Speak Estonian. The text appears where your cursor is.
Diksu turns Estonian speech into text. Dictate into whatever you are writing in, get a meeting written down, or transcribe a recording. Everything happens on your own computer.
- 14 days with every feature
- No payment card asked
- Works offline
Works where you write: Word Outlook Gmail Teams Excel Slack browser and everywhere else
Three things Diksu does
Dictate, get a meeting written down, turn a recording into text.
Ctrl + WinDictate
Hold the keys and speak. The text appears where your cursor is: an e-mail, Word, a chat.
How it goes
microphone + computer soundGet a meeting written down
Diksu listens to you and to the others at once. A Teams, Zoom or Meet meeting becomes text, speaker by speaker.
See an example
mp3, mp4, wav, m4aTurn a recording into text
Pick an audio or video file and get a transcript. An hour of recording takes about ten minutes.
All for one price
Meeting transcription
Let Diksu tick along.
Leave Diksu running in the background during a meeting. It listens to your microphone and to the computer's sound at the same time: what you say is written down, and so is what the others say in Teams, Zoom or Meet.
- Diksu tells the speakers apart itself
- No bot joins the meeting, and the audio never leaves your computer
- It also follows an Estonian sentence with English words in it
- Afterwards the text can be gone over with a slower, more accurate model
- Speaker 1
Does this work in a Teams meeting too?
- Speaker 2
Yes. The program listens to both the microphone and the computer's sound and writes the talk into a text file.
- Speaker 1
And it can tell who is speaking?
- Speaker 2
It can. It tells the speakers apart itself, just like in this text.
- Speaker 1
OK, but what if I mix English words into an Estonian sentence?
Diksu is listening and writingWho it is for
For anyone whose work is writing.
Say the sentence and it is written down.
- patient record
Patsient pöördus vastuvõtule kolm päeva kestnud peavaluga, palavikku ei ole.
Doctor
- contract draft
Leping jõustub 1. jaanuaril 2027 ja kehtib 12 kuud.
Lawyer
- reply letter
Teie taotlus on registreeritud ja vastame sellele 30 päeva jooksul.
Official
- feedback to a parent
Karl on sel veerandil matemaatikas palju edasi läinud, eriti tekstülesannetes.
Teacher
- interview notes
Küsisin ministrilt, millal eelnõu riigikokku jõuab.
Journalist
- client spreadsheet
Soovib pakkumist reedeks, enne seda helistada.
Sales manager
Privacy
Your voice never leaves the computer.
Speech is recognized on your own computer. Your audio, dictated text and transcripts are not sent to us or to anyone else. The program uses the internet only to download speech models and updates and to check the trial and the license key.
That makes Diksu fit for confidential work too: on a doctor's, lawyer's, official's or journalist's computer.
- Works offline
- No account needed
- No usage statistics collected
The Estonian speech models were made by the Laboratory of Language Technology at Tallinn University of Technology. Diksu is an independent product, not affiliated with TalTech.
How it works
Three moves.
Download and install
Installing takes a minute and needs no administrator rights. On first start the program downloads the Estonian speech model (about 160 MB).
Hold Ctrl + Win and speak
The text appears while you are still talking, usually within about a second. The program adds the punctuation itself.
Let go of the keys
For longer speech press Ctrl + Win + R: the program listens until you press it again, and your hands stay free.
Dictation into any program
The text is typed where the cursor is. No separate window, no copying.
Numbers as digits
A spoken “kaks tuhat kakskümmend viis” is written as 2025. Numbers can stay as words if you prefer.
A more accurate second pass
After a meeting the text can be gone over with a slower model that hears whole sentences and errs less.
Price
One price, every feature.
89 €
one payment per user, VAT included

- A license for one person, on up to three of their computers at a time
- Every update released within a year; the version you bought keeps working after that
- The invoice comes by e-mail; a company can add its details and VAT number when buying
- The license key right after payment
- Buy 5 or more licenses at once and the price is 20% lower
Payment is made on Stripe's secure checkout page. By paying you accept the terms of sale and license.
Does Diksu work offline?
Yes. It needs the internet on first start, to download the speech model, to activate your license key, and later for updates. You can dictate with no connection at all.
What computer do I need?
A computer with 64-bit Windows. No graphics card is needed: speech recognition runs on the processor.
How accurate is it?
With clear speech the text is mostly right, but every speech recognizer slips on names and rare words. Read the text before you send it. A meeting transcript can be gone over again with a slower, more accurate model.
Does Diksu write down meetings too?
Yes. Choose “Microphone + computer audio” as the audio input and Diksu writes what you and the others say into a text file, speaker by speaker. It works with any program whose sound comes out of the computer: Teams, Zoom, Meet. No bot joins the meeting, and the audio never leaves your computer.
Which files can be transcribed?
Audio and video files that Windows can open: mp3, m4a, mp4, wav, wma, aac, mov, wmv and 3gp.
Can I try it before buying?
Yes. The program works for 14 days with every feature, and no payment card is asked for.
How do I get my key after buying?
The license key appears on screen right after payment and is also e-mailed to you. Copy it, choose “Enter license key” in Diksu's menu and paste it into the window that opens.
On how many computers can I use a license?
On up to three computers at a time that you use yourself. When you change computers, free the old one's place on the “My computers” page. Each person needs a license of their own; you can buy several at once.
What happens after the year of updates?
The program keeps working. A license includes every version released within a year of the purchase, and those are yours for good. For later versions you can renew the license.
How is Diksu better than Windows voice typing (Win + H)?
Windows voice typing sends the audio to Microsoft's cloud and needs the internet. Diksu recognises speech on your computer, works offline, writes down meetings and transcribes recordings. A comparison with other tools is at diksu.winotto.ee/en/compare.
What languages does it support?
You can dictate in Estonian and English, also mixed in one sentence. The program and its installer are in Estonian, and can be switched to English.
Is there a Mac version?
For now Diksu runs on Windows only. A Mac version is planned.
Can I cancel a purchase?
You get the license key right after payment, so a purchase can't be withdrawn. That is what the 14-day free trial is for: make sure the program suits you before you buy.
Small but mighty!
Try Diksu free for 14 days. Every feature is there from the start, no payment card is asked and no account is needed.
