AI TranslatorDownload
Menu-bar app for macOS

Speak your language.
They hear theirs.

AI Translator translates both sides of a Zoom, Meet, Teams or FaceTime call while it happens. It installs its own speaker and microphone, so your call app needs no plugin — you just pick them.

Download free for macOS

macOS 14 or newer · Signed & notarized · No BlackHole or Loopback

2
directions translated at the same time
29
languages with Gemini or on-device
3
engines: OpenAI, Gemini, or on your Mac
How it works

It sits between your call app and your headphones

Your call app plays the other side into AI Translator Speaker and takes your voice from AI Translator Microphone. The app translates in between, in both directions at once.

How AI Translator routes audio Incoming: the call app sends the other person's voice through the AI Translator Speaker virtual device into the app, which plays it to your headphones in your language. Outgoing: your microphone goes through the app and the AI Translator Microphone virtual device back to the call app in their language. 1 The other person speaks — you hear your language their voiceyour languageCall appZoom · Meet · FaceTimeAI Translator Speakervirtual deviceAI Translatortwo live sessionsYouyour headphones2 You speak — they hear their language your voicetheir languageYouyour microphoneAI Translatortwo live sessionsAI Translator Microphonevirtual deviceCall appZoom · Meet · FaceTime

1 The other person speaks — you hear your language

  1. Call appZoom · Meet · FaceTime
  2. their voiceAI Translator Speakervirtual device
  3. AI Translatortwo live sessions
  4. your languageYouyour headphones

2 You speak — they hear their language

  1. Youyour microphone
  2. your voiceAI Translatortwo live sessions
  3. AI Translator Microphonevirtual device
  4. their languageCall appZoom · Meet · FaceTime
  1. 01

    Install

    Open the installer. It adds the app and two audio devices, AI Translator Speaker and AI Translator Microphone. No restart, nothing else to download.

  2. 02

    Pick the devices

    In your call app, choose AI Translator Speaker as the speaker and AI Translator Microphone as the mic. Or turn on Autopilot: the app switches system audio on Start and restores it on Stop.

  3. 03

    Press Start

    Choose the two languages and press Start Translation (⌃⌥T from any app). Each side now hears the other in their own language, for the whole call.

Nobody installs a plugin. Nobody waits for their turn. You both just talk.

Features

Built to last a whole call

A translator has to hold up for an hour of real conversation, with whatever app the other side happens to use.

  • Both sides at once

    Two separate translation sessions run side by side, one per direction. If one stalls, the other keeps going.

  • The call app you already use

    Zoom, Google Meet, Microsoft Teams, FaceTime, WhatsApp, Discord, Slack and browser calls see an ordinary speaker and microphone. Nothing to install inside them.

  • Its own audio drivers

    The signed, notarized installer brings two audio drivers of its own. No BlackHole, no Loopback, no audio routing to set up by hand.

  • Live captions

    A floating window shows what each side said and its translation as it streams in. Captions live in memory only and are never written to disk.

  • Updates itself, carefully

    New versions download inside the app and install only after the checksum, the developer signature and Apple’s Gatekeeper all check out.

Engines

Three engines. Switch any time.

Two cloud engines run on your own API key. The third runs entirely on your Mac. Pick one in the menu-bar panel — each keeps its own settings.

CloudOpenAI RealtimeCloudGemini LiveOn-deviceOn-device
Runs onOpenAI’s serversGoogle’s serversYour Mac, with no network
Sounds likeThe speaker’s own voiceA natural generated voiceA macOS voice, or a downloaded one that imitates the speaker
SpeaksWhile the other side is still talkingWhile the other side is still talkingAfter each phrase ends
Translates into13 languages29 languages29 languages
CostsYour OpenAI key, billed per minuteYour Gemini key, billed by GoogleNothing
NeedsmacOS 14 or newermacOS 14 or newermacOS 26 or newer
Cloud

OpenAI Realtime

Runs on
OpenAI’s servers
Sounds like
The speaker’s own voice
Speaks
While the other side is still talking
Translates into
13 languages
Costs
Your OpenAI key, billed per minute
Needs
macOS 14 or newer
Cloud

Gemini Live

Runs on
Google’s servers
Sounds like
A natural generated voice
Speaks
While the other side is still talking
Translates into
29 languages
Costs
Your Gemini key, billed by Google
Needs
macOS 14 or newer
On-device

On-device

Runs on
Your Mac, with no network
Sounds like
A macOS voice, or a downloaded one that imitates the speaker
Speaks
After each phrase ends
Translates into
29 languages
Costs
Nothing
Needs
macOS 26 or newer

Inside the on-device engine

Three stages, all on your Mac. Speech recognition uses Apple’s SpeechAnalyzer, or a model you download that keeps English names inside other languages: Whisper large-v3-turbo (1.6 GB), Whisper small (0.5 GB) or Nemotron 3.5 Streaming (0.6 GB), which shows text while the speaker talks. Translation uses Apple’s built-in translator, or a model you download that reads each sentence in the context of the conversation: Hy-MT2 7B (4.2 GB) or TranslateGemma 4B (2.2 GB). Nothing downloads until you pick one and press Download.

