Languages
7 languages
EnglishDutchFrenchGermanItalianPortugueseSpanish
Each version speaks one language. Parakeet hears all seven, and Boubou asks Qwen to reply in that language.
pocket-tts
Pipeline 7 languages 27 voices Mac · Linux with NVIDIA RecommendedKyutai’s Pocket TTS on the CPU. Each version loads one language config, and every config ships the same 27 preset voices. Boubou uses each language’s default voice.
npx boubou@beta run pocket-tts:english_2026-10
Talk on this Mac. It needs Node.js 18 or later; Get started walks you through.
Talk on this Linux PC. It needs an NVIDIA GPU, and the first run also downloads a 3.62 GB engine that every pipeline voice shares; Get started walks you through.
npx boubou@beta serve pocket-tts:english_2026-10
Talk from a phone on the same Wi-Fi, or connect a robot or app. The model still runs on this Mac. Phones and robots walks you through.
Talk from a phone on the same Wi-Fi, or connect a robot or app. The model still runs on this PC. Phones and robots walks you through.
This version runs on Apple Silicon Macs only. On Linux, choose pocket-tts:english_2026-10.
- Download
- 5.80 GB5.48 GB
- Memory, estimated
- 12 GiB
- GPU memory, measured
- 6–8.5 GiB
- Speaks
- English
- Voice
- alba
Recommended device
Apple Silicon Mac
- 16 GB of memory or more
- macOS 15 or later
- Qwen3.5-4B, the language model Boubou maintains
- The voice runs on the CPU
Linux PC with an NVIDIA GPU
- NVIDIA driver 580.65.06 or later
- Tested on GPUs with 16 GB and 32 GB of memory
- x86-64, with glibc 2.39 or later, such as Ubuntu 24.04
- Qwen3.5-4B, the language model Boubou maintains, runs on llama.cpp
Memory needs are estimates. Boubou checks the machine before it downloads anything, and smaller machines still need testing.
Voices and languages
Everything the model can speak, as its own card or files list it. Black marks what Boubou uses.
- Boubou uses
- Available in the model
Voices
27 voices
Press a voice to hear it. Each clip is the pinned model reading one sentence in that voice’s language.
English voices
Default voices of the other languages
All 27 voices ship with every language config. Boubou uses the release without voice cloning, so every voice is a preset.
Versions 26
A version is one ready-to-run setup with its own command. Versions differ in size, language or the models they combine. Suggested ones also appear in the boubou menu.
| Version | What’s different | Speaks | Voice | Download | Mac | Linux |
|---|---|---|---|---|---|---|
| The default English config | English | alba | 5.80 GB | 16 GB+ | NVIDIA GPU | |
| Another English config | English | alba | 5.80 GB | — | NVIDIA GPU | |
| Another English config | English | alba | 5.80 GB | — | NVIDIA GPU | |
| Larger 24-layer English config | English | alba | 6.91 GB | — | NVIDIA GPU | |
| Another English config | English | alba | 5.80 GB | — | NVIDIA GPU | |
| Another English config | English | alba | 5.80 GB | — | NVIDIA GPU | |
| Larger 24-layer English config | English | alba | 6.91 GB | — | NVIDIA GPU | |
| Another English config | English | alba | 5.82 GB | — | NVIDIA GPU | |
| Another English config | English | alba | 5.69 GB | — | NVIDIA GPU | |
| Another English config | English | alba | 5.69 GB | — | NVIDIA GPU | |
| Another English config | English | alba | 5.65 GB | — | NVIDIA GPU | |
| Another English config | English | alba | 5.65 GB | — | NVIDIA GPU | |
| Another English config | English | alba | 5.63 GB | — | NVIDIA GPU | |
| Another English config | English | alba | 5.63 GB | — | NVIDIA GPU | |
| Dutch config | Dutch | daan | 5.80 GB | — | NVIDIA GPU | |
| Larger 24-layer Dutch config | Dutch | daan | 6.27 GB | — | NVIDIA GPU | |
| French config | French | estelle | 5.81 GB | — | NVIDIA GPU | |
| Larger 24-layer French config | French | estelle | 6.28 GB | — | NVIDIA GPU | |
| German config | German | juergen | 5.80 GB | — | NVIDIA GPU | |
| Larger 24-layer German config | German | juergen | 6.28 GB | — | NVIDIA GPU | |
| Italian config | Italian | giovanni | 5.80 GB | — | NVIDIA GPU | |
| Larger 24-layer Italian config | Italian | giovanni | 6.27 GB | — | NVIDIA GPU | |
| Portuguese config | Portuguese | rafael | 5.80 GB | — | NVIDIA GPU | |
| Larger 24-layer Portuguese config | Portuguese | rafael | 6.28 GB | — | NVIDIA GPU | |
| Spanish config | Spanish | lola | 5.80 GB | — | NVIDIA GPU | |
| Larger 24-layer Spanish config | Spanish | lola | 6.27 GB | — | NVIDIA GPU |
How it works
You speak
VAD
Silero + Smart Turn v3.2
--vad sileroSTT
Parakeet TDT 0.6B v3
speech to textLLM
Qwen3.5-4B · 4-bit
--llm_backend mlx-lmTTS
pocket-tts · alba
--tts pocketIt answers
Boubou owns the microphone and speakers. Apple voice processing on a Mac, or WebRTC AEC3 on Linux, cancels echo, so you can talk over a reply. Boubou drops the queued audio and cancels the generation.
Swap a component
Boubou maintains one Hugging Face speech-to-speech 1.1.0 pipeline and pins every model in it. Add an upstream option after the model to replace one part.
Change the language model to any MLX model on Hugging Face
npx boubou@beta run pocket-tts:english_2026-10 --model_name mlx-community/Llama-3.2-3B-Instruct-4bit
Change speech recognition to Whisper
npx boubou@beta run pocket-tts:english_2026-10 --stt mlx-audio-whisper
Options work with serve too. Boubou does not check the models you name: they download from Hugging Face into your own Hugging Face cache. API modules send audio or text to the server you configure. These two examples are for a Mac: on Linux the pipeline stays offline, so options must name local paths.
- Used by Boubou
- Also runs in Boubou
VADVoice activity
--vadHears when you start and stop talking. Smart Turn v3.2 then decides whether a pause ends your turn.
silero
Not in Boubou’s runtime yet: firered
STTSpeech to text
--sttWrites down what you said. Parakeet is the default; Whisper versions detect the language.
parakeet-tdtmlx-audio-whisperwhisperqwen3-asrvllm-realtimeAPIopenaiAPIopenai-realtimeAPI
Not in Boubou’s runtime yet: faster-whisper, whisper-mlx, parakeet-unified, nemotron-streaming, orukeet, paraformer, sense-voice
LLMLanguage model
--llm_backendWrites the reply. Boubou runs pinned 4-bit Qwen3.5 models with mlx-lm; --model_name picks another one.
mlx-lmtransformerschat-completionsAPIresponses-apiAPI
TTSText to speech
--ttsSpeaks the reply while it is still being written. Every pipeline voice in the library is one of these.
qwen3kokoropocketphononkittensupertonicchatTTSfacebookMMSomnivoicedrama3openaiAPIminimaxAPI
License
CC BY 4.0. Read the license. Weights keep their own licenses; check them before commercial use.