On your Mac
Dictation history (when retention is on), dictionary, snippets, settings and your Deepgram key.
Closed beta · macOS
Dictate in Russian and English without switching languages. Scriptus prepares text for your app. You decide when to send.
For macOS 13 and later, Apple Silicon first. The signed build is not published yet.
Speech
okay write I'll send the report by six please check the numbers first
Three actions. Speech recognition and text cleanup are separate steps, and both are visible.
Recording runs while the key is held. Release, and recording stops. Nothing is sent on its own.
Speech is recognised in the Deepgram cloud — under your key or with Scriptus access. Optionally, a model cleans the text up: punctuation, dictionary, snippets.
The text is pasted where the cursor is. If the paste is not confirmed, the text stays in the app — copy it with one click.
Scriptus never presses Enter or sends a message. You read the ready text and decide.
Ready text stays in the app until you close it, even when a paste fails.
No language switching: mixed speech is recognised as spoken.
A Deepgram key is stored locally and is never sent to Scriptus.
A macOS app. The web account is only needed for Scriptus access.
Both options use the same Deepgram cloud recognition. The difference is who pays and where the key lives.
| Your own key | Scriptus access | |
|---|---|---|
| How to start | Enter a Deepgram key in the app settings | Get an invitation and enter the code in your account |
| Who pays for recognition | You, at Deepgram rates | Scriptus, within a weekly word quota |
| Limit | No Scriptus limit | Weekly word quota; resets every Monday |
| Where the key lives | On your Mac | No key on the Mac: the app receives a short-lived token per dictation |
| Status | Works in the beta build | Closed beta, invitation only |
Pricing is not announced. During the closed beta, Scriptus access is granted by invitation.
We separate what stays on your Mac, what goes to the recognition cloud, and what Scriptus knows.
Dictation history (when retention is on), dictionary, snippets, settings and your Deepgram key.
Audio is streamed to Deepgram under your key or a short-lived Scriptus token. Scriptus stores neither audio nor recognised text.
With Scriptus access we count numbers only: words, durations, timestamps. No text, no audio.
No. Speech recognition runs in the Deepgram cloud, so dictation needs a network. History, dictionary, snippets and settings are available offline.
Recognition under a Scriptus key instead of your own. It is invitation-only during the closed beta and limited by a weekly word quota. The app receives a short-lived token per dictation; no permanent key is stored on your Mac.
Not if you have Scriptus access. Your own key has no Scriptus limit and stays on your Mac. You can switch between the two in the app settings.
Audio is streamed to Deepgram during dictation and handled under Deepgram's terms. Scriptus does not store audio. With Scriptus access, the server issues a token and counts numbers only.
To the cursor in the active app and, when retention is on, to the local history on your Mac. Text is never sent to Scriptus servers.
macOS 13 and later. The first builds target Apple Silicon; an Intel build is planned later.
Open the closed beta page and email us. Invitations are issued by hand; we reply by email.
On the account page, under “Delete account”, with a typed confirmation. It removes your email address, plan, redeemed invitations and usage metadata. Local history on your Mac stays with you.
Questions, invitations and sign-in problems — write to us.