The on-device pipeline: recognize, translate, speak audio inaudio outRecognizespeech → textSpeechAnalyzer— or a model you downloadTranslatetext → textApple Translation or Apple Intelligence— or a model you downloadSpeaktext → speechAVSpeechSynthesizerspeaks sentence by sentence
  1. Audio in
  2. Recognizespeech → textSpeechAnalyzer — or Whisper or Nemotron, downloaded on request
  3. Translatetext → textApple Translation or Apple Intelligence — or a model you download
  4. Speaktext → speechAVSpeechSynthesizer — or VoxCPM2 or Qwen3-TTS, downloaded on request
  5. Audio out
Privacy

Your conversations stay yours

A translator sits in the most private spot of a call. This one is built so you don’t have to take that on trust.

  • On-device means on-device

    With the on-device engine, speech recognition, translation and speech all run on your Mac. Audio and text never touch a network.

  • Keys stay in the Keychain

    Your OpenAI or Gemini key lives in the macOS Keychain — never in a settings file, a log or a crash report.

  • No server of ours in the audio path

    Cloud engines talk to OpenAI or Google directly, with your own key. Your calls never pass through us.

  • No analytics in the app

    No tracking, usage pings or crash uploads. The app only asks this site whether a new version exists. The site itself counts page views with cookie-less Cloudflare Web Analytics.

Requirements

What you need

macOS
14 Sonoma or newer
The on-device engine needs macOS 26
Mac
Apple Silicon
M1 or later; Intel Macs are not supported
Headphones
Wired works best
Bluetooth adds 100–300 ms
API key
For OpenAI or Gemini
The on-device engine needs none
FAQ

Questions, answered

Something missing? Write to siarhei@dudko.dev.

Do I need BlackHole or Loopback?

No. AI Translator ships its own two audio drivers inside the installer. There is no third-party virtual-audio software to install or configure.

Is it free?

Yes. The app and its license are free; the license lasts 6 months and renews free with the same email. The on-device engine costs nothing to run. OpenAI and Gemini bill your own API key for what you use.

Which apps does it work with?

Any macOS app with a microphone and speaker — Zoom, Google Meet, Microsoft Teams, FaceTime, WhatsApp, Discord, Slack, and browser-based calls. To them, AI Translator looks like an ordinary speaker and microphone.

Is my audio sent anywhere?

With the on-device engine, no — everything runs on your Mac. With a cloud engine, call audio goes straight to that provider's API for translation — OpenAI or Google — the same way any app built on those APIs works. It never passes through a server of ours.

Does it work offline?

With the on-device engine, yes, once its speech assets and any translation model or voice you picked are downloaded. The license is checked on your Mac, so it works offline too.

Which languages can it translate into?

OpenAI translates into 13 languages; Gemini and the on-device engine into 29. The on-device engine uses Apple’s translator, or a model you download — Hy-MT2 7B or TranslateGemma 4B — for more natural phrasing.

What is the license key for?

Registering your email issues a license key valid for 6 months. The app needs a valid key to start translating and to download updates. Renewing is free — register again with the same email.

How do updates work?

The app checks this site for new versions and tells you when one is out. Updates download inside the app and install only after the checksum, the developer signature and Apple’s Gatekeeper all check out.

What macOS version do I need?

macOS 14 Sonoma or newer on a Mac with Apple Silicon for the OpenAI and Gemini engines. The on-device engine needs macOS 26.

How do I uninstall it?

Open the menu-bar panel, click ⋯ and choose Uninstall AI Translator. It removes the app, both audio drivers, your logs and preferences. API keys stay in the Keychain until you delete them.

Get AI Translator free

Enter your email and we send a confirmation link. Open it to get the installer and your license key.

  • Free. No card, no payment details.
  • License for 6 months; renew free with the same email.
  • Uninstalls from the menu bar — app and drivers.

macOS 14 or newer · Signed & notarized