KONWA Setup SKILLS

---
name: konwa-setup
description: Install, set up and troubleshoot KONWA, a desktop app for macOS and Windows that translates what the user says in real time and sends the translation, through a virtual microphone, into Zoom, Teams, Discord, Tencent Meeting, OBS and other apps. Use when the user wants to install KONWA, connect it to a call or streaming app, or fix "the other side hears nothing / hears my own voice / hears an echo / I hear nothing".
---

# KONWA setup and troubleshooting

KONWA translates what the user says into another language in real time, speaks it with a synthesized voice, and sends that voice into other apps through a virtual microphone, so the other side hears the translation. It can also caption what the other side says (Listen Translation).

## Read first: what only the user can do

When you reach one of these, tell the user what to do and where, and wait until they say it is done. Never type a password or a code for them, and never try to get around these:

- Their computer or administrator password (macOS, when authorizing the virtual microphone)
- Windows asking "Do you want to allow this app to make changes to your device?" and "Would you like to install this device software?"
- Signing in on the web: the email code, the human check, a Google account
- System permission prompts: Microphone, Screen & System Audio Recording, Accessibility
- Restarting the computer

## The route

The translation reaches the other side through four hops; each one is needed:

1. Physical microphone → KONWA (Settings → Devices → You speak)
2. KONWA → virtual microphone (Settings → Devices → The other side hears; KONWA Microphone on macOS, VB-CABLE on Windows)
3. Virtual microphone → the target app's microphone
4. Target app → the other side

There is a separate path: the target app's and the system's sound → the user's headphones. It must be headphones: on speakers the microphone picks up the other side's voice and the translation, and KONWA translates them again as if the user said them.

When something is wrong, find which hop is broken first, then fix that hop.

## Check first

Whatever a command can tell you, find out yourself instead of asking the user.

macOS:

- System version: `sw_vers -productVersion` (14 or later; Listen Translation needs 14.2 or later)
- Chip: `sysctl -in hw.optional.arm64` (`1` means Apple silicon; Intel Macs are not supported. Don't rely on `uname -m`: it says x86_64 when Terminal runs under Rosetta)
- Whether KONWA is installed, and which version: `defaults read /Applications/KONWA.app/Contents/Info CFBundleShortVersionString`
- Whether the virtual microphone is installed: `ls "/Library/Audio/Plug-Ins/HAL/" | grep -i konwa` (expect `KONWA Microphone.driver`)
- Audio devices and the default input: `system_profiler SPAudioDataType`
- Which call or streaming apps are installed: `ls /Applications`

Windows (PowerShell):

- System version: `[Environment]::OSVersion.Version` (Windows 10 / 11, 64-bit; build 22000 or later is Windows 11; Listen Translation needs build 19041 or later, that is Windows 10 version 2004)
- Whether KONWA is installed, and which version:
  `Get-ItemProperty HKCU:\Software\Microsoft\Windows\CurrentVersion\Uninstall\*, HKLM:\Software\Microsoft\Windows\CurrentVersion\Uninstall\*, HKLM:\Software\WOW6432Node\Microsoft\Windows\CurrentVersion\Uninstall\* -ErrorAction SilentlyContinue | Where-Object DisplayName -like 'KONWA*' | Select-Object DisplayName, DisplayVersion`
- Whether VB-CABLE is installed: `Test-Path HKLM:\SYSTEM\CurrentControlSet\Services\VBAudioVACMME` (KONWA checks the same key)
- Whether the virtual devices are working: `Get-PnpDevice -Class AudioEndpoint -PresentOnly | Where-Object FriendlyName -like '*CABLE*' | Select-Object FriendlyName, Status` (expect CABLE Input and CABLE Output, perhaps also CABLE In 16ch, all with status OK)
- The default playback device: no command shows it. Run `mmsys.cpl` and ask the user which device is the default on the Playback tab; it must not be CABLE Input.
- Running apps: `Get-Process | Select-Object -Unique ProcessName`

## Install and set up

Go in order, and reach each step's "Done" before the next.

### 1. Install

Download the installer for the system from https://konwa.live/download.

- macOS: open the dmg and drag KONWA into Applications.
- Windows: run the installer. If "Windows protected your PC" appears, click "More info" → "Run anyway" (KONWA's installer is not code-signed yet). The installer adds VB-CABLE, the virtual microphone KONWA needs, unless it is already there; there is nothing to tick. Hand the system's permission prompt and, if Windows asks, "install this device software" to the user. The installer does not ask for a restart; when it finishes, KONWA opens.

Don't download VB-CABLE or BlackHole separately: KONWA brings the virtual microphone it needs.

Done: KONWA opens.

### 2. Sign in (the user does this)

In KONWA, click "Sign in through your browser". The user signs in on the page that opens, with an email code or a Google account; when the browser asks whether to open KONWA, choose to allow it.

If KONWA does not come back on its own: click "Enter code manually" in KONWA and enter the 8-character code shown on the browser's finished page (not the 6-digit code from the email).

Done: KONWA no longer shows the sign-in screen.

### 3. First run

After sign-in, KONWA walks the user through choosing languages, allowing the microphone and saying something to try. The try plays on the user's own headphones or speakers and never needs the virtual microphone.

- Ask first: which language the user speaks, and which language the other side should hear.
- When the system asks for microphone access, hand it to the user to allow. If it was refused before: on macOS, turn KONWA on in System Settings → Privacy & Security → Microphone; on Windows, in Settings → Privacy & security → Microphone (on Windows 10, Settings → Privacy → Microphone).

Done: when the user speaks, KONWA shows what they said and its translation.

