❓ Frequently Asked Questions (FAQ) — Voxis Live

[Davut Akça (DavutAkca)]

A single reference for the questions people ask most about Voxis Live, the real-time voice translation app for Windows. Each answer is self-contained — jump to whatever applies.

---

What is Voxis Live?

Voxis Live is a real-time voice translation app for Windows. It captures your PC's system audio (a video, a game, or a call), translates it with a simultaneous speech-to-speech model, and plays the translation back in your language a few seconds behind — while the speaker is still talking, like a human interpreter.

Which operating systems does it support?

Windows 10 (version 2004+) and Windows 11 only. It relies on Windows-specific audio APIs (WASAPI loopback, the session-volume API, DPAPI), so there is currently no macOS or Linux build.

Do I need to install audio drivers?

No, not for one-way translation. The default driverless mode uses WASAPI process-exclude loopback — it reads the system audio and excludes Voxis's own output, so no virtual cable is needed. A virtual microphone (VB-CABLE) is only required for two-way Meeting mode, where your translated voice must be fed back into a meeting app.

Can I use my own API key?

Yes. The open-source build is BYOK (bring your own key). Open Settings → API key and paste a Google Gemini key from https://aistudio.google.com/. The key is stored encrypted on your machine via Windows DPAPI (bound to your Windows account) — never in a plaintext file, never sent anywhere except Google's Gemini endpoint.

Which translation engine does it use?

The open-source build uses Google Gemini Live (gemini-3.5-live-translate-preview), a native simultaneous speech-to-speech model. The official Microsoft Store app additionally routes some target languages through a Qwen voice engine for broader spoken-voice coverage.

How many languages does it translate?

79 translation target languages (e.g. Turkish, English, Spanish, French, German, Italian, Portuguese, Russian, Arabic, Chinese, Japanese, Korean, Hindi, and more). The app interface itself is localized in 16 languages. Translation targets and the interface language are set independently.

Can it translate video games and desktop apps, or only browser tabs?

Any Windows app. Because Voxis reads the Windows system audio mix directly, it translates native games, desktop Zoom/Teams/Discord calls, and any local video player. Browser-tab translation extensions can only reach audio playing inside one Chrome tab — Voxis is not limited to the browser.

How does two-way Meeting mode work?

You hear the other person translated into your language through your headphones (no extra install), and your own voice is translated and fed into the call through a virtual microphone (VB-CABLE), which the meeting app selects as its mic. Without a virtual cable installed, Meeting mode automatically runs listen-only (you understand them; your voice goes out untranslated). Setup steps are in the README.

What is the latency / delay?

Roughly the length of the sentence plus a few seconds. That lag is the translation model's designed ear-voice span — it waits for enough context to translate correctly, exactly as a human simultaneous interpreter does. It is not tunable from the client; there is no "go faster" setting, because the model owns its own timing.

Is Voxis Live free?

The open-source developer build in this repo is free — you run it from source and pay only for your own Gemini API usage. The official Microsoft Store app offers a free daily tier (a limited number of translation minutes per day) plus paid prepaid minute packages for heavier use.

Can I export the transcript?

Yes. Every session can be saved and exported as TXT, SRT, or VTT, with bilingual cues (source + translation side by side). Past sessions stay searchable in the in-app History panel.

Is my audio private? Where does it go?

The open-source build sends no telemetry, usage data, or analytics of its own — the only outbound connection is the Gemini Live WebSocket that your own key opens to Google. Your audio is streamed to Google's translation model to be translated; it is not stored by Voxis. There is no in-app auto-updater.

Does it work offline?

No. Translation runs on a hosted cloud model (Gemini Live), so an internet connection is required. There is no on-device/offline translation engine.

How do I install the end-user app (not from source)?

The polished end-user app ships through the Microsoft Store. Setup and download links are at https://voxislive.com. This GitHub repository is the open-source developer build you run from source with your own key.

Where do I get a Gemini API key?

From Google AI Studio: https://aistudio.google.com/. Create a key there, then paste it into Voxis under Settings → API key.

Why is Python 3.14 not supported?

The pinned versions of numpy and onnxruntime do not yet publish stable Python 3.14 (cp314) wheels, so pip install would fail. Use Python 3.11–3.13 (64-bit).

How is Voxis Live licensed?

Under the PolyForm Noncommercial License 1.0.0 — free for personal, hobby, research, and non-commercial use; commercial use, resale, and revenue-generating deployments require a commercial license (https://voxislive.com/licensing).

---

Didn't find your question? Ask in Q&A or open an Issue.

[T. Wasfy (tarekwasfy01)]

Here is my new tool for virtual cables, it uses a different aproach. You can have 32 for free. https://github.com/tarekwasfy01/Virtual-Cables

[Davut Akça (DavutAkca)]

Thank you for your comment, I will definitely review the repo.