jamiepine · at-aepeak-com · Jul 19, 2026 · Jul 19, 2026 · Jul 19, 2026 · Jul 19, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -7,6 +7,27 @@
 
 ## [Unreleased]
 
+### Voice Generation
+
+- **Romanian voice cloning via a new F5-TTS engine.** A new `f5` engine wraps
+  the community Romanian fine-tune of F5-TTS (`MihaiPopa-1/F5-TTS-Romanian`,
+  Apache-2.0) — the first engine with true Romanian voice cloning: record a
+  cloned profile and generate Romanian (or English) speech in that voice,
+  zero-shot. References longer than 12s are trimmed at the quietest pause and
+  both Romanian diacritic conventions are normalized to the model's vocab.
+  Heads-up: generation uses 32-step flow matching and is well below realtime
+  (~12x slower on Apple Silicon MPS, ~20x on CPU); the ~1.2GB checkpoint plus
+  the Vocos vocoder download on first use. Adds one Python dependency
+  (`f5-tts`, pinned); FFmpeg is required on the system (torchcodec audio I/O).
+
+- **Romanian TTS via a new MMS engine.** A new `mms` engine wraps Meta's
+  MMS-TTS Romanian checkpoint (`facebook/mms-tts-ron`) — a ~150MB VITS model
+  that runs realtime on CPU, using the already-bundled transformers runtime
+  (zero new dependencies). It ships as a preset voice (like Kokoro): create a
+  profile with the Romanian MMS voice and generate. Romanian text using either
+  diacritic convention (cedilla ş/ţ or comma-below ș/ț) is normalized
+  automatically so no character is silently dropped by the tokenizer.
+
 ### Linux
 
 - **ROCm setup works on Linux AMD systems.** Docker ROCm builds now keep PyTorch

diff --git a/README.md b/README.md
@@ -67,14 +67,14 @@
 
 ## What is Voicebox?
 
-Voicebox is a **local-first AI voice studio** — a free and open-source alternative to **ElevenLabs** and **WisprFlow** in one app. Clone voices from a few seconds of audio, generate speech in 23 languages across 7 TTS engines, dictate into any text field with a global hotkey, and give any MCP-aware AI agent a voice of your choosing.
+Voicebox is a **local-first AI voice studio** — a free and open-source alternative to **ElevenLabs** and **WisprFlow** in one app. Clone voices from a few seconds of audio, generate speech in 24 languages across 9 TTS engines, dictate into any text field with a global hotkey, and give any MCP-aware AI agent a voice of your choosing.
 
 The two cloud incumbents sit on opposite halves of the voice I/O loop — ElevenLabs on output, WisprFlow on input. Voicebox does both, bridges them with a bundled local LLM for refinement and per-profile personas, and runs the whole thing on your machine.
 
 - **Complete privacy** — models, voice data, and captures never leave your machine
-- **7 TTS engines** — Qwen3-TTS, Qwen CustomVoice, LuxTTS, Chatterbox Multilingual, Chatterbox Turbo, HumeAI TADA, and Kokoro
-- **Voice cloning and preset voices** — zero-shot cloning from a reference sample, or 50+ curated preset voices via Kokoro and Qwen CustomVoice
-- **23 languages** — from English to Arabic, Japanese, Hindi, Swahili, and more
+- **9 TTS engines** — Qwen3-TTS, Qwen CustomVoice, LuxTTS, Chatterbox Multilingual, Chatterbox Turbo, HumeAI TADA, Kokoro, MMS, and F5-TTS
+- **Voice cloning and preset voices** — zero-shot cloning from a reference sample, or 50+ curated preset voices via Kokoro, Qwen CustomVoice, and MMS (Romanian)
+- **24 languages** — from English to Arabic, Japanese, Hindi, Romanian, Swahili, and more
 - **Post-processing effects** — pitch shift, reverb, delay, chorus, compression, and filters
 - **Expressive speech** — paralinguistic tags like `[laugh]`, `[sigh]`, `[gasp]` via Chatterbox Turbo; natural-language delivery control via Qwen CustomVoice
 - **Unlimited length** — auto-chunking with crossfade for scripts, articles, and chapters
@@ -109,7 +109,7 @@ The two cloud incumbents sit on opposite halves of the voice I/O loop — Eleven
 
