The open-source AI voice studio. Clone, dictate, create.
Merge origin/main (Qwen VoiceDesign) into prep/pr-1062, keep both engines
jamiepinebb6ee6a8e863
diff --cc CHANGELOG.mdindex ab61a950,518f3e48..3bdaefe1--- a/CHANGELOG.md+++ b/CHANGELOG.md@@@ -13,11 -17,17 +17,21 @@@### API- **OpenAI-compatible text-to-speech API.** `POST /v1/audio/speech` and `GET /v1/models` accept OpenAI's request shape, so the official `openai` SDKs, Open WebUI, Home Assistant and any other OpenAI TTS client become a local, offline TTS by changing only the base URL. `voice` is a Voicebox profile (name or id, falling back to the per-client binding and default voice like `/speak`); `model` is a Voicebox engine or model variant, with `tts-1`, `tts-1-hd` and `gpt-4o-mini-tts` accepted as aliases for the profile's configured engine; `response_format` supports `wav`, `mp3`, `flac`, `opus`, `pcm` and (with ffmpeg) `aac`; `speed` is a pitch-preserving time stretch; `instructions` feeds the engine's instruct prompt. Errors use OpenAI's `{"error": {...}}` body, and the endpoint is documented on the MCP Server page. — @neuron-tech-ai ([#1170](https://github.com/jamiepine/voicebox/pull/1170), from [#656](https://github.com/jamiepine/voicebox/pull/656); fixes [#10](https://github.com/jamiepine/voicebox/issues/10))+ ### Linux++ - **Installing the CUDA or ROCm backend on Linux downloads a native Linux binary.** The GPU server release assets are now platform-qualified (`voicebox-server-{cuda,rocm}-<platform>.tar.gz`, `{cuda,rocm}-libs-<platform>-<version>.tar.gz`) and a `build-gpu-linux` release job builds the Linux CUDA and ROCm bundles alongside the Windows ones, so Linux no longer fetches a Windows `.exe` and silently falls back to CPU. The ROCm download is offered on Linux hosts with an AMD ROCm driver (`/dev/kfd`), and the CUDA download gate from 0.6 now refuses only platforms without a published asset (macOS, Linux arm64). — @1337hero, @smendola ([#851](https://github.com/jamiepine/voicebox/pull/851), [#561](https://github.com/jamiepine/voicebox/pull/561); fixes [#767](https://github.com/jamiepine/voicebox/issues/767))+ ### Linux++ - **Closing the window quits the app.** With "Keep server running" off, closing the main window on Linux left the Tauri process alive after the backend had stopped (the hidden dictation pill window kept it running), and its speak monitor retried `/events/speak` every 30 seconds forever. On Linux the close flow now exits the application once the frontend finishes its cleanup, or after the 5-second timeout; macOS and Windows keep their existing close behaviour. — @Fulminao ([#1040](https://github.com/jamiepine/voicebox/pull/1040))+ - **`just build-server-cuda` works on Linux.** The recipe builds the CUDA backend with `build_binary.py --cuda` and installs it into `$XDG_DATA_HOME/sh.voicebox.app/backends/cuda` (default `~/.local/share`), matching the Windows recipe. — @Fulminao ([#1044](https://github.com/jamiepine/voicebox/pull/1044))+ ### Export++ - **Stories export as M4B audiobooks or MP3 with chapter markers, and single generations export as MP3.