feat: add Hebrew language support with Chatterbox TTS and ivrit-ai Whisper

.gitignore

@@ -38,9 +38,14 @@ Thumbs.db
 data/profiles/*
 data/generations/*
 data/projects/*
+data/cache/*
+data/finetune/*
 data/voicebox.db
 !data/.gitkeep
 
+# Model binaries (downloaded at runtime)
+backend/models/
+
 # Logs
 *.log
 logs/

CHANGELOG.md

@@ -57,6 +57,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Audio export failing when Tauri save dialog returns object instead of string path
 
 ### Added
+- **Hebrew Language Support** - Full Hebrew voice cloning and transcription
+  - Chatterbox Multilingual TTS backend (`ResembleAI/chatterbox`) for Hebrew voice generation
+  - ivrit-ai Whisper models (`ivrit-ai/whisper-large-v3-turbo` and `ivrit-ai/whisper-large-v3`) for Hebrew transcription
+  - Automatic backend routing: Hebrew requests use Chatterbox, all other languages use Qwen3-TTS
+  - Auto-download of Hebrew models on first use with progress tracking
+  - Trailing silence trimming for Chatterbox output
+- **Model Unloading** - New `/models/{model_name}/unload` endpoint to free memory for individual models
 - **Makefile** - Comprehensive development workflow automation with commands for setup, development, building, testing, and code quality checks
   - Includes Python version detection and compatibility warnings
   - Self-documenting help system with `make help`
@@ -68,9 +75,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ---
 
-## [Unreleased - Planned]
+## Roadmap
 
-### Planned
 - Real-time streaming synthesis
 - Conversation mode with multiple speakers
 - Voice effects (pitch shift, reverb, M3GAN-style)
@@ -82,4 +88,5 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ---
 
+[Unreleased]: https://github.com/nadavox/voicebox/compare/v0.1.0...HEAD
 [0.1.0]: https://github.com/jamiepine/voicebox/releases/tag/v0.1.0

app/src/components/Generation/FloatingGenerateBox.tsx

@@ -13,7 +13,7 @@ import {
 } from '@/components/ui/select';
 import { Textarea } from '@/components/ui/textarea';
 import { useToast } from '@/components/ui/use-toast';
-import { LANGUAGE_OPTIONS } from '@/lib/constants/languages';
+import { LANGUAGE_OPTIONS, type LanguageCode } from '@/lib/constants/languages';
 import { useGenerationForm } from '@/lib/hooks/useGenerationForm';
 import { useProfile, useProfiles } from '@/lib/hooks/useProfiles';
 import { useAddStoryItem, useStory } from '@/lib/hooks/useStories';
@@ -75,6 +75,16 @@ export function FloatingGenerateBox({
     },
   });
 
+  // Auto-sync language when selected profile changes
+  useEffect(() => {
+    if (selectedProfile?.language) {
+      const profileLang = selectedProfile.language as LanguageCode;
+      if (LANGUAGE_OPTIONS.some((l) => l.value === profileLang)) {
+        form.setValue('language', profileLang);
+      }
+    }
+  }, [selectedProfile?.language, form]);
+
   // Click away handler to collapse the box
   useEffect(() => {
     function handleClickOutside(event: MouseEvent) {
@@ -383,7 +393,7 @@ export function FloatingGenerateBox({
                     name="language"
                     render={({ field }) => (
                       <FormItem className="flex-1 space-y-0">
-                        <Select onValueChange={field.onChange} defaultValue={field.value}>
+                        <Select onValueChange={field.onChange} value={field.value}>
                           <FormControl>
                             <SelectTrigger className="h-8 text-xs bg-card border-border rounded-full hover:bg-background/50 transition-all">
                               <SelectValue />

app/src/components/Generation/GenerationForm.tsx

@@ -1,3 +1,4 @@
+import { useEffect } from 'react';
 import { Loader2, Mic } from 'lucide-react';
 import { Button } from '@/components/ui/button';
 import { Card, CardContent, CardHeader, CardTitle } from '@/components/ui/card';
@@ -19,7 +20,7 @@ import {
   SelectValue,
 } from '@/components/ui/select';
 import { Textarea } from '@/components/ui/textarea';
-import { LANGUAGE_OPTIONS } from '@/lib/constants/languages';
+import { LANGUAGE_OPTIONS, type LanguageCode } from '@/lib/constants/languages';
 import { useGenerationForm } from '@/lib/hooks/useGenerationForm';
 import { useProfile } from '@/lib/hooks/useProfiles';
 import { useUIStore } from '@/stores/uiStore';
@@ -30,6 +31,16 @@ export function GenerationForm() {
 
   const { form, handleSubmit, isPending } = useGenerationForm();
 