 ### Multi-Engine Voice Cloning
 
-Seven TTS engines with different strengths, switchable per-generation:
+Nine TTS engines with different strengths, switchable per-generation:
 
 | Engine                      | Languages | Strengths                                                                                                                                |
 | --------------------------- | --------- | ---------------------------------------------------------------------------------------------------------------------------------------- |
@@ -120,6 +120,8 @@ Seven TTS engines with different strengths, switchable per-generation:
 | **Chatterbox Turbo**        | English   | Fast 350M model with paralinguistic emotion/sound tags                                                                                   |
 | **TADA** (1B / 3B)          | 10        | HumeAI speech-language model — 700s+ coherent audio, text-acoustic dual alignment                                                        |
 | **Kokoro**                  | 8         | 50 curated preset voices, tiny 82M model, fast CPU inference                                                                             |
+| **MMS** (Meta)              | Romanian  | Per-language VITS checkpoints (~150MB), CPU realtime, preset voice                                                                       |
+| **F5-TTS Romanian**         | Romanian, English | Community fine-tune with true Romanian voice cloning — zero-shot from a reference sample (slow: flow matching, ~12-20x realtime)   |
 
 ### Emotions & Paralinguistic Tags
 
@@ -369,7 +371,7 @@ Full API documentation available at `http://127.0.0.1:17493/docs`.
 | Frontend      | React, TypeScript, Tailwind CSS                                                 |
 | State         | Zustand, React Query                                                            |
 | Backend       | FastAPI (Python)                                                                |
-| TTS Engines   | Qwen3-TTS, Qwen CustomVoice, LuxTTS, Chatterbox, Chatterbox Turbo, TADA, Kokoro |
+| TTS Engines   | Qwen3-TTS, Qwen CustomVoice, LuxTTS, Chatterbox, Chatterbox Turbo, TADA, Kokoro, MMS, F5-TTS |
 | STT           | Whisper / Whisper Turbo (PyTorch or MLX)                                        |
 | Local LLM     | Qwen3 (0.6B / 1.7B / 4B), shared runtime with TTS / STT                         |
 | MCP Server    | FastMCP mounted at `/mcp` (Streamable HTTP) + bundled stdio shim binary         |
@@ -416,7 +418,7 @@ just dev     # starts backend + desktop app
 
 Install [just](https://github.com/casey/just): `brew install just` or `cargo install just`. Run `just --list` to see all commands.
 
-**Prerequisites:** [Bun](https://bun.sh), [Rust](https://rustup.rs), [Python 3.11+](https://python.org), [Tauri Prerequisites](https://v2.tauri.app/start/prerequisites/), and [Xcode](https://developer.apple.com/xcode/) on macOS.
+**Prerequisites:** [Bun](https://bun.sh), [Rust](https://rustup.rs), [Python 3.11+](https://python.org), [FFmpeg](https://ffmpeg.org) (`brew install ffmpeg` — required for F5-TTS generation and torchaudio ≥2.11 audio I/O), [Tauri Prerequisites](https://v2.tauri.app/start/prerequisites/), and [Xcode](https://developer.apple.com/xcode/) on macOS.
 
 The repo ships a pre-wired `.mcp.json` at the root — running Claude Code inside this checkout picks up the Voicebox MCP tools automatically once the dev app is running.
 

diff --git a/app/src/components/CapturesTab/CapturesTab.tsx b/app/src/components/CapturesTab/CapturesTab.tsx
@@ -64,7 +64,7 @@ import type {
   CaptureSource,
   VoiceProfileResponse,
 } from '@/lib/api/types';
-import type { LanguageCode } from '@/lib/constants/languages';
+import { ENGINE_LANGUAGES, type LanguageCode } from '@/lib/constants/languages';
 import { BOTTOM_SAFE_AREA_PADDING } from '@/lib/constants/ui';
 import { useCaptureRecordingSession } from '@/lib/hooks/useCaptureRecordingSession';
 import { useDictationReadiness } from '@/lib/hooks/useDictationReadiness';
@@ -263,14 +263,23 @@ export function CapturesTab() {
     mutationFn: async ({ capture, voice }: { capture: CaptureResponse; voice: VoiceProfileResponse }) => {
       const text = capture.transcript_refined || capture.transcript_raw;
       if (!text.trim()) throw new Error(t('captures.noTranscriptError'));
-      const language = (capture.language || voice.language) as LanguageCode;
       // Preset profiles (Kokoro etc.) reject the qwen default — honor the
       // profile's stored engine preference. Cloned profiles without an
       // override fall through to whatever the backend picks.
-      const engine = voice.default_engine as
+      const engine = (voice.default_engine ?? voice.preset_engine) as
         | 'qwen' | 'qwen_custom_voice' | 'luxtts' | 'chatterbox'
-        | 'chatterbox_turbo' | 'tada' | 'kokoro'
+        | 'chatterbox_turbo' | 'tada' | 'kokoro' | 'mms' | 'f5'
         | undefined;
+      // Prefer the capture's language, but only if the target engine can
+      // speak it (e.g. MMS is Romanian-only, Kokoro covers 8 languages) —
+      // otherwise fall back to the profile's own language.
+      const captureLanguage = capture.language as LanguageCode | undefined;
+      const supported = engine ? ENGINE_LANGUAGES[engine] : undefined;
+      const language = (
+        captureLanguage && (!supported || supported.includes(captureLanguage))
+          ? captureLanguage
+          : voice.language
+      ) as LanguageCode;
       return apiClient.generateSpeech({
         profile_id: voice.id,
         text,

