-
Notifications
You must be signed in to change notification settings - Fork 0
Feature Reference
DavutAkca edited this page Jul 26, 2026
·
1 revision
A spec-sheet style overview of what Voxis Live does. For narrative context, see the Welcome and FAQ discussions; for how it compares to other tools, see Comparisons.
| Translation targets | 79 languages |
| Interface (UI) languages | 16, set independently from the translation target |
| Engine | Native simultaneous speech-to-speech model — audio in, translated audio out, no separate transcribe → translate → speak steps |
| Speech coverage | Every target language is translated; a subset also has a spoken voice on the free tier (see voxislive.com/languages for the current list) — paid plans have a voice for all 79 |
| Latency | A few seconds behind the speaker, by design — this is the model's "ear-voice span," the same trade-off a human simultaneous interpreter makes for accuracy. It is not a client-side setting. |
| Mode | Direction | What it's for |
|---|---|---|
| Video / Game | One-way | Watching or playing something in another language. The original audio is ducked while the translation speaks. |
| Meeting | Two-way | Live calls. You hear the other party translated into your language; your own voice is translated and sent out through a virtual microphone. Without a virtual mic installed, this mode automatically falls back to listen-only. |
| Path | Requires a driver? | Notes |
|---|---|---|
| Driverless (default) | No | Reads the Windows system audio mix directly and excludes Voxis's own output, so it never re-translates its own translated speech. |
| Virtual-cable-routed | Yes (VB-CABLE) | Used for Meeting mode's outgoing leg, and unlocks stereo-aware, music-preserving ducking (dialogue is suppressed while background music in the stereo sides is kept). |
Four curated latency/quality modes, trading responsiveness against smoothness and (for the most aggressive preset) billed minutes:
- Smooth — balanced default.
- Fast — lower latency.
- Callout — tuned for short, punchy lines.
- Saver — gates translation to speech only, skipping silence, for the lightest usage.
- Every session can be saved and exported as TXT, SRT, or VTT.
- Exports can be bilingual (source and translation side by side) or translation-only.
- Past sessions are searchable in an in-app History panel; each session's files (transcript, exports, optional recordings) live together in one folder under
Documents\Voxis\Transcripts.
- Speaker labels — local, on-device detection tags captions with anonymous speaker labels (S1, S2, …) once more than one voice is present in a session.
- Language swap — a single control to exchange your "hear" and "send" languages in Meeting mode.
- Problem reporting — a one-click report attaches a redacted log and diagnostics; personal identifiers are scrubbed on-device before anything is sent.
| Build | API key | Auth / telemetry | Engine |
|---|---|---|---|
| Official app (Microsoft Store) | Provided by the service, per session | Sign in; usage billed | Gemini Live, with some languages routed through an additional voice engine |
| Open-source build (this repo) | Your own key (BYOK) | None — local, offline | Gemini Live only |
Windows 10 (version 2004 or later) and Windows 11. There is currently no macOS or Linux build — see the roadmap discussion for why and what's being considered instead.