Your first dictation takes about a minute and it works the same everywhere: put the cursor where text should go, hold one key, say a sentence, let go. The text appears at the cursor.
Do it once in a scratch document before you try it on real work, so the rhythm is familiar.
The four steps
- Click into a text field. Anything with a blinking cursor: a new email, a Word document, a chat box, a browser search field.
- Hold the dictation key.
fnon a Mac,Inserton Windows,Control+Spaceon Linux. A small recording bar appears. - Say one sentence. Speak the way you would to a colleague — normal pace, normal volume. Do not dictate punctuation.
- Let go. After a moment the finished text is written at your cursor.
That is the whole loop. Everything else in Sprecho is a refinement of it.
What "good" sounds like
Speech recognition is not fussy, but a few habits make a visible difference:
- Speak in whole sentences. The formatter uses the shape of the sentence to place punctuation. A complete thought punctuates better than three fragments.
- Normal conversational pace. Slowing down and over-enunciating usually makes accuracy worse, not better, because it breaks the natural word boundaries the model expects.
- Keep going through a stumble. If you misspeak, just say the correct version. Sprecho drops false starts and repeated words.
- Stay 15 to 30 cm from the microphone, off to the side of your mouth rather than directly in front of it, so hard consonants do not pop.
A useful first test sentence: "Hi Tom, thanks for the numbers — could you send the updated forecast before Friday?" You should get the dash, the comma and the question mark back without having said any of them.
What Sprecho does with what you said
Two things happen between releasing the key and the text arriving:
- Recognition turns your audio into raw words.
- Formatting turns those words into text: punctuation, capitalisation, paragraph breaks, lists where you clearly dictated a list, and the removal of "um", "uh", "like" and similar fillers.
Both run on Sprecho's servers in Germany, so the app needs to be online. A short sentence typically comes back in about a second; a long one takes proportionally longer, and the recording bar tells you it is still working.
The most common first-time mistake is clicking somewhere else while speaking. Text goes to whatever has keyboard focus at the moment it is inserted, so stay in the target window until the text has landed.
Two ways to hold the key
Holding the key is push-to-talk: record while held, insert on release. That is best for short bursts.
For anything longer, tap the key once instead of holding it. The recording locks on and keeps running hands-free until you press the key again. You do not need a double-tap — one tap is enough. Full details in Hold-to-talk vs lock mode.
If nothing appears
| What you see | What it usually is |
|---|---|
| No recording bar at all | The key is not registered, or another app has claimed it. Check Settings → General → Keyboard Shortcuts |
| Recording bar, but no audio level | Microphone permission or the wrong input device |
| Recording works, no text arrives | On macOS the Accessibility permission is missing, or Sprecho was not restarted after it was granted |
| The text is in the wrong app | Focus moved while you were speaking |
Both permission cases are covered in Grant microphone and Accessibility permissions.
Where to go next
- Teach it your names and jargon: Improve recognition accuracy
- Learn every shortcut: Keyboard shortcuts cheat sheet
- Cancel a dictation you do not want: Cancel or redo a dictation