feat: Add voice cloning support and prepare clone functionality in voice design

qwen3-tts-frontend/src/components/tts/CustomVoiceForm.tsx

@@ -160,6 +160,20 @@ const CustomVoiceForm = forwardRef<CustomVoiceFormHandle>((_props, ref) => {
 
       let result
       if (selectedItem?.source === 'saved-design') {
+        if (selectedItem.backendType === 'local') {
+          result = await ttsApi.createVoiceCloneJob({
+            text: data.text,
+            language: data.language,
+            ref_audio: null,
+            voice_design_id: selectedItem.designId,
+            max_new_tokens: data.max_new_tokens,
+            temperature: data.temperature,
+            top_k: data.top_k,
+            top_p: data.top_p,
+            repetition_penalty: data.repetition_penalty,
+            backend: 'local',
+          })
+        } else {
           result = await ttsApi.createVoiceDesignJob({
             text: data.text,
             language: data.language,
@@ -171,6 +185,7 @@ const CustomVoiceForm = forwardRef<CustomVoiceFormHandle>((_props, ref) => {
             top_p: data.top_p,
             repetition_penalty: data.repetition_penalty,
           })
+        }
       } else {
         result = await ttsApi.createCustomVoiceJob(data)
       }

qwen3-tts-frontend/src/components/tts/VoiceDesignForm.tsx

@@ -52,6 +52,7 @@ const VoiceDesignForm = forwardRef<VoiceDesignFormHandle>((_props, ref) => {
   })
   const [showSaveDialog, setShowSaveDialog] = useState(false)
   const [saveDesignName, setSaveDesignName] = useState('')
+  const [isPreparing, setIsPreparing] = useState(false)
 
   const { currentJob, isPolling, isCompleted, startPolling, elapsedTime } = useJobPolling()
   const { refresh } = useHistoryContext()
@@ -128,13 +129,29 @@ const VoiceDesignForm = forwardRef<VoiceDesignFormHandle>((_props, ref) => {
       toast.error('请输入设计名称')
       return
     }
+
     try {
-      await voiceDesignApi.create({
+      const backend = preferences?.default_backend || 'local'
+      const design = await voiceDesignApi.create({
         name: saveDesignName,
         instruct: instruct,
-        backend_type: preferences?.default_backend || 'local'
+        backend_type: backend
       })
+
       toast.success('音色设计已保存')
+
+      if (backend === 'local') {
+        setIsPreparing(true)
+        try {
+          await voiceDesignApi.prepareClone(design.id)
+          toast.success('音色克隆准备完成')
+        } catch (error) {
+          toast.error('准备克隆失败，但设计已保存')
+        } finally {
+          setIsPreparing(false)
+        }
+      }
+
       setShowSaveDialog(false)
       setSaveDesignName('')
     } catch (error) {
@@ -238,8 +255,8 @@ const VoiceDesignForm = forwardRef<VoiceDesignFormHandle>((_props, ref) => {
             }}>
               取消
             </Button>
-            <Button type="button" onClick={handleSaveDesign}>
-              保存
+            <Button type="button" onClick={handleSaveDesign} disabled={isPreparing}>
+              {isPreparing ? '准备中...' : '保存'}
             </Button>
           </DialogFooter>
         </DialogContent>

qwen3-tts-frontend/src/lib/api.ts

@@ -259,6 +259,9 @@ export const ttsApi = {
     if (data.x_vector_only_mode !== undefined) {
       formData.append('x_vector_only_mode', String(data.x_vector_only_mode))
     }
+    if (data.voice_design_id !== undefined) {
+      formData.append('voice_design_id', String(data.voice_design_id))
+    }
     if (data.max_new_tokens !== undefined) {
       formData.append('max_new_tokens', String(data.max_new_tokens))
     }
@@ -422,6 +425,13 @@ export const voiceDesignApi = {
   delete: async (id: number): Promise<void> => {
     await apiClient.delete(API_ENDPOINTS.VOICE_DESIGNS.DELETE(id))
   },
+
+  prepareClone: async (id: number): Promise<{ message: string; cache_id: number; ref_text: string }> => {
+    const response = await apiClient.post<{ message: string; cache_id: number; ref_text: string }>(
+      API_ENDPOINTS.VOICE_DESIGNS.PREPARE_CLONE(id)
+    )
+    return response.data
+  },
 }
 
 export default apiClient

qwen3-tts-frontend/src/lib/constants.ts

@@ -33,6 +33,7 @@ export const API_ENDPOINTS = {
     GET: (id: number) => `/voice-designs/${id}`,
     UPDATE: (id: number) => `/voice-designs/${id}`,
     DELETE: (id: number) => `/voice-designs/${id}`,
+    PREPARE_CLONE: (id: number) => `/voice-designs/${id}/prepare-clone`,
   },
 } as const
 

qwen3-tts-frontend/src/types/tts.ts

@@ -47,6 +47,7 @@ export interface VoiceCloneForm {
   top_p?: number
   repetition_penalty?: number
   backend?: string
+  voice_design_id?: number
 }
 
 export type SpeakerSource = 'builtin' | 'saved-design'