Changelog

Quiet progress.Loud results.

Every release in the open. What we shipped, what we fixed, and a hint of what's coming next.

27+commands shipped

3platforms supported

100%offline

0data uploaded

Last updated June 13, 2026

v2.1.0 · MinorRead the room

June 13, 2026

Meetings, sharpened.

Meetings grow up. Hide filler words with one toggle, search any transcript and jump between matches, and watch the text highlight line by line as the audio plays. Voice learns spoken formatting commands and a custom dictionary, and screen context is now built in - type @screen to pull in whatever is on your display, no extra software required.

Highlights

✓Clean view - hide "um" and "uh" with one toggle; your original transcript is always kept
✓Search any transcript - find a word, jump between matches, scroll straight to it
✓Follow along - the spoken line highlights as audio plays; click any line to jump there
✓Custom vocabulary - teach names, brands, and jargon; speaker names remembered across meetings
✓Spoken formatting - say "new line", "comma", "question mark" while dictating
✓Screen context built in - @screen pulls in on-screen text; nothing extra to install

Added

Meetings: Clean view toggle hides filler words like "um" and "uh" while keeping the original transcript intact
Meetings: in-transcript search with match-to-match navigation and auto-scroll to the result
Meetings: follow-along playback - the spoken line highlights as audio plays, and any line is click-to-jump
Meetings: custom vocabulary for names, brands, and jargon; rename a speaker once and it sticks across future meetings
Meetings: Settings now surfaces leftover recording audio with a one-click button to reclaim the space
Voice: spoken formatting commands - say "new line", "comma", "question mark", and more while dictating
Voice: custom dictionary with a simple chip editor so dictation gets your words and names right
Screen context: type @screen in any command to include on-screen text, or turn it on for specific commands
Screen context: autocomplete can quietly take what is on your screen into account

Improved

Meeting summaries render as clean, formatted text instead of raw symbols, with copy buttons for the summary and the full transcript
Meeting audio is compressed automatically, taking far less disk space, and leftover audio from re-transcribing is tidied up on its own
Reading screen text is now built in - no extra software to install

v2.0.1 · PatchSteady hand

June 12, 2026

Polish, polished.

A focused fix pass. Polish lets go cleanly after the shortcut, the autocomplete tray stops repeating itself and adding stray spaces, and the local AI engine no longer drops a request and falsely reports that it stopped.

Highlights

✓Polish no longer behaves as if the shortcut keys are still held afterward
✓The Polish key no longer slips a stray character into your text
✓Autocomplete stops repeating words you already wrote
✓Local AI engine no longer drops mid-request with a false "not running" error

Fixed

Polish: releasing the shortcut no longer leaves keys feeling "stuck" - pressing Space right after will not accidentally open the prompt panel
Polish: the trigger key no longer slips a stray character into the text you are polishing
Polish: starting a Polish now clears any open autocomplete suggestion, since the polished text replaces what the suggestion was based on
Autocomplete: suggestions no longer repeat words you have already written
Autocomplete: accepting a suggestion no longer adds a double space
Autocomplete: a suggestion can never contain a line break, so accepting one cannot accidentally send your message in chat apps
Autocomplete: typing normal text right after a link no longer cuts suggestions down to a single word
Autocomplete: typo corrections and completions are more reliable, especially with smaller AI models
Stability: fixed the local AI engine stopping mid-request and falsely showing "The local AI engine isn't running" even though it had just been running

v2.0.0 · MajorQuiet rewrite

June 9, 2026

A new Typilot.

A second act. The biggest addition is Meetings - record any call, transcribe it locally with Whisper, label every speaker automatically, summarize with your local model, then chat with the transcript like a document. Voice mode, Polish, hands-free Notes, Autocomplete, and a brand-new light/dark appearance ship alongside.

Highlights

✓Meetings: record mic + system audio, local Whisper transcription, summary on your model
✓Speaker diarization that learns - name a voice once, recognized in every future meeting
✓Chat with the transcript - "What did Sarah commit to?" answered by your local model
✓Voice mode: hold a key, get clean prose - punctuation, filler-removal, capitalization
✓Polish: select anywhere, rewrite in place with your tone
✓Hands-free Notes - say "note: ..." and Typilot captures, tags, routes
✓Autocomplete: ghost suggestion in the tray after a pause - Tab to accept
✓New appearance - light + dark themes applied system-wide

