Platform documentation

Voice call guide

macOS Windows In development

Mac voice call translation

How to translate voice calls on Mac with TCVoiceAI

Translate two languages or transcribe a same-language call from your Mac. Suggestions work in translated conversations and in Same language sessions. Bilingual cards include the reply you can use and a translation for your partner.

One-sided setup: Install TCVoiceAI on your Mac only. Your partner joins the call normally, and the app translates both your speech and their audio from your side.

Watch the product demo View setup guide

Live translation Ready

macOS setup guide

Languages

🇺🇸 EN 🇪🇸 ES

macOS calling apps FaceTime · Zoom · Teams

Translated voice Both directions stay active

Partner setup Nothing to install

Translated voice · bilingual

Both directions stay active

Ready

Partner

Can you deliver two weeks earlier? Can you deliver two weeks earlier without increasing the budget?

¿Pueden entregar dos semanas antes? ¿Pueden entregar dos semanas antes sin aumentar el presupuesto?

You

Yes—without changing the agreed budget. Yes—we can bring the launch forward by two weeks while keeping the agreed budget.

Sí, sin cambiar el presupuesto acordado. Sí, podemos adelantar el lanzamiento dos semanas manteniendo el presupuesto acordado.

Suggested reply Speak to partner Private guidance

Prepare your session · Live session · Saved outputs

Setup steps

Translate an online call in four steps.

01
Open the app

Sign in on your Mac. A saved Conversation Profile is optional when you want to reuse a recurring setup.
02
Prepare your session

Use the defaults and start, or add only the context that helps.
03
Configure the calling app

Open Calling app setup in TCVoiceAI for app-specific audio steps. FaceTime can be configured automatically after Accessibility permission is granted.
04
From call to useful work

Create a result after the call, review its evidence, then continue with confirmed actions in Work Inbox.

Product details

Choose how the call sounds and what History captures.

Conversation Profiles for recurring calls

Save languages, mode, context, response timing, correction, vocabulary, and optional transcript translation. Each profile opens its own local History.

Sessions

Use the defaults and start, or add only the context that helps. Session settings are optional overrides and do not change the Profile.

Portable settings and profile transfer

In Settings > Data transfer, export one portable JSON file containing your Conversation Profiles and compatible app preferences. Import the same file on another Mac or Windows PC.

Translated voice mode

The outgoing path replaces your routed speech with translated Mac voice audio, while the incoming path speaks your partner's translation through your physical speaker or headphones.

Original audio mode

Both callers keep their real voices. History can hold a bilingual translation or a same-language transcript without spoken translated audio.

Same-language transcription

Turn on Same language and choose one Conversation language. Leave transcript translation off so transcription itself uses no translated words, or translate both transcripts into another language.

Automatic or On request suggestions

Automatic can generate a reply after a completed, meaningful partner turn. On request waits and shows Get suggestion under eligible partner messages in History and Floating captions.

Floating captions over the call

Keep up to four recent turns, bilingual text when available, and the current suggested reply above supported full-screen call apps. The complete conversation remains in local History.

Reply actions follow the audio mode

In Translated voice, review a suggestion and choose Speak to partner. In Original audio, keep the suggestion as private guidance while both real voices stay live.

Live mode switching

During a bilingual session, move between Translated voice and Original audio without ending the conversation. If routing cannot change safely, the app rolls back automatically.

Guided calling-app audio setup

Open installed calling apps from TCVoiceAI, follow exact microphone and speaker steps, or let the app configure FaceTime automatically with permission.

Saved outputs

Turn this session into a useful result you can reopen later.

Session notes & follow-up

Meeting notes · Decisions & action items · Follow-up email · CRM notes · Technical handoff

Evidence-backed review

Review decisions, actions, risks, and open questions against their linked transcript moments before using them.

Work Inbox

Confirmed actions move to Work Inbox, where you can set status, owner, due date, reminders, or start a follow-up Session.