diff --git a/app/src/components/Generation/EngineModelSelector.tsx b/app/src/components/Generation/EngineModelSelector.tsx
@@ -27,6 +27,8 @@ const ENGINE_OPTIONS = [
   { value: 'tada:1B', label: 'TADA 1B', engine: 'tada' },
   { value: 'tada:3B', label: 'TADA 3B Multilingual', engine: 'tada' },
   { value: 'kokoro', label: 'Kokoro 82M', engine: 'kokoro' },
+  { value: 'mms', label: 'MMS Romanian', engine: 'mms' },
+  { value: 'f5', label: 'F5-TTS Romanian', engine: 'f5' },
 ] as const;
 
 const ENGINE_DESCRIPTIONS: Record<string, string> = {
@@ -37,13 +39,15 @@ const ENGINE_DESCRIPTIONS: Record<string, string> = {
   chatterbox_turbo: 'English, [laugh] [cough] tags',
   tada: 'HumeAI, 700s+ coherent audio',
   kokoro: '82M params, CPU realtime, 8 langs',
+  mms: 'Meta MMS, Romanian, CPU realtime',
+  f5: 'Romanian voice cloning, slow but faithful',
 };
 
 /** Engines that only support English and should force language to 'en' on select. */
 const ENGLISH_ONLY_ENGINES = new Set(['luxtts', 'chatterbox_turbo']);
 
 /** Engines that support cloned (reference audio) profiles. */
-const CLONING_ENGINES = new Set(['qwen', 'luxtts', 'chatterbox', 'chatterbox_turbo', 'tada']);
+const CLONING_ENGINES = new Set(['qwen', 'luxtts', 'chatterbox', 'chatterbox_turbo', 'tada', 'f5']);
 
 function getAvailableOptions(selectedProfile?: VoiceProfileResponse | null) {
   if (!selectedProfile) return ENGINE_OPTIONS;

diff --git a/app/src/components/Generation/FloatingGenerateBox.tsx b/app/src/components/Generation/FloatingGenerateBox.tsx
@@ -151,7 +151,9 @@ export function FloatingGenerateBox({
     | 'chatterbox_turbo'
     | 'tada'
     | 'kokoro'
-    | 'qwen_custom_voice';
+    | 'qwen_custom_voice'
+    | 'mms'
+    | 'f5';
   useEffect(() => {
     if (selectedProfile?.language) {
       form.setValue('language', selectedProfile.language as LanguageCode);
@@ -163,7 +165,7 @@ export function FloatingGenerateBox({
     } else if (selectedProfile && selectedProfile.voice_type !== 'preset') {
       // Cloned/designed profile with no default — ensure a compatible (non-preset) engine
       const currentEngine = form.getValues('engine');
-      const presetEngines = new Set(['kokoro', 'qwen_custom_voice']);
+      const presetEngines = new Set(['kokoro', 'qwen_custom_voice', 'mms']);
       if (currentEngine && presetEngines.has(currentEngine)) {
         form.setValue('engine', 'qwen');
       }
@@ -418,13 +420,19 @@ export function FloatingGenerateBox({
                                         ? 'bg-accent text-accent-foreground border border-accent hover:bg-accent/90'
                                         : 'bg-card border border-border hover:bg-background/50',
                                     )}
-                                    aria-label={active ? t('generation.persona.ariaLabelActive') : t('generation.persona.ariaLabelInactive')}
+                                    aria-label={
+                                      active
+                                        ? t('generation.persona.ariaLabelActive')
+                                        : t('generation.persona.ariaLabelInactive')
+                                    }
                                     aria-pressed={active}
                                   >
                                     <Wand2 className="h-4 w-4" />
                                   </Button>
                                   <span className="pointer-events-none absolute bottom-full left-1/2 -translate-x-1/2 mb-2 whitespace-nowrap rounded-md bg-popover px-3 py-1.5 text-xs text-popover-foreground border border-border opacity-0 transition-opacity group-hover:opacity-100 z-[9999]">
-                                    {active ? t('generation.persona.tooltipActive') : t('generation.persona.tooltipInactive')}
+                                    {active
+                                      ? t('generation.persona.tooltipActive')
+                                      : t('generation.persona.tooltipInactive')}
                                   </span>
                                 </div>
                               </FormControl>
@@ -566,7 +574,6 @@ export function FloatingGenerateBox({
                     </div>
                   )}
 
