See It In Action

Your command center

One window holds it all. Drop in an audio or video file, hit the microphone, or capture what's playing through your speakers. Three ways in, one place — no separate tools to juggle, no menus to hunt through.

The last transcription is always visible at the bottom, alongside a training card showing how many voice samples you have stored for personal training — click it to browse the full sample list. From here you can open Settings, view History, correct the last transcription, or start recording with a single click. Closing the window hides the app to the system tray, where it keeps running in the background, ready the moment you press your hotkey.

Press, speak, done

Press, speak, done The voice keyboard. Press, speak, let go, and the words are where the cursor is — in a browser, an IDE, a chat app, any field that takes a keyboard. Want the translation instead of the transcript? Type that instead.

Transcribe any file — in any language

Drop in a two-hour meeting recording and go make coffee. No length limit, nothing to babysit — it works through the whole file and hands you the transcript. Forget you left it recording and nothing is lost; it finishes the job by itself. Everything happens on the machine in front of you, in your language, in your accent.

Record and translate live

Record a call, a meeting, a conversation, and watch the words appear as they're spoken. Add translation and you get a second language in the same breath. Your accent doesn't get in the way, because it has already learned you.

Capture what plays on your speakers

The meeting you joined as a guest. The video with no transcript. The webinar you missed half of. It hears what comes out of your speakers, not just what comes out of your microphone.

Translate into many languages at once

Speak once in Polish, have it typed in English, Japanese, French and Spanish at the same time — one per line, or side by side in the same line. Pick as many targets as you want. You said it once; it says it in all of them.

Translate text and subtitles

Translate plain text, or translate a subtitle file and keep every timing intact. SRT or plain text, timed by sentence or word by word when you need to edit precisely. At quality size, Tencent's translation engine reaches 97.9% of Google Gemini 3.1 Pro on the FLORES-200 benchmark, and beats it on real-world and minority-language translation — Tencent's own published figures. The quality size is what was measured, not the fast one.

Teach it your vocabulary

Your brand names, your jargon, your colleagues' names — pin them once, and they stop coming out wrong. The same pinned list steers the transcription and the translation, so a name is never localised by accident.

Fix mistakes, improve the model

Correct a misheard word once and it becomes training data. No sample sentences to read out. The fixes you already make teach it.

Your voice, your dataset

It trains on your voice, and the dataset is yours — it stays on your machine. 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. The call with your lawyer, the unreleased product, the patient notes — transcribed with nothing to leak.

Personalize the AI — on your machine

Training is one click, runs on your own hardware, and the new model is ready to switch to when it finishes. Two model sizes — the small one for laptops and integrated graphics, the larger one for accented or noisy audio. A choice, not a tier, and you can switch any time.

Models & devices

It runs on the laptop you already have. The small model needs 4 GB of graphics memory, the larger 6 GB, and with the quality translation engine loaded too, 8 GB. No graphics card at all? That works too — 8 GB of RAM and a four-core CPU. On Linux it's a single portable file with nothing to install; on Windows, an installer that needs no administrator rights. If your Linux is older than glibc 2.38 — Ubuntu 22.04, Debian 12 — it won't start. Fedora 39, Ubuntu 24.04, Debian 13 and newer are home.

Full control over input

Microphone, file, or speakers — the input is your choice. Transcription and translation can each pick their own processor, so one runs on the graphics card while the other runs on the CPU.

Noise reduction is there for recordings made in a noisy room, but it's off by default — because it was measured to make short clean microphone clips worse.

30 languages + 22 Chinese dialects, one click

Transcribe in 30 languages, translate across 38, and 23 work in both directions. On its own, it recognises 22 Chinese regional dialects. Your grandmother's Sichuan comes out right.

Everything you said, searchable

Every transcript is saved, on your machine, and searchable. The two-hour meeting you forgot about is still there. Nothing had to be uploaded for it to be kept.

Make it yours

Make it yours Give your AI agent ears. It can transcribe and translate through MCP, on your machine, with no API key — part of the paid licence. If your agent only reaches things by URL, the app can expose a local endpoint that requires a token and refuses to start without one. A three-hour call is no different from a three-minute one.

Stay up to date

Buy once, it's yours. A year of updates is included; after that, updates are optional, and without them the app keeps working on the version you have, indefinitely. It still works in five years, whether or not you ever pay us again. That's the whole deal.