** `GET /stories/{id}/export-audio` takes `format=wav|m4b|mp3` and `chapters=none|auto`; `auto` writes one chapter per story item titled from its first sentence (CJK aware, multi-track items deduped), so the export navigates like a real audiobook in Apple Books and Plex. `GET /history/{id}/export-audio?format=mp3` transcodes a single generation. The Stories "Export Audio" button is now a WAV / MP3 / Audiobook (M4B) menu and the history row menu gains "Export Audio (MP3)". WAV stays the default everywhere, so existing callers get the same bytes. Encoding shells out to ffmpeg (Homebrew, MacPorts and WinGet paths are searched even when the sidecar's PATH is minimal; `VOICEBOX_FFMPEG_DIR` overrides) and a missing ffmpeg returns a 503 with install instructions instead of an empty download. — @suifeng156, @joneldiablo ([#787](https://github.com/jamiepine/voicebox/pull/787), MP3 export from [#1048](https://github.com/jamiepine/voicebox/pull/1048); addresses [#165](https://github.com/jamiepine/voicebox/issues/165), [#270](https://github.com/jamiepine/voicebox/issues/270))+### Engines++- **OmniVoice TTS engine (non-commercial).** Adds k2-fsa/OmniVoice as an opt-in eighth engine: zero-shot cloning and attribute-based voice design (gender, age, pitch, accent) across the 23 languages Voicebox already lists, 24 kHz output. Its `HiggsAudioV2` codec only exists in transformers 5.x, so the modelling code is vendored under `backend/vendor/` and grafted onto the pinned 4.57 at load time; `omnivoice` installs `--no-deps` like Chatterbox and TADA. The weights are CC-BY-NC, so the engine is labelled non-commercial in the engine picker and model manager and is never a default. — @pippo73 ([#1062](https://github.com/jamiepine/voicebox/pull/1062); fixes [#380](https://github.com/jamiepine/voicebox/issues/380), [#517](https://github.com/jamiepine/voicebox/issues/517), [#791](https://github.com/jamiepine/voicebox/issues/791))+## [0.6.0] - 2026-10-04**The stability release.** 0.5 made Voicebox a dictation and agent-voice studio; 0.6 makes it hold up. Sixty-plus contributor pull requests landed since 0.5 and most of them fix something that used to go wrong in daily use: MLX crashed with `There is no Stream(gpu, 1) in current thread` when a model was loaded on one thread and used on another, memory grew with every generation and never came back on unload, a fully-cached model retried HuggingFace five times per file before it would load offline, the Docker image could not write its own data volume, dictation refused to arm without an LLM it was never going to call, and the backend test suite would not even collect. All of that is fixed. The headline features ride on top: native AMD ROCm on Windows and in Docker, five new interface languages, and Chatterbox multilingual running on Apple Silicon through MLX.diff --cc app/src/components/CapturesTab/CapturesTab.tsxindex b387f6a5,cdf50384..401b8a3c--- a/app/src/components/CapturesTab/CapturesTab.tsx+++ b/app/src/components/CapturesTab/CapturesTab.tsx@@@ -268,8 -268,8 +268,8 @@@ export function CapturesTab()// profile's stored engine preference. Cloned profiles without an// override fall through to whatever the backend picks.const engine = voice.default_engine as- | 'qwen' | 'qwen_custom_voice' | 'luxtts' | 'chatterbox'+ | 'qwen' | 'qwen_custom_voice' | 'qwen_voice_design' | 'luxtts' | 'chatterbox'- | 'chatterbox_turbo' | 'tada' | 'kokoro'+ | 'chatterbox_turbo' | 'tada' | 'kokoro' | 'omnivoice'| undefined;return apiClient.generateSpeech({profile_id: voice.id,diff --cc app/src/components/Generation/EngineModelSelector.tsxindex 2bd91fbf,a4159831..bdf6dd9d--- a/app/src/components/Generation/EngineModelSelector.tsx+++ b/app/src/components/Generation/EngineModelSelector.tsx@@@ -45,15 -45,21 +47,28 @@@ const ENGINE_DESCRIPTIONS: Record<strinconst 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',+ 'omnivoice',+]);+ /** Engines that synthesise a voice from a written description. */+ export const DESIGN_ENGINES = new Set(['qwen_voice_design']);++ /**+ * Engines that actually honor the instruct kwarg at model level. Base+ * Qwen3-TTS accepts it but ignores it, so it is deliberately absent.