How can we help?

Download the installer from the Download page — it runs on Windows 10 or 11, or Linux. On Linux it ships as a single portable file with nothing to install; on Windows it's a normal installer that doesn't need administrator rights. It runs without a graphics card at all, and with one, any Vulkan 1.2+ GPU works — NVIDIA, AMD, and Intel Arc alike. The smaller voice model runs comfortably on any 4 GB graphics card.

Just three ways in, all in one place: drop in an audio or video file, speak into the microphone, or capture what's playing through your speakers — the meeting you joined as a guest, the video with no transcript. And the voice keyboard types wherever your cursor is — browsers, editors, chat apps, any field that takes a keyboard — and can type the translation instead of the transcript.

Not at the moment. Windows 10 and 11 and Linux are supported with the same features, including translation. There is no date for macOS. There's a 14-day trial that needs only an email — no card. It unlocks everything except personal training and the agent connection.

That's what it's built for. It learns your voice, your jargon, and your proper nouns, on your own machine. Correct a misheard word once and that correction becomes the training example — no sample sentences to read out. Training is one click, runs on your hardware, and the new model is ready to switch to when it finishes.

1. Pin them once and they stop coming out wrong — in the transcript and in the translation. The same list steers both, so a name is never localised by accident.
2. Yes. Say it once and have it typed in English, Japanese, French and Spanish at the same time — one per line, or side by side in the same line. Add as many target languages as you like.
3. There's no length limit. The agent can sit through a two-hour meeting and hand you the transcript afterwards. Forget you left it recording and nothing is lost — it finishes the job by itself and keeps the transcript for you.
4. Yes — plain text or SRT, timed by sentence or word by word when you need to edit precisely. It subtitles in 38 languages, and translating a subtitle file keeps every timing intact.

Personal voice training (fine-tuning) is on the roadmap for Q3 2026 — see the roadmap post. Once it ships, training on your own voice will become the top fix for recognition issues.

Any microphone works. For best results, use a headset or USB condenser microphone. Built-in laptop microphones work but may pick up more background noise. The built-in noise reduction helps compensate.

Any modern GPU with Vulkan or CUDA support. NVIDIA: GTX 10-series or newer. AMD: RX 500-series or newer. Intel Arc GPUs and recent integrated graphics also work. You can also run in CPU-only mode — transcription will be slower but fully functional.

Is this a subscription? No. You pay once and it's yours. Every purchase is a perpetual licence with a year of updates included. After that year, updates are optional — and without them, the app keeps working on the version you have, indefinitely. It still works in five years, on the version you bought, whether you ever pay us again. What does the trial include? The 14-day trial unlocks everything except personal training and the agent connection. No card needed — only an email. What are the paid plans? Two. One for individuals, covering two devices, and one per-seat for teams, with volume discounts, priority support, and a team dashboard.

Yes. Deactivate the license on your old device from Settings → License, then log in on your new device. Alternatively, log into your account to manage devices.

Can my AI agent use it? Yes. Through MCP, the agent can transcribe, translate, and record a meeting of any length — a three-hour call is no different from a three-minute one. It works on your machine, and it's included with a paid licence. Is it open source? No. Brethof Voice Pro is proprietary software under its own licence.

The Linux build won't start on my machine. The Linux download needs glibc 2.38 or newer — a hard floor in how it's built. It runs on Fedora 39+, Ubuntu 24.04+, Debian 13+, and Arch and CachyOS. Below that version, the loader refuses to start it at all; there's no workaround. Do I need a graphics card? No. It runs on the CPU alone. With a GPU, transcription and translation each pick their own processor, so one can run on the graphics card while the other runs on the CPU.

What exactly leaves my machine? Nothing that matters. Three things reach the network and nothing else: a licence check at launch, an update check at launch, and the model downloads you start yourself. No audio, no text — ever. Once the models are on disk and the update check is off, nothing goes out at all.

My recording was made in a noisy room. The optional noise reduction is built for exactly that. Turn it on for the recording and it cleans it up.