Text to speech in Zoom, Teams and Google Meet
There are many reasons to let a synthetic voice speak in a video call: losing your voice, a speech disability, a presentation with prepared narration, or a meeting in a language you are still learning. The meeting apps carry your microphone, so the question is how to get a text to speech voice into it.
This guide shows two ways that work in Zoom, Microsoft Teams and Google Meet: playing prepared audio with computer sound, and routing a voice through a virtual microphone, using the free SuperClonica engine for the voice.
Prepare the audio
Write what you want to say, one short file per point, and voice it with the free engine: install it with python -m pip install superclonica-cpu and run, for example, superclonica --text-file intro.txt --voice voice.json --lang en --normalize --output intro.wav. --normalize gives every file the same loudness. To pick a voice first, type a line of 30 to 100 characters on any voice page and listen, free and without an account.
Share computer sound
In the Zoom desktop app, tick Share computer sound at the bottom of the Share Screen window before you share; under Advanced there is also a sound-only option, and computer sound cannot be shared while several screens are shared. In Teams, open Share content and turn on Include computer sound; on a Mac, Teams installs a driver the first time, and every sound of the computer, notifications included, goes into the meeting. In Google Meet in Chrome, present a Chrome tab: the tab’s audio is shared, so play your file in that tab.
Use a virtual microphone
For typing and speaking live, a virtual audio cable turns the engine’s output into a microphone: VB-CABLE on Windows and macOS, or the free BlackHole on macOS. Set the player or text to speech app to play into the cable and choose the cable as the microphone in Zoom, Teams or Meet. You will not hear yourself unless you also listen to the cable, and your real microphone is no longer used while the cable is selected.
Your own voice
If you are losing your voice to an illness, record a clean 5 to 10 second sample while you still can and train a voice in the studio; training on your own NVIDIA graphics card is free. Then every prepared line can sound like you. The engine speaks 31 languages; Chinese is not supported yet.
Tell people it is a synthetic voice
Say at the start of the meeting, in chat or in the first line, that you use text to speech, so nobody is confused by pauses before your replies. Never use the voice of another real person in a call without their consent, and never use a voice to pretend to be someone else.