Two-Way Voice Translation in Teams, Zoom and Other Meeting Apps
Overview
This guide sets up the Sokuji desktop app for a video call with someone who speaks another language. They hear what you say, spoken in their language. You hear what they say, translated into yours and read aloud, and you can read both sides as subtitles. It uses KizunaAI (Powered by Soniox) and works with any meeting app that lets you choose a microphone, including the Microsoft Teams and Zoom desktop apps. The screenshots are from Sokuji 0.44.0 on Windows with Microsoft Teams.
1Set up Sokuji with KizunaAI
If you have not done so yet, follow the KizunaAI quick start to install Sokuji and sign in.
2Choose the two-way scenario
In the setup guide, choose Two-way online conversation. It puts Sokuji in Both mode, speaks your translation into the meeting and reads the other side’s translation to you. If you have already finished the setup, click Both in the mode switch at the bottom left of the main window instead. The first time, a panel asks where the other side is: keep In a meeting. Clicking Both does not switch on Translation I hear; you do that in step 5.
3Set the two languages
In Settings, under Translation Languages, set I speak to your own language and they hear to the other person’s. The line underneath shows the other direction: it reads “They speak English → I read Chinese” until Sokuji listens to one app such as Teams (step 5), then “→ I hear Chinese”, because the other side’s translation is read aloud to you. For a call with a client in the UK, for example: I speak Chinese, they hear English.
Tip
Do not leave I speak on Auto Detect. In Both mode Sokuji has to know your language to translate the other side into it, and Start stays disabled until you choose one.
4Install the virtual audio cable
Your translated voice reaches the meeting through a virtual microphone. On Windows this is VB-CABLE, a free driver. While it is missing, Sokuji asks to install it each time it starts: click Install Now, and restart Windows if it asks you to. From then on, Sokuji sends its translated voice to VB-CABLE by itself.
Tip
On macOS the Sokuji installer adds a virtual audio driver, and on Linux Sokuji creates its own virtual microphone. The rest of this guide is the same.
5Choose your microphone, the meeting app and what you hear
Still in Settings: under Microphone, pick your real microphone. Under Other’s audio, pick your meeting app, here Microsoft Teams. Sokuji then listens only to that app, so sound from other programs is not translated, and you keep hearing the other person’s own voice through Teams as usual. In the Speech block, Translation the other side hears goes to the virtual microphone, and Translation I hear reads the other person’s translation to you on the device beside it. Switch Translation I hear off if you would rather only read it.
Tip
If your meeting app is not in the list, join the call first (an app often appears only once it plays sound) and click the refresh icon. System Audio (All Applications) works too, but then Translation I hear is greyed out, so that Sokuji does not translate its own voice again: you only read the other side.
6Choose VB-CABLE as the microphone in the meeting app
In Microsoft Teams, open Settings → Devices and set the microphone to CABLE Output (VB-Audio Virtual Cable). In Zoom the same setting is under Settings → Audio → Microphone. Leave the speaker on your usual headset or speakers.
Tip
Your real microphone is selected in Sokuji, not in the meeting app: Sokuji listens to you, and the meeting app receives only the translated voice.
7Start translating
Click Start in Sokuji. When you speak, the other side hears your words in their language through the meeting app, and Sokuji shows both sides as subtitles. When they speak, you hear their own voice through Teams, then Sokuji reads the translation to you in your language.
Tips for Best Results
- Wear headphones. Otherwise your microphone picks up the other person from your speakers, and Sokuji translates them as if they were you.
- Make a short test call first, for example with a colleague, and check that they hear the translated voice.
- Pause briefly at the end of each sentence: the translation is spoken once you finish a thought.
- Hearing the other side read aloud costs extra. If subtitles are enough, switch off Translation I hear before you start: it cannot change while a session runs.
Frequently Asked Questions
Does the other person need to install anything?
No. Only you run Sokuji. The other person just hears a microphone that speaks their language.
Does this work with the Teams desktop app?
Yes. The Sokuji desktop app works with any meeting app that lets you choose a microphone, including the Teams and Zoom desktop apps.
The other person hears nothing. What should I check?
Check that the meeting app’s microphone is CABLE Output (VB-Audio Virtual Cable), that Translation the other side hears is switched on in Sokuji, and that Sokuji is in Me or Both mode with a session running.
The other person’s words are not translated. What should I check?
Check that Other’s audio is set to your meeting app or to System Audio (All Applications), that Sokuji is in Both mode with the other side in a meeting, and that the call is actually playing sound.
I read the other side’s translation but do not hear it. What should I check?
Check that Translation I hear is switched on and plays on the device you listen to, and that Other’s audio is your meeting app rather than System Audio (All Applications). Stop the session to change the switch.
Does translating both directions cost twice as much?
About that, by default: Both mode runs a separate Soniox session for each direction, which copes best when both people talk at once. Switching on Shared session in Both mode (Settings → Advanced → Provider) translates both sides in one session at lower cost. Reading the other side aloud adds a second voice on top. The billing guide has the details.
Troubleshooting
Still stuck? Send us feedback from the account menu, or write to support@kizuna.ai. You can also ask in GitHub Discussions.