What LiveCaption is, and which mode suits whom

LiveCaption is the desktop app of Live Caption Pro, for Windows and macOS. It shows what is said on the microphone or the speakers in a transparent, always-on-top caption window – translated if you ask for it, and, if needed, it speaks the translation out loud. The speech is not processed by a cloud service but by servers running on your own machine or your own network; the internet is only needed to download the models.

What LiveCaption does

The work has three steps, and each one is done by a server. The steps build on each other: translation needs captions, and interpreting needs translation.

Step	What you get	Which server does it
Captions	the speech as text, split into sentences, a few seconds behind	the caption service (captionserver, with the Whisper model)
Translation	every sentence in the chosen language too, under the original	the translator (translatorserver)
Interpreting	the translated sentence is also spoken out loud while the original audio is lowered	the voice service (voiceserver), or the app's own built-in voice

The servers can run on the same machine as LiveCaption, or on a more powerful machine with a graphics card on the network – setting them up is covered in the server manual.

Where the captions come from

Captions can have two sources:

• Your own caption server (the default): the route to an accurate transcript. You can choose the model, the spoken language and the segmentation mode; it needs a machine with a graphics card – yours or one on the network.
• Windows Live Captions (Windows only): the system's built-in speech recognition. It needs neither a server nor a graphics card; in exchange, the recognition quality and the sentence boundaries are Windows', and the spoken language must be set in advance, exactly.

The source can be switched while running.

Which mode suits whom

Situation	Suggested mode	What it needs
Following a lecture, video or meeting in your own language	captions	a caption server, or Live Captions on Windows
Following a lecture or video in a foreign language	captions and translation	a caption server and a translator
A call with a partner who speaks another language, answering in their language	captions, translation and interpreting, with the outgoing direction	caption server, translator, voice service and a virtual audio cable
A machine without a graphics card, but with a powerful machine on the network	any mode, with remote servers	the server addresses and a key
A machine with neither a graphics card nor a server	captions with Windows Live Captions	Windows; translation still needs a reachable translator

The interpreter's voice

Interpreting can speak in three kinds of voice, chosen in the settings window:

• the speaker's cloned voice – the voice service learns from a few of the speaker's sentences and speaks the translation in their voice;
• the voice service's built-in voice – a fixed voice you can choose;
• the app's own voice – a fixed voice inside LiveCaption itself, with no graphics card and no voice service; the voice model has to be downloaded once.

If the chosen voice is not available, the interpreter stays silent and the caption window shows the error – the app does not switch to another voice on its own.

Windows and macOS

LiveCaption offers the same modes on both systems; the differences are in the audio sources and in the system menus (the tray on Windows, the menu bar on macOS). Where a step works differently on the two systems, the chapter says so. By default the caption window does not appear in screen sharing or in screenshots: the captions are visible on your own screen only.

Where next

• System requirements and installation
• First start and connecting to the servers
• The steps for captions without a graphics card: the Captions without a GPU guide.