+  // Auto-sync language when selected profile changes
+  useEffect(() => {
+    if (selectedProfile?.language) {
+      const profileLang = selectedProfile.language as LanguageCode;
+      if (LANGUAGE_OPTIONS.some((l) => l.value === profileLang)) {
+        form.setValue('language', profileLang);
+      }
+    }
+  }, [selectedProfile?.language, form]);
+
   async function onSubmit(data: Parameters<typeof handleSubmit>[0]) {
     await handleSubmit(data, selectedProfileId);
   }
@@ -105,7 +116,7 @@ export function GenerationForm() {
                 render={({ field }) => (
                   <FormItem>
                     <FormLabel>Language</FormLabel>
-                    <Select onValueChange={field.onChange} defaultValue={field.value}>
+                    <Select onValueChange={field.onChange} value={field.value}>
                       <FormControl>
                         <SelectTrigger>
                           <SelectValue />

app/src/components/ServerSettings/ModelManagement.tsx

@@ -1,5 +1,5 @@
 import { useMutation, useQuery, useQueryClient } from '@tanstack/react-query';
-import { Download, Loader2, Trash2 } from 'lucide-react';
+import { Download, Loader2, Power, Trash2 } from 'lucide-react';
 import { useCallback, useState } from 'react';
 import {
   AlertDialog,
@@ -136,6 +136,27 @@ export function ModelManagement() {
     },
   });
 
+  const unloadMutation = useMutation({
+    mutationFn: async (modelName: string) => {
+      return apiClient.unloadModel(modelName);
+    },
+    onSuccess: async (_data, modelName) => {
+      const model = modelStatus?.models.find((m) => m.model_name === modelName);
+      toast({
+        title: 'Model unloaded',
+        description: `${model?.display_name || modelName} has been unloaded from memory.`,
+      });
+      await queryClient.invalidateQueries({ queryKey: ['modelStatus'], refetchType: 'all' });
+    },
+    onError: (error: Error) => {
+      toast({
+        title: 'Unload failed',
+        description: error.message,
+        variant: 'destructive',
+      });
+    },
+  });
+
   const formatSize = (sizeMb?: number): string => {
     if (!sizeMb) return 'Unknown';
     if (sizeMb < 1024) return `${sizeMb.toFixed(1)} MB`;
@@ -164,7 +185,7 @@ export function ModelManagement() {
               </h3>
               <div className="space-y-2">
                 {modelStatus.models
-                  .filter((m) => m.model_name.startsWith('qwen-tts'))
+                  .filter((m) => m.model_name.startsWith('qwen-tts') || m.model_name === 'chatterbox-tts')
                   .map((model) => (
                     <ModelItem
                       key={model.model_name}
@@ -178,7 +199,9 @@ export function ModelManagement() {
                         });
                         setDeleteDialogOpen(true);
                       }}
+                      onUnload={() => unloadMutation.mutate(model.model_name)}
                       isDownloading={downloadingModel === model.model_name}
+                      isUnloading={unloadMutation.isPending && unloadMutation.variables === model.model_name}
                       formatSize={formatSize}
                     />
                   ))}
@@ -206,7 +229,9 @@ export function ModelManagement() {
                         });
                         setDeleteDialogOpen(true);
                       }}
+                      onUnload={() => unloadMutation.mutate(model.model_name)}
                       isDownloading={downloadingModel === model.model_name}
+                      isUnloading={unloadMutation.isPending && unloadMutation.variables === model.model_name}
                       formatSize={formatSize}
                     />
                   ))}
@@ -271,14 +296,16 @@ interface ModelItemProps {
   };
   onDownload: () => void;
   onDelete: () => void;
+  onUnload: () => void;
   isDownloading: boolean;  // Local state - true if user just clicked download
+  isUnloading: boolean;
   formatSize: (sizeMb?: number) => string;
 }
 