-
                   <FormField
                     control={form.control}
                     name="language"

diff --git a/app/src/components/ServerSettings/ModelManagement.tsx b/app/src/components/ServerSettings/ModelManagement.tsx
@@ -414,7 +414,9 @@ export function ModelManagement() {
         m.model_name.startsWith('luxtts') ||
         m.model_name.startsWith('chatterbox') ||
         m.model_name.startsWith('tada') ||
-        m.model_name.startsWith('kokoro'),
+        m.model_name.startsWith('kokoro') ||
+        m.model_name.startsWith('mms') ||
+        m.model_name.startsWith('f5'),
     ) ?? [];
   const whisperModels = modelStatus?.models.filter((m) => m.model_name.startsWith('whisper')) ?? [];
   const llmModels = modelStatus?.models.filter((m) => m.model_name.startsWith('qwen3-')) ?? [];

diff --git a/app/src/components/VoiceProfiles/ProfileCard.tsx b/app/src/components/VoiceProfiles/ProfileCard.tsx
@@ -22,6 +22,7 @@ import { useUIStore } from '@/stores/uiStore';
 const ENGINE_DISPLAY_NAMES: Record<string, string> = {
   kokoro: 'Kokoro',
   qwen_custom_voice: 'CustomVoice',
+  mms: 'MMS',
 };
 
 interface ProfileCardProps {

diff --git a/app/src/components/VoiceProfiles/ProfileForm.tsx b/app/src/components/VoiceProfiles/ProfileForm.tsx
@@ -61,7 +61,7 @@ import { AudioSampleUpload } from './AudioSampleUpload';
 import { SampleList } from './SampleList';
 
 const MAX_AUDIO_DURATION_SECONDS = 30;
-const PRESET_ONLY_ENGINES = new Set(['kokoro', 'qwen_custom_voice']);
+const PRESET_ONLY_ENGINES = new Set(['kokoro', 'qwen_custom_voice', 'mms']);
 const DEFAULT_ENGINE_OPTIONS = [
   { value: 'qwen', label: 'Qwen3-TTS' },
   { value: 'qwen_custom_voice', label: 'Qwen CustomVoice' },
@@ -70,6 +70,8 @@ const DEFAULT_ENGINE_OPTIONS = [
   { value: 'chatterbox_turbo', label: 'Chatterbox Turbo' },
   { value: 'tada', label: 'TADA' },
   { value: 'kokoro', label: 'Kokoro 82M' },
+  { value: 'mms', label: 'MMS Romanian' },
+  { value: 'f5', label: 'F5-TTS Romanian' },
 ] as const;
 
 function makeProfileSchema(t: (key: string) => string) {
@@ -898,6 +900,7 @@ export function ProfileForm() {
                               <SelectContent>
                                 <SelectItem value="kokoro">Kokoro 82M</SelectItem>
                                 <SelectItem value="qwen_custom_voice">Qwen CustomVoice</SelectItem>
+                                <SelectItem value="mms">MMS Romanian</SelectItem>
                               </SelectContent>
                             </Select>
                           </FormItem>
@@ -1207,9 +1210,7 @@ export function ProfileForm() {
                             {...field}
                           />
                         </FormControl>
-                        <FormDescription>
-                          {t('profileForm.fields.personalityHint')}
-                        </FormDescription>
+                        <FormDescription>{t('profileForm.fields.personalityHint')}</FormDescription>
                         <FormMessage />
                       </FormItem>
                     )}

diff --git a/app/src/components/VoiceProfiles/ProfileList.tsx b/app/src/components/VoiceProfiles/ProfileList.tsx
@@ -9,7 +9,7 @@ import { ProfileCard } from './ProfileCard';
 import { ProfileForm } from './ProfileForm';
 