+ */+ const INSTRUCT_ENGINES = new Set(['qwen_custom_voice', 'qwen_voice_design']);++ export function engineSupportsInstruct(engine?: string): boolean {+ return !!engine && INSTRUCT_ENGINES.has(engine);+ }+function getAvailableOptions(selectedProfile?: VoiceProfileResponse | null) {if (!selectedProfile) return ENGINE_OPTIONS;return ENGINE_OPTIONS.filter((opt) => isProfileCompatibleWithEngine(selectedProfile, opt.engine));diff --cc app/src/components/Generation/FloatingGenerateBox.tsxindex bbf4e903,d28eb769..5f3daaa4--- a/app/src/components/Generation/FloatingGenerateBox.tsx+++ b/app/src/components/Generation/FloatingGenerateBox.tsx@@@ -152,8 -151,8 +152,9 @@@ export function FloatingGenerateBox(| 'chatterbox_turbo'| 'tada'| 'kokoro'- | 'omnivoice'- | 'qwen_custom_voice';+ | 'qwen_custom_voice'- | 'qwen_voice_design';++ | 'qwen_voice_design'++ | 'omnivoice';useEffect(() => {if (selectedProfile?.language) {form.setValue('language', selectedProfile.language as LanguageCode);diff --cc app/src/lib/constants/engines.tsindex 36b3219a,00000000..22dd3fb5mode 100644,000000..100644--- a/app/src/lib/constants/engines.ts+++ b/app/src/lib/constants/engines.ts@@@ -1,18 -1,0 +1,22 @@@+/** Engine capability sets shared across the generation UI and its form hook. */++/**+ * Engines whose model actually honours the `instruct` field.+ *+ * Base Qwen3-TTS accepts the kwarg and silently ignores it, so it stays out.+ *+ * Note the two members read the field very differently: Qwen CustomVoice takes+ * free-form prose ("speak in an angry tone"), while OmniVoice takes a closed+ * vocabulary of attributes ("female, low pitch, british accent") and raises on+ * anything outside it.+ */- export const INSTRUCT_ENGINES: ReadonlySet<string> = new Set(['qwen_custom_voice', 'omnivoice']);++export const INSTRUCT_ENGINES: ReadonlySet<string> = new Set([++ 'qwen_custom_voice',++ 'qwen_voice_design',++ 'omnivoice',++]);++/** Whether the given engine should show and forward the instruct field. */+export function engineSupportsInstruct(engine: string | undefined): boolean {+ return engine !== undefined && INSTRUCT_ENGINES.has(engine);+}diff --cc app/src/lib/constants/languages.tsindex 3c1420f6,61626bf9..c075892e--- a/app/src/lib/constants/languages.ts+++ b/app/src/lib/constants/languages.ts@@@ -70,33 -70,7 +70,34 @@@ export const ENGINE_LANGUAGES: Record<stada: ['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'],+ qwen_voice_design: ['zh', 'en', 'ja', 'ko', 'de', 'fr', 'ru', 'pt', 'es', 'it'],+ // OmniVoice covers 600+ languages. Listed here are the ones ALL_LANGUAGES+ // already knows -- widening that list is a separate change.+ omnivoice: [+ 'ar',+ 'da',+ 'de',+ 'el',+ 'en',+ 'es',+ 'fi',+ 'fr',+ 'he',+ 'hi',+ 'it',+ 'ja',+ 'ko',+ 'ms',+ 'nl',+ 'no',+ 'pl',+ 'pt',+ 'ru',+ 'sv',+ 'sw',+ 'tr',+ 'zh',+ ],} as const;/** Helper: get language options for a given engine. */diff --cc app/src/lib/hooks/useGenerationForm.tsindex 110451ed,1f0219fd..551a2d06--- a/app/src/lib/hooks/useGenerationForm.ts+++ b/app/src/lib/hooks/useGenerationForm.ts@@@ -102,11 -101,11 +103,13 @@@ export function useGenerationForm(optio: 'tada-1b': engine === 'kokoro'? 'kokoro'- : engine === 'omnivoice'- ? 'omnivoice'- : engine === 'qwen_custom_voice'- ? `qwen-custom-voice-${data.modelSize}`- : `qwen-tts-${data.modelSize}`;+ : engine === 'qwen_voice_design'+ ? 