R-CtrlPush-to-talk dictation for desktop
Dictation that stays out of your way.
Hold a key and speak. When you let go, the finished text is inserted at your cursor — in an email, a terminal, an ERP form or a patient record. typeout never takes focus and never asks you to paste anything.
Click. Speak. Done.
Click the pill and talk. Click it again, or just stop talking, and it types out exactly what you said — the real recogniser, with nothing rewritten.
Click the pill and talk
The bars are your actual input level, sampled ten times a second — the same rate the desktop overlay reads it at.
Where the 178 milliseconds go.
Key down to text in the field. Inference is the dominant cost at 93 ms; the network round trip adds 44, and capture, encoding and the paste into the target application together account for the remaining 41.
- Capture & encode
- 36ms
- Network round trip
- 44ms
- Inference
- 93ms
- Insert
- 5ms
- End to end
- 178ms
- You—
- Microphone10 ms
- Browser20 ms
- Encoding6 ms
- TLS1 ms
- Router1 ms
- ISP8 ms
- Edge12 ms
- Speech-to-Text93 ms
- Clean-upoptional · not in budget
- Streaming22 ms
- Insert5 ms
Built for how your field actually talks.
A profile is a set of instructions for what happens after you stop speaking. That makes typeout adaptable to a vocabulary rather than locked to one.
Specialised vocabularies
The Medicine profiles prime recognition with a medical vocabulary block. The same mechanism works for legal, technical or trade language — a profile is a recognition primer plus a set of formatting instructions, and both are yours to write.
Dialects and regional speech
Recognition is trained on natural speech rather than dictation-style enunciation, so South Tyrolean German, Swiss German or a strong regional Italian generally come through. It is not a claim we will make in the abstract — try it on your own voice in the demo above and judge it in a minute.
Whole practices and whole teams
Profiles are per-user, so a practice can run clinical dictation and correspondence side by side without either getting in the other's way. Enterprise puts the whole team on one invoice.
General office work
Nothing about typeout is medical by default. E-Mail and Prompts ship as standard profiles, and most people use it for correspondence, tickets, notes and chat all day without ever touching a specialised one.
Something we have not built yet
If your field needs a vocabulary or an output shape that the profile system cannot express, say so. That is a product conversation, not a support ticket, and it is how the Medicine profiles came to exist.
It types into the field that already has focus
No window to switch to and nothing to paste. typeout sends the finished text the same way your keyboard does, so it does not need to know what the application is.
Post-processing reshapes what you said
It does not invent content or answer questions. The active profile decides how your words are formatted — and a profile with post-processing switched off returns exactly what you said.
Profiles adapt the output to the task
Medicine, Medicine AI, E-Mail and Prompts ship by default. Each one changes the recognition primer, the formatting, or both.
Eight dictation languages, plus auto-detect
Speak German, get German. The formatting instructions never bleed the wrong language into your text.
A small floating pill, and the system tray
The pill shows what typeout is doing and lets you switch profile in one click. It can be hidden entirely.
Updates that do not interrupt you
They download quietly and apply the next time you quit. Nothing restarts underneath you and no dialog appears mid-sentence.
From your key to your cursor.
- 01
Hotkey
R-Ctrl, held. Nothing is captured until it is.
- 02
Audio capture
Only the phrase you are speaking.
- 03
Recognition
Runs on our servers, not on your machine.
- 04
Text processing
The active profile reshapes what you said.
- 05
Insertion at caret
Typed into the field that already has focus.
Recognition runs on our servers, which is why typeout needs an internet connection and has no offline mode. That is a deliberate trade: on-device models are either much slower or noticeably worse with accents, punctuation and proper nouns.
Answered plainly.
Including the answers that concede a limitation. Those are the ones worth reading.
Yes. typeout types into the focused field the same way your keyboard does, so it does not need to integrate with anything — email clients, browsers, ERP systems, patient record software, terminals and chat apps all work.
In whatever field your cursor is already in. There is no window to switch to and nothing to copy and paste.
No. It does not sit in calls and it does not listen continuously. typeout is dictation: you speak on purpose, into the field you are working in, right now.
No. Speech recognition runs on our servers, so typeout needs an internet connection. That is deliberate — on-device models are either much slower or noticeably worse with accents, punctuation and proper nouns.
No. Nothing is recorded unless you are holding the key. There is no always-on microphone and nothing is captured between phrases.
You can dictate in English, German, French, Spanish, Italian, Japanese, Korean and Chinese, plus automatic detection. The app interface is available in English, German, Italian, Spanish and Japanese.
No. typeout is a desktop product for Windows and macOS. There is no iPhone or Android client.
Yes. Profiles with post-processing switched off give you exactly what you said. The Medicine profile works this way on purpose — an assistant improving a clinical note is a liability, not a feature.