SUZU
SUZU
SUZU
SUZUv0.9

A real native Mac app.

Suzu works seamlessly across your system and runs your terminal agents exactly as you expect. Audio and history stay on your Mac.

Neural on-device voices
Hands-free Suzu Ask
60-language translation
Works with any agent
Smart audio ducking
Menu bar quota gauges
Animated mascot
Local-first privacy

Bring your own agents. Claude Code, Pi, Codex, Cursor, Gemini, or custom scripts: Suzu drives the tools you already run in your terminal, and anything with a CLI hooks right in.

On-device neural voice. Powered by local MLX and Apple Metal. Natural speech with zero latency, zero cloud subscriptions, and no audio data sent over the wire.

Smart audio ducking. Automatically dips background music, podcasts, or videos when your agent starts speaking, and smoothly restores volume the moment it finishes.

Suzu Ask for instant decisions. Talk back to your running agent with local speech-to-text. Tap your hotkey, state your choice out loud, and keep your hands on the keyboard.

Zero cloud telemetry. Your source code, logs, session outputs, and spoken prompts never leave your Mac. Everything runs 100% locally through compiled Rust and Swift.

Multi-agent orchestration. Track multiple agents working simultaneously across workspaces. Suzu manages voice queues cleanly so overlapping tasks never speak over each other.

Mascot, Orb, or Ambient. Choose the companion style that fits your flow: an animated Rive mascot, a minimal glowing orb, or discreet click-through ambient corner light.

Auto-configured hooks. One command equips your environment rules and notification hooks across Claude, Pi, Codex, and Cursor with zero manual setup.

Universal CLI bridge. High performance private Unix socket with dynamic port fallback. Trigger vocal alerts, questions, and decisions from any script or CI job.

Real-time quota monitor. Live token counters and rate limit tracking across 16 API providers right in your menu bar popover, before you hit sudden limits.

Never steals your focus. Suzu displays surfaces and speaks alerts without ever taking active window focus. Your cursor stays in your editor and Escape always dismisses.

Native, and only a few MB. No Electron, no bundled browser: a genuine Swift app with a Rust core that launches in milliseconds and barely touches your memory or battery.

30 languages spoken. 60 translated.

Two MLX speech engines, nine transcription engines and a local translation model. Everything runs on your Mac, nothing leaves it.

Text to Speech

2 engines · 30 languages

Qwen3-TTS

10 languages
Chinese (Mandarin)
English
French
German
Italian
Japanese
Korean
Portuguese
Russian
Spanish

Native English accents (British, American, Regional) and 9 Mandarin dialects, incl. Cantonese, Sichuanese, Hokkien & Wu.

VoxCPM2

+20 languages
Arabic
Burmese
Danish
Dutch
Finnish
Greek
Hebrew
Hindi
Indonesian
Khmer
Lao
Malay
Norwegian
Polish
Swedish
Swahili
Tagalog
Thai
Turkish
Vietnamese

A superset of Qwen3-TTS: it also speaks all 10 of its languages, 30 in total.

Safety net: Apple Speech follows the voices installed on your Mac, and 6 more languages (Bengali, Czech, Persian, Romanian, Ukrainian, Urdu) currently ride on system voices.

Speech to Text

9 engines · up to 99 languages

Qwen3-ASR

17 languages · live
Chinese
Czech
English
French
German
Greek
Italian
Japanese
Korean
Persian
Portuguese
Romanian
Russian
Spanish
Swedish
Thai
Vietnamese

MLX on the GPU: certifies 17 languages, auto-detects the one you speak and streams partial text while you talk.

Whisper Large v3 & Turbo

99 languages
Arabic
Chinese
English
French
German
Hindi
Italian
Japanese
Korean
Portuguese
Russian
Spanish+87 more

WhisperKit on Core ML, automatic language detection. Large v3 for maximum accuracy, Turbo to run faster on smaller machines.

Whisper Distil Large v3

English only
English

Trades the 99 languages for speed when English is all you need.

Parakeet TDT v3

28 languages
Belarusian
Bosnian
Bulgarian
Croatian
Czech
Danish
Dutch
English
Estonian
Finnish
French
German
Greek
Hungarian
Italian
Latvian
Lithuanian
Maltese
Polish
Portuguese
Romanian
Russian
Serbian
Slovak
Slovenian
Spanish
Swedish
Ukrainian

FluidAudio int8 on the Neural Engine: blazing fast full-paragraph transcription.

Parakeet TDT v2 & Flash

English only
English

Flash is the lightest model of the lineup, from 240 MB.

Nemotron 3.5 Streaming

10 languages · live
Chinese
English
French
German
Italian
Japanese
Korean
Portuguese
Russian
Spanish

Live partials, certified language by language.

Apple Speech

system voices
Follows your Mac's dictation languages

Translation

LMT-60 · 60 languages

LMT-60 0.6B

on-device

NiuTrans LMT-60, Apache 2.0, custom 8-bit MLX conversion. Runs fully offline.

