Stats
SuggestionsMined from your corrections (what the recogniser wrote versus what you changed it to) and from the fixes the model makes itself — only where the misheard word sounds like the term. 3 of your corrections or 7 model fixes of the same word add it by themselves (marked "auto" in the dictionary, with Undo). Accept adds the entry to your dictionary and lets the model decide from the sentence whether the misheard word means it ("cloth" → Claude only where it is about Claude); Always replaces it everywhere — for names and jargon that cannot mean anything else. Dismiss hides it. Runs by itself twice a day, or now with the button.
none open
Your styleFormatting habits learned from your edits: how you write greetings and sign-offs, where you put line breaks, which apps should always get the full cleanup. Every night at 03:30 your edited dictations are compared with what was pasted; a pattern is suggested only when it shows in at least three edits and, tried on your past dictations, brings them closer to what you sent and never further away. Nothing changes until you accept a suggestion. Accepted rules go to the cleanup model (also the cloud route) and greeting and sign-off rules are applied after it, even for short dictations.
Cleanup
verbatim: fillers and punctuation only · casual: light · professional: complete sentences · formal: Sie-form in German
Tone per app
Apps you dictated into. A per-app tone overrides the default; the learn switch can also be set per app.
LearningAfter a paste the client re-reads the field it pasted into (every second for 30 s, then every 5 s up to 2 min) and remembers the last version it saw. When the text is sent or the window ends, that last version is compared with what was pasted and a difference is stored as a correction of that dictation. Only that passage, never other fields, never passwords, never keystrokes. Web-based apps (Claude, Teams, Chrome, Slack) get their accessibility switched on for this.
Learn from my edits in all apps
Paste into the window I dictated in
Account default · a device can override it in its This Mac / This PC card
MeetingsWhen a Teams, Zoom or Webex meeting starts on your Mac (on Windows: Teams), Vocando asks whether to transcribe it — and whether everyone in the meeting agreed. Only then does it listen: your microphone as "Me", the meeting app's sound as "Them". The audio is transcribed piece by piece and never stored; the text appears live under Transcripts. Needs macOS 14.2 or newer and, once, the "System Audio Recording" permission for "Vocando Audio".
Offer to transcribe online meetings
Summary after the meeting
Your account
Plan & subscriptionPayment runs through Stripe: the card details go to Stripe only, never to Vocando. Stripe e-mails the invoices (VAT included; businesses can add their VAT number at checkout). Change seats, the payment method or cancel under “Manage subscription” — a cancelled subscription runs to the end of the period already paid.
PasskeysSign in on the public address with Touch ID, Face ID, Windows Hello or a security key — no password. Admins need a passkey there for the Admin tab. “Add a passkey” makes a one-time link, valid 15 minutes: open it in Safari, Chrome, Edge or Firefox on the device that should keep the passkey (passkeys in iCloud Keychain or Google Password Manager reach your other devices by themselves).
Invite & downloadsSend the invite link to someone on your Tailscale network: the page recognises their computer, offers the right download and explains the pairing code. Their request then appears under Admin → Devices asking to join. The links carry no secrets.
Phones: the browser client at /app on this server — signed in by the Tailscale network, no key needed.
Client · server · STT .
This Mac⚠ permission missing
Toggle modeOn: press the key to start, press again (or pause) to stop. Off: hold the key while speaking.
HotkeyThe key that starts a dictation. Right ⌥ is the default.
Floating pillThe small strip at the bottom of the screen: level meter while you speak, spinner, check mark, errors.
Start at loginmacOS: registers Vocando as a login item (System Settings → General → Login Items). Windows: a shortcut in the Startup folder.
Paste into the window I dictated inOn: the text goes into the window where you started speaking, even if you switched windows meanwhile. Off: into the window in front. "Account setting" follows Settings → Learning; on or off here applies to this device only.
Restore the clipboard afterThe pasted text is removed from the clipboard again after this delay. Raise it if an app pastes the previous clipboard content.
ms
Hotkey, floating pill, permissions and pairing belong to a device: open this page in the Vocando app on your Mac or PC (menu bar or tray → Open Vocando) to change them.
Continuous improvement
0 = never write audio to disk. Everything stays on your own server.
UsageYour own dictations on this server. Words are counted in the cleaned text; minutes are recorded speech; corrections are the texts you fixed afterwards.
Apps
LearningWhat happened after each paste in apps with "learn from my edits" on: sent/kept unchanged, or edited (a correction was stored).
Devices asking to joinA freshly installed client shows a six-digit code and waits here. Approve gives that device a key for the chosen account (a pending account becomes active); the device picks it up within seconds. Codes expire after 15 minutes. People can approve their own devices on /app/pair. Approve page
…
Trial & plansThe values every trial starts with (D-063/D-069). A changed default applies to every account that has no limit of its own; per account you set plan, limits and retention in the Accounts list below (Plan…). Usage counts from the activation or the last plan change.
Trial: words
dictated words in total
Trial: meeting minutes
transcribed meeting time
Trial: retention
days recordings and texts are kept (fixed for trial users)
Paid plans: retention
days, the starting value — the user may change it
Activation
while fewer than trials are active (now: –)