Added

Meetings recorder - capture microphone and system audio locally; no audio ever uploaded
Local Whisper transcription for meetings (Metal-accelerated on Apple Silicon)
Speaker diarization - every voice labeled automatically; rename once and Typilot remembers across all future meetings
Auto-summary written by your local Ollama model when the recording stops
Chat-with-transcript - ask follow-up questions like "what did marcus commit to?" and the local model answers from the meeting text
Configurable retention - keep or delete the raw audio file after the summary is generated
Voice mode - local Whisper STT with auto-punctuation, filler-word removal, capitalization, paragraphs
Voice AI replies - speak a question, get an answer routed through any system prompt you write
Polish - select text in any app, press the Polish key, rewrite in place; non-typeable shortcut so it never overwrites your selection
Custom polish prompts - brand voice presets you can swap per use case
Autocomplete tray - small floating window suggests the next phrase after a short pause; Tab to accept; domain-aware (code, email, chat, doc)
Background Notes - record sessions silently; notes are timestamped, auto-tagged (Meeting / Action / Idea), grouped by date and project
Custom command editor - name, prefix, system prompt, shortcut binding, per-command model selection
Light + dark appearance applied across main window, floating bar, tray menus
New marketing surfaces - homepage, features (with deep-dives), pricing, download, security, roadmap, compare pages
Independent audit reports surfaced inline (GDPR EU · GDPR DE · CCPA)

Improved

Brand and product UI rebuilt - warm cream, sage accent, Instrument Serif headlines, fewer gradients
Atmosphere across the app - layered radial glows, vignettes, refined shadow stacks
Typography: editorial serif for headings, mono labels with consistent letter-spacing
Reduced-motion paths wired throughout - animations gracefully halt for users who prefer no motion
Keyboard activation feel - lower input lag, cleaner cancel zones
Model manager UX - clearer install progress, per-model status pills
Whisper model picker now shows speed/quality trade-off inline

Fixed

Edge case where the suggestion tray could linger after focus moved
Inconsistent shortcut binding between macOS and Linux modifier keys
Polish replacement losing trailing punctuation in some apps
System audio capture occasionally muting after sleep/wake on macOS
Several minor layout shifts during initial paint on large displays

v1.0.0 · InitialFirst keystroke

November 9, 2025

The first release.

The original Typilot. A keyboard shortcut, a small set of commands, and a way to pull text from anything on screen.

Highlights

✓Hotkey-driven local AI for every app
✓27 built-in commands shipped on day one
✓Screenshot OCR for image-aware prompts
✓Cross-platform - macOS, Windows, Linux

Added

27 built-in commands - gen:, fix:, exp:, trl:, sum: and more
Custom command prefixes - write your own system prompts
Screenshot / OCR command (img:) - read text from any region of the screen
Ollama integration - pull and pick from any local model
Cross-platform builds for macOS, Windows, Linux
3-day full trial - no card, no signup wall

What's next

We're working on

something quiet.

A few things are taking shape. We'd rather show than tell. Watch this page - you'll know when it lands.

Drop us a line

Changelog

Quiet progress.Loud results.

Every release in the open. What we shipped, what we fixed, and a hint of what's coming next.

27+commands shipped

3platforms supported

100%offline

0data uploaded

Last updated June 13, 2026

v2.1.0 · MinorRead the room

June 13, 2026

Meetings, sharpened.

Highlights

✓Clean view - hide "um" and "uh" with one toggle; your original transcript is always kept
✓Search any transcript - find a word, jump between matches, scroll straight to it
✓Follow along - the spoken line highlights as audio plays; click any line to jump there
✓Custom vocabulary - teach names, brands, and jargon; speaker names remembered across meetings
✓Spoken formatting - say "new line", "comma", "question mark" while dictating
✓Screen context built in - @screen pulls in on-screen text; nothing extra to install

Added

Meetings: Clean view toggle hides filler words like "um" and "uh" while keeping the original transcript intact
Meetings: in-transcript search with match-to-match navigation and auto-scroll to the result
Meetings: follow-along playback - the spoken line highlights as audio plays, and any line is click-to-jump
Meetings: custom vocabulary for names, brands, and jargon; rename a speaker once and it sticks across future meetings
Meetings: Settings now surfaces leftover recording audio with a one-click button to reclaim the space
Voice: spoken formatting commands - say "new line", "comma", "question mark", and more while dictating
Voice: custom dictionary with a simple chip editor so dictation gets your words and names right
Screen context: type @screen in any command to include on-screen text, or turn it on for specific commands
Screen context: autocomplete can quietly take what is on your screen into account

Improved

Meeting summaries render as clean, formatted text instead of raw symbols, with copy buttons for the summary and the full transcript
Meeting audio is compressed automatically, taking far less disk space, and leftover audio from re-transcribing is tidied up on its own
Reading screen text is now built in - no extra software to install

v2.0.1 · PatchSteady hand

June 12, 2026

Polish, polished.

