apiKey: sk_88f5a560e1bbde0e5b8b6b6eb1812163a98bfb98554acbec
     modelId: eleven_turbo_v2_5
 # 语音转文本
 whisper:

vetti-common/src/main/java/com/vetti/common/ai/gpt/OpenAiStreamClient.java

@@ -115,21 +115,22 @@ public class OpenAiStreamClient {
                                         .getJSONObject("delta")
                                         .getStr("content");
 
-                                if (content != null && !content.isEmpty()) {
+//                                if (content != null && !content.isEmpty()) {
-                                    if(punctuationSet.contains(content)){
+//                                    if(punctuationSet.contains(content)){
-                                        //说明有标点啦,直接返回
+//                                        //说明有标点啦,直接返回
-                                        bufferStr.append(content);
+//                                        bufferStr.append(content);
-                                        listener.onMessage(bufferStr.toString());
+//                                        listener.onMessage(bufferStr.toString());
-                                    }else{
+//                                    }else{
-                                        //加入缓冲区
+//                                        //加入缓冲区
-                                        if(StrUtil.isEmpty(bufferStr.toString())){
+//                                        if(StrUtil.isEmpty(bufferStr.toString())){
-                                            bufferStr.append(content);
+//                                            bufferStr.append(content);
-                                        }else {
+//                                        }else {
-                                            bufferStr.append(" ").append(content);
+//                                            bufferStr.append(" ").append(content);
-                                        }
+//                                        }
-                                    }
+//                                    }
-
+//
-                                }
+//                                }
+                                listener.onMessage(content);
                             } catch (Exception e) {
                                 listener.onError(new IOException("Parse error: " + e.getMessage()));
                             }