AIspeaky
Product

A desktop voice multitool for dictation, translation and quick commands.

Push-to-talk dictation with two insertion modes, realtime translation, and simple commands that run inside the app you're already focused on.

Download betaSee the tour

AIspeaky is currently in beta and free to use within beta limits. We do not read or store the content of what you dictate.

Laptop with dictation microphone — hand-drawn
Capabilities

What ships in the beta

Dictation that fits how you write

Speak naturally and insert text into the app you're already using — your editor, your inbox, your chat.

Available in beta
  • As-is mode: fast insertion with minimal transformation.
  • Post-corrected mode: punctuation, casing and obvious recognition errors cleaned up before insertion.
  • Push-to-talk and toggle modes; pick what fits your workflow.
As-is insertion
"thanks sending the deck after lunch"
Post-corrected
"Thanks — sending the deck after lunch."

Realtime translation

Translate speech and text in real time during cross-language calls, messages and notes.

Available in beta
  • Powered by managed cloud models through AIspeaky's blind relay.
  • Subject to beta usage limits so we can keep infrastructure costs predictable.
  • Insert translated output directly into the focused app.

Quick voice commands

Use your voice for simple desktop actions without leaving the keyboard workflow.

Available in beta
  • Available now: undo, copy, paste, text selection and similar deterministic commands.
  • Coming later: rewrite selected text, summarize, change tone.
  • We mark text-rewriting commands as upcoming instead of pretending they're ready.
Undo Copy Paste Select text

Text-rewriting commands

We're expanding the command set with rewrite, summarize and tone-change actions. Not ready for prime time yet.

Coming later
  • Backed by managed models when shipped.
  • Will be clearly marked as available only when quality holds up across apps.
  • Until then: use dictation post-correction for cleanups.

Cloud power, blind-relay architecture

Managed models give you the strongest recognition and translation. We hide provider credentials behind our relay and don't read your content.

Available in beta
  • Blind relay keeps provider keys server-side.
  • Technical metadata only: session identifiers, usage counters, crash/bug events.
  • We don't sell dictated text. We don't train on user content.
Where it runs today

Fedora/Linux and Windows 10/11 beta — macOS waitlist

Plain platform availability. No Web App. No browser add-on yet.

Fedora / LinuxAvailable in beta
Windows 10 / 11Available in beta
macOSWaitlist
Download beta Privacy architecture