'qwen-voice-design-1.7B'- : engine === 'qwen_custom_voice'- ? `qwen-custom-voice-${data.modelSize}`- : `qwen-tts-${data.modelSize}`;++ : engine === 'omnivoice'++ ? 'omnivoice'++ : engine === 'qwen_custom_voice'++ ? `qwen-custom-voice-${data.modelSize}`++ : `qwen-tts-${data.modelSize}`;const displayName =engine === 'luxtts'? 'LuxTTS'@@@ -120,15 -119,15 +123,17 @@@: 'TADA 1B': engine === 'kokoro'? 'Kokoro 82M'- : engine === 'omnivoice'- ? 'OmniVoice'- : engine === 'qwen_custom_voice'- ? data.modelSize === '1.7B'- ? 'Qwen CustomVoice 1.7B'- : 'Qwen CustomVoice 0.6B'- : data.modelSize === '1.7B'- ? 'Qwen TTS 1.7B'- : 'Qwen TTS 0.6B';+ : engine === 'qwen_voice_design'+ ? 'Qwen VoiceDesign 1.7B'- : engine === 'qwen_custom_voice'- ? data.modelSize === '1.7B'- ? 'Qwen CustomVoice 1.7B'- : 'Qwen CustomVoice 0.6B'- : data.modelSize === '1.7B'- ? 'Qwen TTS 1.7B'- : 'Qwen TTS 0.6B';++ : engine === 'omnivoice'++ ? 'OmniVoice'++ : engine === 'qwen_custom_voice'++ ? data.modelSize === '1.7B'++ ? 'Qwen CustomVoice 1.7B'++ : 'Qwen CustomVoice 0.6B'++ : data.modelSize === '1.7B'++ ? 'Qwen TTS 1.7B'++ : 'Qwen TTS 0.6B';// Check if model needs downloadingtry {diff --cc backend/models.pyindex cfa61fd7,3ce8db23..409a0c96--- a/backend/models.py+++ b/backend/models.py@@@ -85,7 -85,7 +85,7 @@@ class GenerationRequest(BaseModel)seed: Optional[int] = Field(None, ge=0)model_size: Optional[str] = Field(default="1.7B", pattern="^(1\\.7B|0\\.6B|1B|3B)$")instruct: Optional[str] = Field(None, max_length=500)- engine: Optional[str] = Field(default=None, pattern="^(qwen|qwen_custom_voice|luxtts|chatterbox|chatterbox_turbo|tada|kokoro|omnivoice)$")- engine: Optional[str] = Field(default=None, pattern="^(qwen|qwen_custom_voice|qwen_voice_design|luxtts|chatterbox|chatterbox_turbo|tada|kokoro)$")++ engine: Optional[str] = Field(default=None, pattern="^(qwen|qwen_custom_voice|qwen_voice_design|luxtts|chatterbox|chatterbox_turbo|tada|kokoro|omnivoice)$")personality: bool = Field(default=False,description="When true and the profile has a personality prompt, the input text is rewritten in-character before TTS.",@@@ -352,7 -352,7 +352,7 @@@ class MCPClientBindingResponse(BaseModeprofile_id: Optional[str] = Nonedefault_engine: Optional[str] = Field(None,- pattern="^(qwen|qwen_custom_voice|luxtts|chatterbox|chatterbox_turbo|tada|kokoro|omnivoice)$",- pattern="^(qwen|qwen_custom_voice|qwen_voice_design|luxtts|chatterbox|chatterbox_turbo|tada|kokoro)$",++ pattern="^(qwen|qwen_custom_voice|qwen_voice_design|luxtts|chatterbox|chatterbox_turbo|tada|kokoro|omnivoice)$",)default_personality: bool = Falselast_seen_at: Optional[datetime] = None@@@ -371,7 -371,7 +371,7 @@@ class MCPClientBindingUpsert(BaseModel)profile_id: Optional[str] = Nonedefault_engine: Optional[str] = Field(None,- pattern="^(qwen|qwen_custom_voice|luxtts|chatterbox|chatterbox_turbo|tada|kokoro|omnivoice)$",- pattern="^(qwen|qwen_custom_voice|qwen_voice_design|luxtts|chatterbox|chatterbox_turbo|tada|kokoro)$",++ pattern="^(qwen|qwen_custom_voice|qwen_voice_design|luxtts|chatterbox|chatterbox_turbo|tada|kokoro|omnivoice)$",)default_personality: bool = False@@@ -390,7 -390,7 +390,7 @@@ class SpeakRequest(BaseModel))engine: Optional[str] = Field(None,- pattern="^(qwen|qwen_custom_voice|luxtts|chatterbox|chatterbox_turbo|tada|kokoro|omnivoice)$",- pattern="^(qwen|qwen_custom_voice|qwen_voice_design|luxtts|chatterbox|chatterbox_turbo|tada|kokoro)$",++ pattern="^(qwen|qwen_custom_voice|qwen_voice_design|luxtts|chatterbox|chatterbox_turbo|tada|kokoro|omnivoice)$",)personality: Optional[bool] = Field(None,diff --cc backend/services/profiles.pyindex 7f300ee3,2470c638..22bdf009--- a/backend/services/profiles.py+++ b/backend/services/profiles.py@@@ -32,8 -32,13 +32,13 @@@ from . import historlogger = logging.getLogger(__name__)-CLONING_ENGINES = {"qwen", "luxtts", "chatterbox", "chatterbox_turbo", "tada"}+CLONING_ENGINES = {"qwen", "luxtts", "chatterbox", "chatterbox_turbo", "tada", "omnivoice"}+ # Engines that synthesise a voice from a natural-language description+ # (voice_type "designed") rather than from samples or a preset id.+ DEFAULT_DESIGN_ENGINE = "qwen_voice_design"+ DESIGN_ENGINES = {DEFAULT_DESIGN_ENGINE}+def _profile_to_response(profile: DBVoiceProfile,diff --cc backend/voicebox-server.specindex d360fbed,ab0b03b7..8c4c8033--- a/backend/voicebox-server.spec+++ b/backend/voicebox-server.spec@@@ -5,7 -5,7 +5,7 @@@ from PyInstaller.utils.hooks import copdatas = []binaries = []- hiddenimports = ['backend', 'backend.main', 'backend.config', 'backend.database', 'backend.models', 'backend.services.profiles', 'backend.services.history', 'backend.services.tts', 'backend.services.transcribe', 'backend.utils.platform_detect', 'backend.backends', 'backend.backends.pytorch_backend', 'backend.backends.qwen_custom_voice_backend', 'backend.utils.audio', 'backend.utils.cache', 'backend.utils.progress', 'backend.utils.hf_progress', 'backend.services.cuda', 'backend.services.effects', 'backend.utils.effects', 'backend.services.versions', 'pedalboard', 'chatterbox', 'chatterbox.tts_turbo', 'chatterbox.mtl_tts', 'backend.backends.chatterbox_backend', 'backend.backends.chatterbox_turbo_backend', 'backend.backends.luxtts_backend', 'zipvoice', 'zipvoice.luxvoice', 'torch', 'transformers', 'fastapi', 'uvicorn', 'sqlalchemy', 'soundfile', 'qwen_tts', 'qwen_tts.inference', 'qwen_tts.inference.qwen3_tts_model', 'qwen_tts.inference.qwen3_tts_tokenizer', 'qwen_tts.core', 'qwen_tts.cli', 'requests', 'pkg_resources.extern', 'backend.backends.hume_backend', 'tada', 'tada.modules', 'tada.modules.tada', 'tada.modules.encoder', 'tada.modules.decoder', 'tada.modules.aligner', 'tada.modules.acoustic_spkr_verf', 'tada.nn', 'tada.nn.vibevoice', 'tada.utils', 'tada.utils.gray_code', 'tada.utils.text', 'backend.utils.dac_shim', 'torchaudio', 'backend.backends.kokoro_backend', 'backend.backends.omnivoice_backend', 'backend.utils.transformers5_compat', 'omnivoice', 'omnivoice.models.omnivoice', 'omnivoice.utils.audio', 'omnivoice.utils.text', 'omnivoice.utils.voice_design', 'omnivoice.utils.duration', 'omnivoice.utils.lang_map', 'pydub', 'en_core_web_sm', 'loguru', 'backend.mcp_server', 'backend.mcp_server.server', 'backend.mcp_server.tools', 'backend.mcp_server.context', 'backend.mcp_server.resolve', 'backend.mcp_server.events', 'sse_starlette', 'backend.backends.mlx_backend', 'mlx', 'mlx.core', 'mlx.nn', 'mlx_audio', 'mlx_audio.tts', 'mlx_audio.stt', 'mlx_lm', 'backend.backends.qwen_llm_backend', 'backend.backends.chatterbox_mlx_backend']-hiddenimports = ['backend', 'backend.main', 'backend.config', 'backend.database', 'backend.models', 'backend.services.profiles', 'backend.services.history', 'backend.services.tts', 'backend.services.transcribe', 'backend.utils.platform_detect', 'backend.backends', 'backend.backends.pytorch_backend', 'backend.backends.qwen_custom_voice_backend', 'backend.backends.qwen_voice_design_backend', 'backend.utils.audio', 'backend.utils.cache', 'backend.utils.progress', 'backend.utils.hf_progress', 'backend.services.cuda', 