 /** Engines that use preset (built-in) voices instead of cloned profiles. */
-const PRESET_ENGINES = new Set(['kokoro', 'qwen_custom_voice']);
+const PRESET_ENGINES = new Set(['kokoro', 'qwen_custom_voice', 'mms']);
 
 export function ProfileList() {
   const { t } = useTranslation();

diff --git a/app/src/lib/api/types.ts b/app/src/lib/api/types.ts
@@ -78,7 +78,9 @@ export interface GenerationRequest {
     | 'chatterbox'
     | 'chatterbox_turbo'
     | 'tada'
-    | 'kokoro';
+    | 'kokoro'
+    | 'mms'
+    | 'f5';
   instruct?: string;
   /** When true and the profile has a personality prompt, input text is rewritten in-character before TTS. */
   personality?: boolean;

diff --git a/app/src/lib/constants/languages.ts b/app/src/lib/constants/languages.ts
@@ -6,6 +6,7 @@
  * Chatterbox Multilingual supports 23 languages.
  * Chatterbox Turbo is English-only.
  * Kokoro supports 8 languages.
+ * MMS is Romanian-only (per-language Meta checkpoints).
  */
 
 /** All languages that any engine supports. */
@@ -28,6 +29,7 @@ export const ALL_LANGUAGES = {
   no: 'Norwegian',
   pl: 'Polish',
   pt: 'Portuguese',
+  ro: 'Romanian',
   ru: 'Russian',
   sv: 'Swedish',
   sw: 'Swahili',
@@ -70,6 +72,9 @@ export const ENGINE_LANGUAGES: Record<string, readonly LanguageCode[]> = {
   tada: ['en', 'ar', 'zh', 'de', 'es', 'fr', 'it', 'ja', 'pl', 'pt'],
   kokoro: ['en', 'es', 'fr', 'hi', 'it', 'pt', 'ja', 'zh'],
   qwen_custom_voice: ['zh', 'en', 'ja', 'ko', 'de', 'fr', 'ru', 'pt', 'es', 'it'],
+  mms: ['ro'],
+  // F5-TTS Romanian community fine-tune — Romanian cloning; retains English.
+  f5: ['ro', 'en'],
 } as const;
 
 /** Helper: get language options for a given engine. */

diff --git a/app/src/lib/hooks/useGenerationForm.ts b/app/src/lib/hooks/useGenerationForm.ts
@@ -27,6 +27,8 @@ const generationSchema = z.object({
       'chatterbox_turbo',
       'tada',
       'kokoro',
+      'mms',
+      'f5',
     ])
     .optional(),
   personality: z.boolean().optional(),
@@ -100,9 +102,13 @@ export function useGenerationForm(options: UseGenerationFormOptions = {}) {
                   : 'tada-1b'
                 : engine === 'kokoro'
                   ? 'kokoro'
-                  : engine === 'qwen_custom_voice'
-                    ? `qwen-custom-voice-${data.modelSize}`
-                    : `qwen-tts-${data.modelSize}`;
+                  : engine === 'mms'
+                    ? 'mms-tts-ron'
+                    : engine === 'f5'
+                      ? 'f5-tts-romanian'
+                      : engine === 'qwen_custom_voice'
+                        ? `qwen-custom-voice-${data.modelSize}`
+                        : `qwen-tts-${data.modelSize}`;
       const displayName =
         engine === 'luxtts'
           ? 'LuxTTS'
@@ -116,13 +122,17 @@ export function useGenerationForm(options: UseGenerationFormOptions = {}) {
                   : 'TADA 1B'
                 : engine === 'kokoro'
                   ? 'Kokoro 82M'
-                  : engine === 'qwen_custom_voice'
-                    ? data.modelSize === '1.7B'
-                      ? 'Qwen CustomVoice 1.7B'
-                      : 'Qwen CustomVoice 0.6B'
-                    : data.modelSize === '1.7B'
-                      ? 'Qwen TTS 1.7B'
-                      : 'Qwen TTS 0.6B';
+                  : engine === 'mms'
+                    ? 'MMS Romanian (Meta)'
+                    : engine === 'f5'
+                      ? 'F5-TTS Romanian (community)'
+                      : engine === 'qwen_custom_voice'
+                        ? data.modelSize === '1.7B'
+                          ? 'Qwen CustomVoice 1.7B'
+                          : 'Qwen CustomVoice 0.6B'
+                        : data.modelSize === '1.7B'
+                          ? 'Qwen TTS 1.7B'
+                          : 'Qwen TTS 0.6B';
 
       // Check if model needs downloading
       try {

diff --git a/app/src/lib/utils/format.ts b/app/src/lib/utils/format.ts
@@ -57,6 +57,8 @@ const ENGINE_DISPLAY_NAMES: Record<string, string> = {
   luxtts: 'LuxTTS',
   chatterbox: 'Chatterbox',
   chatterbox_turbo: 'Chatterbox Turbo',
+  mms: 'MMS TTS',
+  f5: 'F5-TTS',
 };
 
 export function formatEngineName(engine?: string, modelSize?: string): string {