Amharic
Arabic
Armenian
Azerbaijani
Bengali
Bulgarian
Burmese
Chinese
Chinese (Traditional)
Croatian
Czech
Danish
Dutch
English
Finnish
French
Georgian
German
Greek
Hebrew
Hindi
Hungarian
Icelandic
Indonesian
Italian
Japanese
Javanese
Kazakh
Khmer
Korean
Kyrgyz
Lao
Malay
Marathi
Nepali
Norwegian
Pashto
Persian
Polish
Portuguese
Portuguese (Brazil)
Romanian
Russian
Sinhala
Slovak
Spanish
Swedish
Swahili
Tagalog
Tajik
Tamil
Telugu
Thai
Tibetan
Turkish
Ukrainian
Urdu
Uyghur
Uzbek
Vietnamese

Built for modern developers

Everything you need to keep your eyes on your code while staying informed.

Premium voices

Qwen3-TTS neural voices across 10 languages, with native accents and dialects.

Audio effects & ducking

Automatically ducks IDE and music volume when your agent starts speaking.

Sound notification hooks

Pre-message audio chimes mapping to done, error, or question states.

Simplified integration

Just copy-paste a line of context instructions into your agent's system prompt.

Suivi des quotas

Auto-detects token usage and limits across 16 different API providers.

Universal CLI bridge

Trigger vocal alerts via terminal from any tool, environment or language.

Animated mascot

Pip displays the agent’s execution status via native Rive animations.

Blazing fast native app

Built in Swift/SwiftUI. Zero cloud telemetry. Low RAM and battery footprint.

By the numbers

Designed for speed, privacy, and seamless pair programming with AI.

SUZU
SUZU
Prompts Spoken

Voice alerts delivered live.

Focus Saved

Cmd+Tab switches avoided.

Smart Audio Ducking

Spotify & Apple Music volume ducked smoothly when agents speak.

AUTO_FADE_IN_OUT
Token Overhead
~

Consumes less tokens than a standard 3-line system prompt.

Local Voices

Offline neural voices.

IPC Bridge Latency
<

Socket Unix IPC bridge connects terminal agents to Pip instantly.

Local Privacy
Offline

Zero cloud telemetry. 100% local macOS processing.

Card Stack

Interactive stacked cards auto-flipping every 5 seconds.

These cards are amazing, I want to use them in my project. Framer motion is a godsend ngl tbh fam 🙏

Manu Arora

Senior Software Engineer

I dont like this Twitter thing, deleting it right away because yolo. Instead, I would like to call it X.com so that it can easily be confused with adult sites.

Elon Musk

Senior Shitposter

The first rule ofFight Club is that you do not talk about fight club. The second rule ofFight club is that you DO NOT TALK about fight club.

Tyler Durden

Manager Project Mayhem

Hands-Free Interaction

SUZU ASK

Talk back to your AI agents with seamless Speech-to-Text. When Pip asks a question or requests a decision, press your hotkey, speak naturally, and Suzu transcribes your voice instantly to your running terminal agent.

Listening to your voice...WhisperKit · on-device
"Proceed with option 2 and deploy to staging."
Latency: <120ms100% On-device
Coming Soon

SUZU CLOUD & MOBILE

Voice notifications anywhere. Step away from your desk while your agent runs long benchmarks or builds. Get real-time audio alerts pushed straight to your iPhone, Apple Watch, or Android device through Suzu Cloud.

Arguments de vente

  • Peu coûteux en tokens
  • Environ 300 tokens au départ
  • 20 tokens par cycle de dev
  • Ducking audio quand on écoute de la zik pour entendre le message
  • Silencieux pendant les appels visio
  • Créer ses propres voix sur mesure

Share your feedback

Have questions, feature requests, or found a bug? We'd love to hear from you. Try the demo below!

Scroll down

Going Zero to One

If you're navigating a new business... breaking into a new market.

Scaling from One to N

If you've achieved Product/Market Fit...

Need Quick Solutions

If you know exactly what you want and need...

So cool, right?

Testimonials

What our users say

See what developers and creators have to say about Suzu.

Suzu has completely changed my developer workflow. Having Pip speak build status aloud keeps me 100% focused on code.
Briana Patton
Briana Patton
Senior AI Engineer
The local voice setup and audio ducking when music plays is pure genius. Essential menu bar tool for any solo dev.
Bilal Ahmed
Bilal Ahmed
Fullstack Developer
Zero cloud telemetry and native Swift performance. It's rare to see developer tools designed with such attention to detail.
Saman Malik
Saman Malik
Lead Platform Architect
Suzu has completely changed my developer workflow. Having Pip speak build status aloud keeps me 100% focused on code.
Briana Patton
Briana Patton
Senior AI Engineer
The local voice setup and audio ducking when music plays is pure genius. Essential menu bar tool for any solo dev.
Bilal Ahmed
Bilal Ahmed
Fullstack Developer
Zero cloud telemetry and native Swift performance. It's rare to see developer tools designed with such attention to detail.
Saman Malik
Saman Malik
Lead Platform Architect