-function ModelItem({ model, onDownload, onDelete, isDownloading, formatSize }: ModelItemProps) {
+function ModelItem({ model, onDownload, onDelete, onUnload, isDownloading, isUnloading, formatSize }: ModelItemProps) {
   // Use server's downloading state OR local state (for immediate feedback before server updates)
   const showDownloading = model.downloading || isDownloading;
-  
+
   return (
     <div className="flex items-center justify-between p-3 border rounded-lg">
       <div className="flex-1">
@@ -305,9 +332,25 @@ function ModelItem({ model, onDownload, onDelete, isDownloading, formatSize }: M
       <div className="flex items-center gap-2">
         {model.downloaded && !showDownloading ? (
           <div className="flex items-center gap-2">
-            <div className="flex items-center gap-1 text-sm text-muted-foreground">
-              <span>Ready</span>
-            </div>
+            {model.loaded ? (
+              <Button
+                size="sm"
+                onClick={onUnload}
+                variant="outline"
+                disabled={isUnloading}
+                title="Unload model from memory"
+              >
+                {isUnloading ? (
+                  <Loader2 className="h-4 w-4 animate-spin" />
+                ) : (
+                  <Power className="h-4 w-4" />
+                )}
+              </Button>
+            ) : (
+              <div className="flex items-center gap-1 text-sm text-muted-foreground">
+                <span>Ready</span>
+              </div>
+            )}
             <Button
               size="sm"
               onClick={onDelete}

app/src/lib/api/client.ts

@@ -310,13 +310,10 @@ class ApiClient {
   }
 
   async triggerModelDownload(modelName: string): Promise<{ message: string }> {
-    console.log('[API] triggerModelDownload called for:', modelName, 'at', new Date().toISOString());
-    const result = await this.request<{ message: string }>('/models/download', {
+    return this.request<{ message: string }>('/models/download', {
       method: 'POST',
       body: JSON.stringify({ model_name: modelName } as ModelDownloadRequest),
     });
-    console.log('[API] triggerModelDownload response:', result);
-    return result;
   }
 
   async deleteModel(modelName: string): Promise<{ message: string }> {
@@ -325,6 +322,12 @@ class ApiClient {
     });
   }
 
+  async unloadModel(modelName: string): Promise<{ message: string }> {
+    return this.request<{ message: string }>(`/models/${modelName}/unload`, {
+      method: 'POST',
+    });
+  }
+
   // Task Management
   async getActiveTasks(): Promise<ActiveTasksResponse> {
     return this.request<ActiveTasksResponse>('/tasks/active');

app/src/lib/api/types.ts

@@ -34,6 +34,7 @@ export interface GenerationRequest {
   language: LanguageCode;
   seed?: number;
   model_size?: '1.7B' | '0.6B';
+  instruct?: string;
 }
 
 export interface GenerationResponse {

app/src/lib/constants/languages.ts

@@ -1,6 +1,6 @@
 /**
- * Supported languages for Qwen3-TTS
- * Based on: https://github.com/QwenLM/Qwen3-TTS
+ * Supported languages for voice generation.
+ * Most languages use Qwen3-TTS; Hebrew uses Chatterbox TTS.
  */
 
 export const SUPPORTED_LANGUAGES = {
@@ -14,6 +14,7 @@ export const SUPPORTED_LANGUAGES = {
   pt: 'Portuguese',
   es: 'Spanish',
   it: 'Italian',
+  he: 'Hebrew',
 } as const;
 
 export type LanguageCode = keyof typeof SUPPORTED_LANGUAGES;

app/src/lib/hooks/useGenerationForm.ts

@@ -67,8 +67,15 @@ export function useGenerationForm(options: UseGenerationFormOptions = {}) {
     try {
       setIsGenerating(true);
 
-      const modelName = `qwen-tts-${data.modelSize}`;
-      const displayName = data.modelSize === '1.7B' ? 'Qwen TTS 1.7B' : 'Qwen TTS 0.6B';
+      // Determine model name for download tracking
+      // Hebrew uses Chatterbox TTS, other languages use Qwen
+      const isHebrew = data.language === 'he';
+      const modelName = isHebrew ? 'chatterbox-tts' : `qwen-tts-${data.modelSize}`;
+      const displayName = isHebrew
+        ? 'Chatterbox TTS (Hebrew)'
+        : data.modelSize === '1.7B'
+          ? 'Qwen TTS 1.7B'
+          : 'Qwen TTS 0.6B';
 
       try {
         const modelStatus = await apiClient.getModelStatus();