'backend.services.effects', 'backend.utils.effects', 'backend.services.versions', 'pedalboard', 'chatterbox', 'chatterbox.tts_turbo', 'chatterbox.mtl_tts', 'backend.backends.chatterbox_backend', 'backend.backends.chatterbox_turbo_backend', 'backend.backends.luxtts_backend', 'zipvoice', 'zipvoice.luxvoice', 'torch', 'transformers', 'fastapi', 'uvicorn', 'sqlalchemy', 'soundfile', 'qwen_tts', 'qwen_tts.inference', 'qwen_tts.inference.qwen3_tts_model', 'qwen_tts.inference.qwen3_tts_tokenizer', 'qwen_tts.core', 'qwen_tts.cli', 'requests', 'pkg_resources.extern', 'backend.backends.hume_backend', 'tada', 'tada.modules', 'tada.modules.tada', 'tada.modules.encoder', 'tada.modules.decoder', 'tada.modules.aligner', 'tada.modules.acoustic_spkr_verf', 'tada.nn', 'tada.nn.vibevoice', 'tada.utils', 'tada.utils.gray_code', 'tada.utils.text', 'backend.utils.dac_shim', 'torchaudio', 'backend.backends.kokoro_backend', 'en_core_web_sm', 'loguru', 'backend.mcp_server', 'backend.mcp_server.server', 'backend.mcp_server.tools', 'backend.mcp_server.context', 'backend.mcp_server.resolve', 'backend.mcp_server.events', 'sse_starlette', 'backend.backends.mlx_backend', 'mlx', 'mlx.core', 'mlx.nn', 'mlx_audio', 'mlx_audio.tts', 'mlx_audio.stt', 'mlx_lm', 'backend.backends.qwen_llm_backend', 'backend.backends.chatterbox_mlx_backend']++hiddenimports = ['backend', 'backend.main', 'backend.config', 'backend.database', 'backend.models', 'backend.services.profiles', 'backend.services.history', 'backend.services.tts', 'backend.services.transcribe', 'backend.utils.platform_detect', 'backend.backends', 'backend.backends.pytorch_backend', 'backend.backends.qwen_custom_voice_backend', 'backend.backends.qwen_voice_design_backend', 'backend.utils.audio', 'backend.utils.cache', 'backend.utils.progress', 'backend.utils.hf_progress', 'backend.services.cuda', 'backend.services.effects', 'backend.utils.effects', 'backend.services.versions', 'pedalboard', 'chatterbox', 'chatterbox.tts_turbo', 'chatterbox.mtl_tts', 'backend.backends.chatterbox_backend', 'backend.backends.chatterbox_turbo_backend', 'backend.backends.luxtts_backend', 'zipvoice', 'zipvoice.luxvoice', 'torch', 'transformers', 'fastapi', 'uvicorn', 'sqlalchemy', 'soundfile', 'qwen_tts', 'qwen_tts.inference', 'qwen_tts.inference.qwen3_tts_model', 'qwen_tts.inference.qwen3_tts_tokenizer', 'qwen_tts.core', 'qwen_tts.cli', 'requests', 'pkg_resources.extern', 'backend.backends.hume_backend', 'tada', 'tada.modules', 'tada.modules.tada', 'tada.modules.encoder', 'tada.modules.decoder', 'tada.modules.aligner', 'tada.modules.acoustic_spkr_verf', 'tada.nn', 'tada.nn.vibevoice', 'tada.utils', 'tada.utils.gray_code', 'tada.utils.text', 'backend.utils.dac_shim', 'torchaudio', 'backend.backends.kokoro_backend', 'en_core_web_sm', 'loguru', 'backend.mcp_server', 'backend.mcp_server.server', 'backend.mcp_server.tools', 'backend.mcp_server.context', 'backend.mcp_server.resolve', 'backend.mcp_server.events', 'sse_starlette', 'backend.backends.mlx_backend', 'mlx', 'mlx.core', 'mlx.nn', 'mlx_audio', 'mlx_audio.tts', 'mlx_audio.stt', 'mlx_lm', 'backend.backends.qwen_llm_backend', 'backend.backends.chatterbox_mlx_backend', 'backend.backends.omnivoice_backend', 'backend.utils.transformers5_compat', 'omnivoice', 'omnivoice.models.omnivoice', 'omnivoice.utils.audio', 'omnivoice.utils.text', 'omnivoice.utils.voice_design', 'omnivoice.utils.duration', 'omnivoice.utils.lang_map', 'pydub']datas += copy_metadata('qwen-tts')datas += copy_metadata('requests')datas += copy_metadata('transformers')