DeepL Voice Buddy

Real-time translated captions and voiceover, powered by the DeepL Voice API.

Watch foreign-language videos with captions in your own language. Join a call in German and read along in English. Speak into your mic and let other apps hear your translated voice. DeepL Voice Buddy runs on Windows, macOS, and any Chromium-based browser.

Three ways to run it

Client	What it hears	What it shows
Windows app	Any system audio, any app, any mic, game voice chat	On-screen overlay, and optional translated voice out to any audio device (including virtual cables for OBS / Discord / Zoom)
macOS app	Any microphone, system audio via ScreenCaptureKit	Menu-bar overlay, and optional translated voice out to any output device (works with BlackHole for virtual mic routing)
Chrome extension	Audio from the active browser tab	Subtitle overlay anchored to the video element, and optional translated audio played back through the tab

All three share a settings format and route through the same DeepL Voice Realtime session protocol, so behaviour is consistent across platforms.

Features

• Live captions that update as DeepL transcribes. Unfinished text shows dimmed while the model is still deciding; it sharpens into full brightness once a segment is finalised.
• Optional translated voiceover. DeepL streams the translated voice as audio; the app plays it through any output device you pick.
• Bring-your-own DeepL Pro key. Nothing is proxied through third parties, your audio goes directly to DeepL.
• Per-platform overlay styling: font, colour, size, background opacity, anchor point, max visible lines, click-through.
• Debug consoles that log every REST and WebSocket frame for easy troubleshooting (desktop apps).

Getting started

You need a DeepL API Pro subscription. The Voice API is not currently on the Free tier. Grab a key from deepl.com/pro-api.

Windows

Two install options:

Via 0install (recommended — auto-updates, smaller downloads):

1. Download VoiceBuddy-Setup.exe from the latest release and run it. It installs Zero Install and subscribes to the VoiceBuddy feed.
2. If you already have 0install, skip the bootstrap and just run: 0install run https://deejaytc.github.io/VoiceBuddy/voice-buddy.xml

0install pulls in the .NET 10 Desktop Runtime as a shared dependency, so app updates stay small and don't re-bundle the runtime each time.

Manual download:

1. Download the latest voice-buddy-<version>-win-x64.zip from the releases page, unzip anywhere, and run VoiceBuddy.exe. Requires the .NET 10 Desktop Runtime to be installed.

Once running:

1. Open the Settings tab and paste your DeepL API key.
2. On the Overview tab, pick an audio source from the Device dropdown. Loopback devices (system audio) show a speaker icon, microphones show a mic icon.
3. Click Start capture and play some audio through the selected source. Captions appear in the overlay.

Optional: check Voice: stream translated audio to also play the translated voice. Pick an output device and, if you like, a voice gender.

macOS

1. Download the latest VoiceBuddy.app from the releases page and move it to Applications.
2. On first launch, macOS will ask for Microphone and Screen Recording permissions. Screen Recording is how ScreenCaptureKit exposes system audio; no video is actually recorded. Grant both in System Settings → Privacy & Security.
3. Open Settings and paste your DeepL API key.
4. On the Overview tab, pick an input device and click Start capture.

The menu-bar icon gives you quick access to start, stop, and lock the overlay.

Chrome / Edge / Brave

1. Clone this repo (or download the latest release zip).
2. Open chrome://extensions, enable Developer mode, and click Load unpacked. Point it at src/VoiceBuddy.Chrome/.
3. Click the DeepL Voice Buddy icon in your toolbar, open Settings, and paste your DeepL API key.
4. Go to a tab with audio (YouTube, Twitch, Google Meet, Microsoft Teams web, a podcast, anything). Open the popup, choose the output language, and click Start.

Captions appear overlaid on the video. You can also enable Play translated audio to have the tab speak in your chosen language instead of the original. The extension follows your system's light/dark mode.

Virtual microphone (OBS, Discord, Zoom)

On the desktop clients you can route the translated voice into other apps as if it were a microphone. This needs a virtual audio cable:

• Windows: install VB-CABLE (free, signed by VB-Audio).
  1. In DeepL Voice Buddy, pick CABLE Input (VB-Audio Virtual Cable) as the output device.
  2. In OBS / Discord / Zoom, set the microphone to CABLE Output.
• macOS: install BlackHole 2ch (free, signed).
  1. In DeepL Voice Buddy, pick BlackHole 2ch as the output device.
  2. In OBS / Zoom / Discord, set the microphone to BlackHole 2ch.

Speak into your real mic. DeepL translates, the translated voice streams into the virtual cable, and other apps receive it as your microphone input.

The Chrome extension cannot expose a virtual microphone. Browsers do not have that capability.

Settings

Platform	Location
Windows	%AppData%\VoiceBuddy\settings.json
macOS	~/Library/Application Support/VoiceBuddy/settings.json
Chrome	Browser-managed chrome.storage.local (per profile)

Desktop settings files share the same JSON schema, so you can copy one across machines. API keys are stored in plaintext inside a user-scoped directory; don't commit the file.

Delete the file (or clear extension storage) to reset to defaults.

System requirements

• Windows 10 version 1809 or newer, or Windows 11.
• macOS 14.0 or newer.
• For the extension: any Chromium-based browser from 2024 onwards (Chrome, Edge, Brave, Arc, Opera).
• An active DeepL API Pro subscription.

Building from source

The three clients are independent projects:

• Windows: see the project files under src/VoiceBuddy.App/. Requires the .NET 10 SDK. Build with dotnet build, run with dotnet run --project src/VoiceBuddy.App.
• macOS: see src/mac/VoiceBuddy/README.md. Requires Xcode 15+ and xcodegen.
• Chrome: see src/VoiceBuddy.Chrome/README.md. Load unpacked from the folder. No build step, no bundler.

Known limitations

• DeepL Pro is required. The Voice API is not on the Free tier today. The Free host option remains in the UI so it will keep working if DeepL ever adds Free-tier access.
• 30-second inactivity timeout. If no audio reaches DeepL for 30 seconds the server closes the session. The desktop apps surface this as a status message; restart capture to reconnect.
• 1-hour session cap on the DeepL side. There is no auto-reconnect yet; restart capture.
• Unsigned binaries on both Windows and macOS until the projects are code-signed and notarised. Windows SmartScreen and macOS Gatekeeper will warn on first launch.
• Chrome extension cannot capture non-browser audio. It only sees the active tab.
• No shipped virtual audio driver. Virtual-mic routing relies on VB-CABLE or BlackHole, which the user installs.

License

MIT.