tabi is a Mac voice app that turns speech into cleaned-up text at the cursor in any Mac app, reads selected text or documents aloud, and answers questions or rewrites text on voice command.
tabi is a Mac voice app that turns speech into cleaned-up text at the cursor in any Mac app, reads selected text or documents aloud, and answers questions or rewrites text on voice command.
More detail
All speech-to-text, cleanup, and text-to-speech run on Mistral AI models.
Best for
Mac users on macOS 14+ who write across Mail, Slack, and other apps and want dictation that inserts finished text anywhere, plus hands-free read-aloud of long documents and books.
More detail
Also suits multilingual users who need dictation in 56 languages and read-aloud translation in 37.
Pricing
Free while in beta, with every feature switched on and no account or card required.
More detail
No subscription tiers, one-time fees, or usage caps are published — unknown.
A paid plan may come later, with in-app notice before anything changes and no charge without consent.
Non-monetary costs: every take needs
Main limitation
Windows, Linux, iOS, or Android users, and anyone who needs offline or air-gapped operation.
More detail
Poor fit for privacy-maximalists who object to audio and read-aloud text going to Mistral, or for buyers needing contractual stability while the app is in beta.
tabi is a Mac voice app that turns speech into cleaned-up text at the cursor in any Mac app, reads selected text or documents aloud, and answers questions or rewrites text on voice command.
All speech-to-text, cleanup, and text-to-speech run on Mistral AI models.
A practical view of what is available and where limitations apply.
Offering
Available
Detail
Dictation at the cursor
Yes
Hold fn, speak, release; cleaned-up text lands wherever the cursor is in any Mac app. Tap fn twice for hands-free mode.
Ask via fn + Control
Yes
Turns a take into a request: answer a question, rewrite selected text in place, or draft a reply; with no text field, output appears above the on-screen orb.
Read aloud any selection
Yes
⌃⌘R or the menu-bar command reads selected text in a real voice, with near-instant start.
Listen mode for documents
Yes
Import PDF, Word, EPUB, TXT, or a photo of a page; page numbers and running headers are skipped, sentences are highlighted, and scanned PDFs/photos go through Mistral OCR.
Read-aloud translation
Yes
Translates and reads text in 37 spoken languages, with a switch to show the text as written next to what you hear.
Audio export
Yes
Saves the whole Listen text as an M4A file in the chosen voice and language.
Custom Dictionary
Yes
Fixes recurring mis-transcriptions and records the user's pronunciation of names for read-aloud.
Per-app writing styles
Yes
Each app can own a style, so cleanup in Slack differs from cleanup in Mail.
Offline operation
No
Speech-to-text, cleanup, and text-to-speech run on Mistral's cloud models, so an internet connection is required.
Independent third-party reviews
Unknown
None found in the supplied corpus; review themes and real-world limitations cannot be evidenced from independent sources.
Bot trust
How confidently AI systems can verify this profile from public evidence.
48of 100
ChatGPT50/100
Claude45/100
What bots could verify
• Comprehensive disclosure of data flows, Mistral 30-day retention, and preview-model training risks.