update: 增加TTS音色语种选择

2026-07-22 15:13:55 +08:00 · 2026-02-09 15:28:31 +08:00
parent 520e24a54e
commit 4ea3bea85c
21 changed files with 220 additions and 61 deletions
@@ -41,6 +41,9 @@ public class AgentUpdateDTO implements Serializable {
    @Schema(description = "音色标识", example = "voice_02", nullable = true)
    private String ttsVoiceId;

+    @Schema(description = "音色语言", example = "普通话", nullable = true)
+    private String ttsLanguage;
+
    @Schema(description = "记忆模型标识", example = "mem_model_02", nullable = true)
    private String memModelId;

@@ -45,6 +45,9 @@ public class AgentEntity {
    @Schema(description = "音色标识")
    private String ttsVoiceId;

+    @Schema(description = "音色语言")
+    private String ttsLanguage;
+
    @Schema(description = "记忆模型标识")
    private String memModelId;

@@ -64,6 +64,11 @@ public class AgentTemplateEntity implements Serializable {
     */
    private String ttsVoiceId;

+    /**
+     * 音色语言
+     */
+    private String ttsLanguage;
+
    /**
     * 记忆模型标识
     */
@@ -282,6 +282,9 @@ public class AgentServiceImpl extends BaseServiceImpl<AgentDao, AgentEntity> imp
        if (dto.getTtsVoiceId() != null) {
            existingEntity.setTtsVoiceId(dto.getTtsVoiceId());
        }
+        if (dto.getTtsLanguage() != null) {
+            existingEntity.setTtsLanguage(dto.getTtsLanguage());
+        }
        if (dto.getMemModelId() != null) {
            existingEntity.setMemModelId(dto.getMemModelId());
        }
@@ -87,6 +87,7 @@ public class ConfigServiceImpl implements ConfigService {
                null,
                null,
                null,
+                null,
                agent.getVadModelId(),
                agent.getAsrModelId(),
                null,
@@ -135,15 +136,24 @@ public class ConfigServiceImpl implements ConfigService {
        String voice = null;
        String referenceAudio = null;
        String referenceText = null;
+        String language = null;
        TimbreDetailsVO timbre = timbreService.get(agent.getTtsVoiceId());
        if (timbre != null) {
            voice = timbre.getTtsVoice();
            referenceAudio = timbre.getReferenceAudio();
            referenceText = timbre.getReferenceText();
+            // 优先使用用户选择的语言，如果没有则使用音色支持的第一个语言
+            if (StringUtils.isNotBlank(agent.getTtsLanguage())) {
+                language = agent.getTtsLanguage();
+            } else if (StringUtils.isNotBlank(timbre.getLanguages())) {
+                language = timbre.getLanguages().split("、")[0].trim();
+            }
        } else {
            VoiceCloneEntity voice_print = cloneVoiceService.selectById(agent.getTtsVoiceId());
            if (voice_print != null) {
                voice = voice_print.getVoiceId();
+                // 优先使用用户选择的语言，如果没有则使用默认值
+                language = StringUtils.isNotBlank(agent.getTtsLanguage()) ? agent.getTtsLanguage() : "普通话";
            }
        }
        // 构建返回数据
@@ -208,6 +218,7 @@ public class ConfigServiceImpl implements ConfigService {
                voice,
                referenceAudio,
                referenceText,
+                language,
                agent.getVadModelId(),
                agent.getAsrModelId(),
                agent.getLlmModelId(),
@@ -385,6 +396,7 @@ public class ConfigServiceImpl implements ConfigService {
            String voice,
            String referenceAudio,
            String referenceText,
+            String language,
            String vadModelId,
            String asrModelId,
            String llmModelId,
@@ -423,6 +435,8 @@ public class ConfigServiceImpl implements ConfigService {
                        ((Map<String, Object>) model.getConfigJson()).put("ref_audio", referenceAudio);
                    if (referenceText != null)
                        ((Map<String, Object>) model.getConfigJson()).put("ref_text", referenceText);
+                    if (language != null)
+                        ((Map<String, Object>) model.getConfigJson()).put("language", language);

                    // 火山引擎声音克隆需要替换resource_id
                    Map<String, Object> map = (Map<String, Object>) model.getConfigJson();
@@ -23,6 +23,9 @@ public class VoiceDTO implements Serializable {
    @Schema(description = "音频播放地址")
    private String voiceDemo;
    
+    @Schema(description = "语言类型")
+    private String languages;
+    
    @Schema(description = "是否为克隆音色")
    private Boolean isClone;

@@ -31,6 +34,7 @@ public class VoiceDTO implements Serializable {
        this.id = id;
        this.name = name;
        this.voiceDemo = null;
+        this.languages = null;
        this.isClone = false; // 默认不是克隆音色
    }
    
@@ -39,6 +43,7 @@ public class VoiceDTO implements Serializable {
        this.id = id;
        this.name = name;
        this.voiceDemo = voiceDemo;
+        this.languages = null;
        this.isClone = false;
    }

@@ -129,6 +129,7 @@ public class TimbreServiceImpl extends BaseServiceImpl<TimbreDao, TimbreEntity>
                .map(entity -> {
                    VoiceDTO dto = new VoiceDTO(entity.getId(), entity.getName());
                    dto.setVoiceDemo(entity.getVoiceDemo());
+                    dto.setLanguages(entity.getLanguages()); // 设置语言类型
                    dto.setIsClone(false); // 设置为普通音色
                    return dto;
                })
@@ -146,6 +147,7 @@ public class TimbreServiceImpl extends BaseServiceImpl<TimbreDao, TimbreEntity>
                voiceDTO.setName(MessageUtils.getMessage(ErrorCode.VOICE_CLONE_PREFIX) + entity.getName());
                // 保留从数据库查询到的voiceDemo字段
                voiceDTO.setVoiceDemo(entity.getVoiceDemo());
+                voiceDTO.setLanguages(entity.getLanguages());
                voiceDTO.setIsClone(true); // 设置为克隆音色
                redisUtils.set(RedisKeys.getTimbreNameById(voiceDTO.getId()), voiceDTO.getName(),
                        RedisUtils.NOT_EXPIRE);
@@ -11,4 +11,10 @@ SET languages = CASE
    WHEN languages = '中文(粤语)及中英文混合' THEN '粤语、英语'
    WHEN languages = '粤语及粤英混合' THEN '粤语、英语'
    ELSE languages
-END;
+END;
+
+-- 添加音色语言字段到 ai_agent 表
+ALTER TABLE `ai_agent` ADD COLUMN `tts_language` VARCHAR(50) NULL COMMENT '音色语言' AFTER `tts_voice_id`;
+
+-- 添加音色语言字段到 ai_agent_template 表
+ALTER TABLE `ai_agent_template` ADD COLUMN `tts_language` VARCHAR(50) NULL COMMENT '音色语言' AFTER `tts_voice_id`;
@@ -16,6 +16,7 @@
        <result column="llmModelId" property="llmModelId"/>
        <result column="ttsModelId" property="ttsModelId"/>
        <result column="ttsVoiceId" property="ttsVoiceId"/>
+        <result column="ttsLanguage" property="ttsLanguage"/>
        <result column="memModelId" property="memModelId"/>
        <result column="intentModelId" property="intentModelId"/>

@@ -45,6 +46,7 @@
               a.vllm_model_id     AS vllmModelId,
               a.tts_model_id      AS ttsModelId,
               a.tts_voice_id      AS ttsVoiceId,
+               a.tts_language      AS ttsLanguage,
               a.mem_model_id      AS memModelId,
               a.intent_model_id   AS intentModelId,
               COALESCE(