Parrot is in active development. Here's what we've built so far and what's coming next.
v0.2.3
LatestJuly 20, 2026
Added
Cleanup quality tiers — pick a faster, lighter model or a higher-quality 1.5B / 3B tier so you can trade speed for polish
Formality presets in cleanup settings — dial the tone of cleaned-up text to match how you write
Improved
Faster on-device cleanup — warm model context, cached system prompt, and flash attention for lower latency
Short dictations get cleaned too — brief utterances are no longer skipped by cleanup
v0.2.2
July 19, 2026
Improved
More stable AI cleanup — the cleanup model runs in its own isolated process, so it no longer conflicts with the transcription engine
Better cleanup output — improved handling of filler words, disfluencies, and formatting
Cleaner recording HUD — the chip shows only the listening timer while you’re recording
Fixed
Cleanup no longer crashes — resolved a memory-corruption issue in the cleanup engine
More reliable Bluetooth mic — the always-open input stream is scoped to Bluetooth devices only
v0.2.1
July 19, 2026
Improved
Parrot is fully local-only — no account, login, or cloud sync. Dictation, cleanup, and history stay on your Mac
Simpler onboarding: set up a local profile and start dictating without cloud migration or subscription steps
Fixed
More reliable paste into the focused app — Parrot pastes with ⌘V and falls back to typing if paste doesn’t land
v0.2.0
July 18, 2026
Improved
Dictation feels dramatically faster — text lands at your cursor with less wait after you release the hotkey
Higher first-pass accuracy, especially on names, jargon, and technical terms
Setup is simpler: download once in the app and start dictating — no extra tools to install
Added
Live transcript in the floating HUD while you hold the hotkey — see your words form as you speak
Smarter vocabulary that learns from how you write and suggests terms worth saving
Per-app writing polish so Slack, docs, and code can each sound the way you want
Drop an audio file onto History (or pick one) for a full transcript
v0.1.6
May 7, 2026
Added
Custom vocabulary now biases the transcriber itself — Whisper, Deepgram, and ElevenLabs get hinted with your saved terms so words like “Tauri” or “Gujarati” come through correctly the first time
Optional “Use when” context per vocabulary word — Parrot can spell “Tauri” in software contexts and leave “tory” alone in political ones
Periodic update checks so you’re prompted when a new version ships
macOS microphone permission status now surfaces in onboarding and settings, with a banner when access is missing
Improved
The HUD chip now expands from the center of the orb, not from a corner, so it stays visually anchored to where the icon was
Fixed
Empty dictations (silence, accidental hotkey taps) no longer create history entries or paste blanks
Very short local-whisper inputs return an empty transcript instead of surfacing an error
Hero site favicon now resolves (migrated from .ico to .png)
v0.1.5
May 3, 2026
Added
Manual Ollama setup fallback — if the automatic install fails, the wizard now walks you through installing Ollama yourself and resumes from where it left off
Fixed
Onboarding now persists setup mode before saving your profile, so local-setup and tour steps no longer lose track of the local user
Improved
Cleaner onboarding transitions — dropped the route cross-fade in favor of stable layout heights and instant step changes
v0.1.4
May 3, 2026
Fixed
Local setup no longer hangs on “Installing Ollama…” — installer now elevates through a native macOS password prompt instead of a stuck sudo call
Setup correctly detects an existing Ollama install — including users on Ollama.app who never ran “Install command line”
Improved
Local mode no longer requires Homebrew — CMake (a build-from-source dependency) is the only prerequisite, and the official cmake.org installer works fine
Default dictation hotkey shown as fn across the marketing site and download page
Blog posts now render markdown tables properly
v0.1.3
May 2, 2026
Added
Floating HUD orb shows dictation status and can be dragged anywhere on screen
History table with grouping by date, one-click copy, and delete with confirmation
Delete a dictation removes it from local storage and the cloud (including S3 audio)
Improved
Local Ollama cleanup now respects your custom vocabulary
Reuses an existing Ollama instance, warms up the model, and cleans up child processes on exit
Tuned history group header contrast for readability
Fixed
Paste now uses CGEvent for Cmd+V instead of osascript (more reliable)
Audio buffer is cleared before each recording to prevent stale data
Microphone permission prompt now appears correctly on macOS
v0.1.2
April 30, 2026
Added
Competitor comparison pages and Compare link in the marketing site nav
Four new blog posts
Open Graph images and SEO metadata across marketing pages
Enhanced download page with FAQ
Improved
Subscribe flow replaces the old waitlist (new /subscribe route and component)
API consolidated to PostgreSQL (SQLite support removed)
Fixed
PATH resolution for GUI-launched apps on macOS
Removed
Framer Motion animations from the setup wizard
v0.1.1
April 29, 2026
Added
Download page dynamically fetches the latest release from GitHub
Improved
CI now ships Apple Silicon-only builds; download page updated to match
v0.1.0
Initial releaseApril 29, 2026
Added
Native macOS menu bar app — press a hotkey, speak, and text appears at your cursor
In-process Whisper transcription with Metal acceleration (whisper-rs)
Customizable dictation hotkey with fn-key support
Silent background auto-updater with sidebar banner
Accessibility permission step in onboarding
Centralized error handling with Sonner toast notifications
GitHub Releases CI workflow for Tauri builds
Marketing site, blog, and download page
Improved
Local-first launch — free local mode for life, no account required
Simplified onboarding with local-first default
Friendlier model names in the setup wizard
Safer copy-paste that handles missing accessibility permission
Cleanup prompts hardened against instruction-following
Removed
BYOK API keys section, cloud-setup route, and pricing page (cloud mode coming later)
Standalone whisper.cpp server (replaced by in-process transcription)
Coming up
What we're working on
Even snappier dictation and smarter cleanup
Cross-device sync when you want it
Windows support
Multi-lingual transcription with automatic language detection
Parrot