Highlights

✓Polish no longer behaves as if the shortcut keys are still held afterward
✓The Polish key no longer slips a stray character into your text
✓Autocomplete stops repeating words you already wrote
✓Local AI engine no longer drops mid-request with a false "not running" error

Fixed

Polish: releasing the shortcut no longer leaves keys feeling "stuck" - pressing Space right after will not accidentally open the prompt panel
Polish: the trigger key no longer slips a stray character into the text you are polishing
Polish: starting a Polish now clears any open autocomplete suggestion, since the polished text replaces what the suggestion was based on
Autocomplete: suggestions no longer repeat words you have already written
Autocomplete: accepting a suggestion no longer adds a double space
Autocomplete: a suggestion can never contain a line break, so accepting one cannot accidentally send your message in chat apps
Autocomplete: typing normal text right after a link no longer cuts suggestions down to a single word
Autocomplete: typo corrections and completions are more reliable, especially with smaller AI models
Stability: fixed the local AI engine stopping mid-request and falsely showing "The local AI engine isn't running" even though it had just been running

v2.0.0 · MajorQuiet rewrite

June 9, 2026

A new Typilot.

Highlights

✓Meetings: record mic + system audio, local Whisper transcription, summary on your model
✓Speaker diarization that learns - name a voice once, recognized in every future meeting
✓Chat with the transcript - "What did Sarah commit to?" answered by your local model
✓Voice mode: hold a key, get clean prose - punctuation, filler-removal, capitalization
✓Polish: select anywhere, rewrite in place with your tone
✓Hands-free Notes - say "note: ..." and Typilot captures, tags, routes
✓Autocomplete: ghost suggestion in the tray after a pause - Tab to accept
✓New appearance - light + dark themes applied system-wide

Added

Meetings recorder - capture microphone and system audio locally; no audio ever uploaded
Local Whisper transcription for meetings (Metal-accelerated on Apple Silicon)
Speaker diarization - every voice labeled automatically; rename once and Typilot remembers across all future meetings
Auto-summary written by your local Ollama model when the recording stops
Chat-with-transcript - ask follow-up questions like "what did marcus commit to?" and the local model answers from the meeting text
Configurable retention - keep or delete the raw audio file after the summary is generated
Voice mode - local Whisper STT with auto-punctuation, filler-word removal, capitalization, paragraphs
Voice AI replies - speak a question, get an answer routed through any system prompt you write
Polish - select text in any app, press the Polish key, rewrite in place; non-typeable shortcut so it never overwrites your selection
Custom polish prompts - brand voice presets you can swap per use case
Autocomplete tray - small floating window suggests the next phrase after a short pause; Tab to accept; domain-aware (code, email, chat, doc)
Background Notes - record sessions silently; notes are timestamped, auto-tagged (Meeting / Action / Idea), grouped by date and project
Custom command editor - name, prefix, system prompt, shortcut binding, per-command model selection
Light + dark appearance applied across main window, floating bar, tray menus
New marketing surfaces - homepage, features (with deep-dives), pricing, download, security, roadmap, compare pages
Independent audit reports surfaced inline (GDPR EU · GDPR DE · CCPA)

Improved

Brand and product UI rebuilt - warm cream, sage accent, Instrument Serif headlines, fewer gradients
Atmosphere across the app - layered radial glows, vignettes, refined shadow stacks
Typography: editorial serif for headings, mono labels with consistent letter-spacing
Reduced-motion paths wired throughout - animations gracefully halt for users who prefer no motion
Keyboard activation feel - lower input lag, cleaner cancel zones
Model manager UX - clearer install progress, per-model status pills
Whisper model picker now shows speed/quality trade-off inline

Fixed

Edge case where the suggestion tray could linger after focus moved
Inconsistent shortcut binding between macOS and Linux modifier keys
Polish replacement losing trailing punctuation in some apps
System audio capture occasionally muting after sleep/wake on macOS
Several minor layout shifts during initial paint on large displays

v1.0.0 · InitialFirst keystroke

November 9, 2025

The first release.

The original Typilot. A keyboard shortcut, a small set of commands, and a way to pull text from anything on screen.

Highlights

✓Hotkey-driven local AI for every app
✓27 built-in commands shipped on day one
✓Screenshot OCR for image-aware prompts
✓Cross-platform - macOS, Windows, Linux

Added

27 built-in commands - gen:, fix:, exp:, trl:, sum: and more
Custom command prefixes - write your own system prompts
Screenshot / OCR command (img:) - read text from any region of the screen
Ollama integration - pull and pick from any local model
Cross-platform builds for macOS, Windows, Linux
3-day full trial - no card, no signup wall

What's next

We're working on

something quiet.

A few things are taking shape. We'd rather show than tell. Watch this page - you'll know when it lands.

Drop us a line