大模型交互-超拟人合成
1、超���人合成:将文字转化为自然流畅的人声,在实时语音合成的基础上,精准模拟人类的副语言现象,如呼吸、叹气、语速变化等,使得语音不仅流畅自然,更富有情感和生命力。
()2、唤醒的持久运行--->合成能力加持(唤醒成功后语音答复:主人 我在)--->调用在线或离线听写能力(建议用讯飞在线效果好)--->识别用户说的语音成文字后发给大模型--->建议调用讯飞星火认知大模型--->获取大模型答案后调用语音合成(合成在线离线均可)进行答案输出。这样就顺利实现了用纯语音与大模型进行交互!
3、在获取大模型答案后调用语音合成(合成在线离线均可)进行答案输出环节,讯飞推出超拟人合成,交互更像真人。
()4、通过对大模型返回及时性与合成及时性的结合,逻辑编排使得模拟真人交互成为可能。
5、超拟人合成Java调用示例给大家!
package com.iflytek; import com.google.gson.Gson; import okhttp3.HttpUrl; import org.java_websocket.WebSocket; import org.java_websocket.client.WebSocketClient; import org.java_websocket.handshake.ServerHandshake; import javax.crypto.Mac; import javax.crypto.spec.SecretKeySpec; import java.io.*; import java.net.URI; import java.net.URL; import java.nio.charset.StandardCharsets; import java.text.SimpleDateFormat; import java.util.*; /** * 语音合成流式 WebAPI 接口调用示例 接口文档(必看):https://www.xfyun.cn/doc/tts/online_tts/API.html * 发音人使用方式:登陆开放平台https://www.xfyun.cn/后,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 * 错误码链接:https://www.xfyun.cn/document/error-code (code返回错误码时必看) * 小语种需要传输小语种文本、使用小语种发音人vcn、tte=unicode以及修改文本编码方式 * */ public class WebTtsWs { // 地址与鉴权信息 // public static final String hostUrl = "https://tts-api.xfyun.cn/v2/tts"; // 普通合成 public static final String hostUrl = "https://cbm01.cn-huabei-1.xf-yun.com/v1/private/medd90fec"; // 超拟人合成 // 均到控制台-语音合成页面获取 public static final String appid = ""; public static final String apiSecret = ""; public static final String apiKey = ""; // 合成文本 public static final String TEXT = "商鞅(约公元前390年—前338年),卫国人,公孙氏,姬姓,名鞅。他是一位重要的战国时期政治家、改革家、思想家和军事家。作为法家派别的代表人物,商鞅积极实行变法。在秦孝公的请求下,他提出并实施了一系列深远影响的改革措施,这被称为“商鞅变法”。这些措施包括废除井田制、重视农业、奖励军功、统一度量衡和建立县制等。虽然在他的领导下,秦国强大并繁荣,但他也因其严苛的法律而饱受争议。总的来说,商鞅是一位影响力深远的历史人物。"; // 合成文本编码格式 public static final String TTE = "UTF8"; // 小语种必须使用UNICODE编码作为值 // 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"; // 合成文件名称 public static final String OUTPUT_FILE_PATH = "src/main/resources/tts/" + System.currentTimeMillis() + ".pcm"; // json public static final Gson gson = new Gson(); public static boolean wsCloseFlag = false; public static long startTime = System.currentTimeMillis(); public static long endTime = System.currentTimeMillis(); public static boolean ttsWorkingFlag = false; public static void main(String[] args) throws Exception { String wsUrl = getAuthUrl(hostUrl, apiKey, apiSecret).replace("https://", "wss://"); OutputStream outputStream = new FileOutputStream(OUTPUT_FILE_PATH); websocketWork(wsUrl, outputStream); } // Websocket方法 public static void websocketWork(String wsUrl, OutputStream outputStream) { try { URI uri = new URI(wsUrl); WebSocketClient webSocketClient = new WebSocketClient(uri) { @Override public void onOpen(ServerHandshake serverHandshake) { try { //实时播放 Constants.TTS_SOURCE_DATA_LINE.open(Constants.TTS_AUDIO_FORMAT); Constants.TTS_SOURCE_DATA_LINE.start(); } catch (Exception e) { e.printStackTrace(); } ttsWorkingFlag = true; startTime = System.currentTimeMillis(); System.out.println("ws建立连接成功..."); } @Override public void onMessage(String text) { System.out.println("返回结果打印:" + text); JsonParse myJsonParse = gson.fromJson(text, JsonParse.class); if (myJsonParse.header.code != 0) { System.out.println("发生错误,错误码为:" + myJsonParse.header.code); System.out.println("本次请求的sid为:" + myJsonParse.header.sid); } /* if (myJsonParse.header.status == 2) { try { outputStream.close(); } catch (IOException e) { e.printStackTrace(); } endTime = System.currentTimeMillis(); System.out.println("本次合成耗时:" + (endTime - startTime) + "ms"); System.out.println("本次请求的sid==>" + myJsonParse.header.sid); System.out.println("合成成功,文件保存路径为==>" + OUTPUT_FILE_PATH); // 可以关闭连接,释放资源 ttsWorkingFlag = false; wsCloseFlag = true; Constants.TTS_SOURCE_DATA_LINE.stop(); Constants.TTS_SOURCE_DATA_LINE.close(); }*/ if (myJsonParse.payload.audio.audio != null) { try { byte[] textBase64Decode = Base64.getDecoder().decode(myJsonParse.payload.audio.audio); outputStream.write(textBase64Decode); outputStream.flush(); } catch (Exception e) { e.printStackTrace(); } // 实时播报 try { byte[] textBase64Decode = Base64.getDecoder().decode(myJsonParse.payload.audio.audio); /* outputStream.write(textBase64Decode); outputStream.flush();*/ if (ttsWorkingFlag) { Constants.TTS_SOURCE_DATA_LINE.write(textBase64Decode, 0, textBase64Decode.length); //实时写音频流 // System.err.println("写入"); } else { Constants.TTS_SOURCE_DATA_LINE.stop(); Constants.TTS_SOURCE_DATA_LINE.close(); } } catch (Exception e) { e.printStackTrace(); } } } @Override public void onClose(int i, String s, boolean b) { System.out.println("ws链接已关闭,本次请求完成..."); } @Override public void onError(Exception e) { if (e.getMessage() != null) { System.out.println("发生错误 " + e.getMessage()); } } }; // 建立连接 webSocketClient.connect(); while (!webSocketClient.getReadyState().equals(WebSocket.READYSTATE.OPEN)) { //System.out.println("正在连接..."); Thread.sleep(100); } MyThread webSocketThread = new MyThread(webSocketClient); webSocketThread.start(); } catch (Exception e) { System.out.println(e.getMessage()); } } // 线程来发送音频与参数 static class MyThread extends Thread { WebSocketClient webSocketClient; public MyThread(WebSocketClient webSocketClient) { this.webSocketClient = webSocketClient; } public void run() { String requestJson;//请求参数json串 try { requestJson = "{\n" + " \"common\": {\n" + " \"app_id\": \"" + appid + "\"\n" + " },\n" + " \"business\": {\n" + " \"aue\": \"raw\",\n" + " \"tte\": \"" + TTE + "\",\n" + " \"ent\": \"intp65\",\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" + //" \"text\": \"" + Base64.getEncoder().encodeToString(TEXT.getBytes("UTF-16LE")) + "\"\n" + " }\n" + "}"; requestJson = "{\n" + " \"header\": {\n" + " \"app_id\": \"" + appid + "\",\n" + " \"status\": 0\n" + " },\n" + " \"parameter\": {\n" + " \"oral\": {\n" + " \"spark_assist\": 