Context and transcript correction

You can add a short context note during the call so terms, names, tone, and pronouns stay natural. With Correct likely transcript errors enabled, that note and recent turns can help resolve only obvious speech-recognition word errors.

Translation timing

Translated voice waits for Apple Speech to mark a result final by default. Faster responses can translate stable speech sooner in either direction, but longer sentences may be split. Original audio always waits for final speech results.

Independent call directions

The foreground app recognizes and translates you and your partner on separate paths. Both directions can make progress at the same time, while translation—and Translated voice playback—remain ordered within each direction.

Background audio continuity

When available, the installed helper passes normal audio through while the main app is idle or closed, so compatible calling apps can keep the virtual microphone and speaker selected.

Custom vocabulary for relevant terms

Add paired names, company terms, products, and industry language to a profile. They can guide recognition and provide a preferred translation when the matching term appears, but cannot guarantee a correction.

From the macOS app

The page reflects the actual app flow.

The app uses Apple Speech Recognition and macOS text-to-speech voices on the current Mac.
Import validates the file and asks for confirmation before replacing the current portable settings and profiles.
Importing settings and profiles does not import, delete, or merge local conversation History.
The versioned settings and profile format is designed for both macOS and Windows, so one export can move compatible setup between platforms.
Relevant custom vocabulary can help speech recognition and translation, but it cannot recover speech that was omitted or guarantee that a term will be recognized correctly.
Correct likely transcript errors is on by default. It can use the context note and recent turns only to interpret obvious recognition word errors. It does not rewrite the source transcript in History or recover omitted speech; uncertain text is sent for translation as recognized.
Translated voice marks its translation requests for speech playback, and the server instructs the translation model to use natural spoken phrasing and punctuation while returning translation only.
Each direction has its own ordered translation and playback queues. A phrase in one direction does not replace voice already playing in that direction, while the opposite direction can continue independently.
Translated speech is read aloud using Mac voices, so it may not sound like the original speaker’s personal voice.
Original audio mode keeps the real voices live and saves bilingual History without text-to-speech playback.
The Same language switch replaces the two participant-language pickers with one Conversation language picker and requires Original audio.
With transcript translation off, same-language transcription itself sends no translation request and consumes no translated-word allowance.
Suggestions still work in bilingual and same-language sessions. Generated primary-language reply words share the translated-word allowance at the active plan rate.
On request suggestion classification can expose Get suggestion in History and Floating captions without generating or charging for a reply first.
Floating captions reuse the current session’s History and suggestion state; they do not start another audio capture, recognition task, translation job, or polling loop.
A suggestion used with the same meaning can be marked Suggested reply used and is not charged a second time.
History and CSV export omit empty translation sections for transcription-only turns.
A bilingual session can switch between Translated voice and Original audio without stopping; failed route changes roll back automatically.
Result creation is manual by default. You can explicitly enable one reviewable result after Stop; nothing is published or sent automatically.
One AI output unit covers up to 5,000 source conversation words; each additional started block of 5,000 words uses another unit.
Summaries and saved work results share this allowance.
Your partner does not need to install TCVoiceAI; they can join the call normally.
TCVoiceAI Mic sends translated local speech into the calling app.
TCVoiceAI Speaker captures partner audio from the calling app for translation.
When the main app is idle or closed, the background helper only relays original audio between the selected physical and virtual devices.
The background helper does not recognize speech, create transcripts, translate, record, store, inspect, or upload conversation content, and it makes no server requests.
The calling app or browser call must allow the selected virtual microphone and virtual speaker/output devices.
The server processes translation text, bounded suggestion context, and user-requested AI conversation output for the requested response without retaining audio, transcripts, prompts, generated replies, or v1 plain output; OpenAI requests use storage disabled (store: false). When structured Outcome recovery is enabled, only the generated Outcome is kept as encrypted replay for up to 24 hours by default and deleted early after the app acknowledges saving it.

Related guides

Back to homepage

Sessions