### 4. Virtual microphone (hop 2)

- macOS: after the first run, KONWA offers to add KONWA Microphone. The user clicks "Authorize" and enters this Mac's password (an administrator account is needed). If they clicked "Later", click "Authorize" in Settings → Devices.
- Windows: after the first run, KONWA checks VB-CABLE, which the installer added. If it is ready, KONWA goes straight to the home and shows once "In your voice app, choose CABLE Output as the microphone". Otherwise it shows one page: "Restart" (only when KONWA says a restart is needed; hand the restart to the user) or "Authorize" (the user allows Windows' permission prompt and, if Windows asks, "install this device software"). If they clicked "Later", the same button is on the home and in Settings → Devices.

Then open Settings → Devices and read the device name under "The other side hears": KONWA Microphone on macOS, CABLE Output (VB-Audio Virtual Cable) on Windows. You will need its exact text below.

Done: you have the device name, and that row shows no "Authorize" or "Restart".

### 5. KONWA's input (hop 1)

In Settings → Devices → You speak, choose the user's physical microphone, or keep "Follow system default" if the system's default input is the physical microphone.

Done: "You speak" is the physical microphone, not the virtual one.

### 6. The target app (hop 3)

Look at the installed or running apps, list the candidates and ask the user which one. In that app, set the microphone to the device name from step 4, and keep its output on the headphones.

- Zoom: Settings → Audio → Microphone; also turn off "Automatically adjust microphone volume"
- Microsoft Teams: Settings and more (…) → Settings → Devices → Audio settings → Microphone
- Google Meet: More options → Settings → Audio → Microphone
- Discord: User Settings → Voice & Video → Voice tab → Input Device
- Tencent Meeting: before a meeting, Settings at the bottom left of the main window → Audio; in a meeting, Settings at the top right → Audio. Also turn off automatic microphone volume adjustment
- OBS Studio: add an "Audio Input Capture" source and choose the virtual microphone as its device; mute or remove the physical microphone's source. Don't also choose the same device in Settings → Audio → Mic/Auxiliary Audio: that echoes.
- TikTok LIVE Studio (Windows and Mac): the audio mixer in the control bar → Microphones → Add → choose the virtual microphone as the Microphone device → Add
- Steam: Steam Settings → Voice → Voice Input Device
- Any other app: find "Microphone" or "Input device" in its audio or voice settings.

Done: that app's microphone is the virtual microphone and its output is the headphones.

### 7. Languages

In the control bar at the bottom of KONWA's home, the first language is the one the user speaks and the second is the one the other side hears. Set them to what you asked in step 3.

Done: both languages are right.

### 8. Verify

1. In Settings → Devices, click "Preview" to confirm KONWA gets sound into the virtual microphone (hop 2).
2. On KONWA's home, click "Start translating" and have the user say a line.
3. With the target app's own test, or by asking the other side, confirm the other side hears the translation, not the user's own voice. For example Zoom's "Test microphone" in its audio settings, Teams' "Make a test call", Tencent Meeting's microphone check, Discord's Mic Test.

Done: the other side hears the translation.

### 9. Optional: Listen Translation (captions of the other side)

Needs macOS 14.2 or later, or Windows 10 version 2004 or later. In "Listen" on the home, choose the app the other side's voice comes from.

macOS asks whether KONWA may record system audio; hand it to the user to allow. If it was refused before, turn KONWA on in System Settings → Privacy & Security → Screen & System Audio Recording, then choose the app again. Windows asks for no permission.

Done: when the other side speaks, KONWA shows their captions.

## Troubleshooting

| Symptom | Where it breaks | Fix |
| --- | --- | --- |
| The other side hears nothing | hop 3 not connected, or KONWA not translating | set the target app's microphone to the virtual microphone; click "Start translating" in KONWA |
| The other side hears the user's own voice | hop 3 is still the physical microphone | set the target app's microphone to the virtual microphone |
| The other side hears it twice, or an echo | on Windows, the target app's output is set to CABLE Input | set its output back to the headphones |
| The other side hears the volume swell and fade | the target app adjusts the microphone volume automatically | turn that setting off |
| The user hears nothing at all | on Windows, the system output is set to CABLE Input | set it back to the headphones in Settings → Devices → You hear, or in Windows' Sound settings (KONWA also switches it back to a real speaker at launch and when a translation starts) |
| KONWA does not hear the user | hop 1 | check "You speak" and the system's microphone permission |
| "Another app is using the microphone" | hop 1 | close the app holding the microphone |
| The other side's words show up in the user's own captions | the user is on speakers | use headphones |
| The home or Settings → Devices says "The virtual microphone isn’t ready" | hop 2 not set up, removed or broken | click "Authorize": it is also the retry and the repair; restart only if KONWA then asks |
| It says "The virtual microphone works after a restart" | the virtual microphone was just installed | click "Restart" (hand it to the user) |
| The test says the device "is not passing sound back" | the virtual device is stuck | restart the computer |
| "The audio component did not load" | KONWA's install is incomplete | reinstall KONWA |
| On a Mac, the target app still uses BlackHole 2ch | left from an older KONWA | change it to KONWA Microphone (BlackHole may stay or go) |
| "Push to talk" is chosen but the key does nothing (macOS) | no Accessibility permission | the user turns KONWA on in System Settings → Privacy & Security → Accessibility |
| "Out of time" | the account's time is used up | hand it to the user |

## Don't

- Don't choose the virtual microphone as KONWA's input ("You speak").
- On Windows, don't set CABLE Input as the system's or the target app's output.
- Don't install BlackHole, VB-CABLE or any other virtual audio device separately.
- Don't type passwords or codes for the user, and don't get around a human check or a system permission.