1,\n" + " \"oral_level\": \"mid\"\n" + " },\n" + " \"tts\": {\n" + " \"vcn\": \"x4_lingxiaoxuan_oral\",\n" + " \"speed\": 50,\n" + " \"volume\": 50,\n" + " \"pitch\": 50,\n" + " \"bgs\": 0,\n" + " \"reg\": 0,\n" + " \"rdn\": 0,\n" + " \"rhy\": 0,\n" + " \"scn\": 0,\n" + " \"version\": 0,\n" + " \"L5SilLen\": 1000,\n" + " \"ParagraphSilLen\": 0,\n" + " \"audio\": {\n" + " \"encoding\": \"raw\",\n" + " \"sample_rate\": 16000,\n" + " \"channels\": 1,\n" + " \"bit_depth\": 16,\n" + " \"frame_size\": 0\n" + " },\n" + " \"pybuf\": {\n" + " \"encoding\": \"utf8\",\n" + " \"compress\": \"raw\",\n" + " \"format\": \"plain\"\n" + " }\n" + " }\n" + " },\n" + " \"payload\": {\n" + " \"text\": {\n" + " \"encoding\": \"utf8\",\n" + " \"compress\": \"raw\",\n" + " \"format\": \"json\",\n" + " \"status\": 0,\n" + " \"seq\": 0,\n" + " \"text\": \"" + Base64.getEncoder().encodeToString(TEXT.getBytes(StandardCharsets.UTF_8)) + "\"\n" + " },\n" + " \"user_text\": {\n" + " \"encoding\": \"utf8\",\n" + " \"compress\": \"raw\",\n" + " \"format\": \"json\",\n" + " \"status\": 0,\n" + " \"seq\": 0,\n" + " \"text\": \"" + Base64.getEncoder().encodeToString(TEXT.getBytes(StandardCharsets.UTF_8)) + "\"\n" + " }\n" + " }\n" + "}"; System.err.println(requestJson); webSocketClient.send(requestJson); // 等待服务端返回完毕后关闭 while (!wsCloseFlag) { Thread.sleep(200); } webSocketClient.close(); } catch (Exception e) { e.printStackTrace(); } } } // 鉴权方法 public static String getAuthUrl(String hostUrl, String apiKey, String apiSecret) throws Exception { URL url = new URL(hostUrl); // 时间 SimpleDateFormat format = new SimpleDateFormat("EEE, dd MMM yyyy HH:mm:ss z", Locale.US); format.setTimeZone(TimeZone.getTimeZone("GMT")); String date = format.format(new Date()); // 拼接 String preStr = "host: " + url.getHost() + "\n" + "date: " + date + "\n" + "GET " + url.getPath() + " HTTP/1.1"; //System.out.println(preStr); // SHA256加密 Mac mac = Mac.getInstance("hmacsha256"); SecretKeySpec spec = new SecretKeySpec(apiSecret.getBytes(StandardCharsets.UTF_8), "hmacsha256"); mac.init(spec); byte[] hexDigits = mac.doFinal(preStr.getBytes(StandardCharsets.UTF_8)); // Base64加密 String sha = Base64.getEncoder().encodeToString(hexDigits); // 拼接 String authorization = String.format("api_key=\"%s\", algorithm=\"%s\", headers=\"%s\", signature=\"%s\"", apiKey, "hmac-sha256", "host date request-line", sha); // 拼接地址 HttpUrl httpUrl = Objects.requireNonNull(HttpUrl.parse("https://" + url.getHost() + url.getPath())).newBuilder().// addQueryParameter("authorization", Base64.getEncoder().encodeToString(authorization.getBytes(StandardCharsets.UTF_8))).// addQueryParameter("date", date).// addQueryParameter("host", url.getHost()).// build(); return httpUrl.toString(); } //返回的json结果拆解 class JsonParse { Header header; Payload payload; } class Header { int code; String sid; int status; } class Payload { Audio audio; } class Audio { String audio; int seq; } }
The End