doing.toolsDoing

Doing

Changelog.

v0.1.0

February 22, 2026

Latest (build 56)

  • App renamed from Toucan to Doing: new executable, config directory (~/.config/doing), transcript directory (~/Documents/Doing), and Sparkle feed. Existing settings, API keys, skills, and transcripts are automatically migrated on first launch.
  • Fix Option+Space hotkey stuck in heavy apps (Teams, Slack): Re-enable CGEvent tap when macOS disables it due to processing timeout, and reset trigger state so the pip doesn't get stuck. Also improved pip animations: instant visual feedback on key release, seamless recording-to-processing transition for cloud providers, and snappier dismiss after paste.
  • Smart hotkey recommendations: Detects external keyboards (Logitech, etc.) and promotes Option+Space as the default trigger since fn keys on third-party keyboards don't reach macOS. Removed false conflict warnings for Ctrl+V/Z/X/A (Windows conventions, not macOS).
  • Onboarding discount code hints: Improved discount code entry and license key flow during onboarding.
  • Fix main thread hangs on recording start: Moved AVAudioEngine and AVAudioRecorder init to background queues — these Core Audio calls were blocking the UI for 2+ seconds. Extracted audio level math to ToucanCore with 23 tests.
  • YOLO toggle polish: Toggle pill is now near-black when engaged (light mode) for clearer on/off state
  • Loading phrases: 20 rotating phrases (up from 10) cycling every 3s for more variety during model load
  • "Hold fn to speak": Status card, menu bar, and hero keycap now say "to speak" instead of "to record"
  • YOLO Mode toggle: Dedicated toggle row inside the compact status bar to enable auto-return (press Return after pasting). Includes pill toggle, hover tooltip, confirmation feedback on toggle, and ⏎ indicator in the mouse pip during recording when active.
  • Transcript inbox redesign: Main window now shows transcripts at all times — recording, transcribing, and model loading are compact status cards instead of full-screen takeovers. Hero keycap only appears when no transcripts exist. Removes pattern background and view mode toggle for a cleaner layout.
  • Smart paste: Toucan no longer tries to paste into its own window (which caused a beep); falls back to copy-only when the main window is focused.
  • YOLO Return Mode: Auto-return setting now accessible in Settings > General (previously only in menu bar). Automatically presses Return after pasting — perfect for chat apps.
  • Benchmark mode: Side-by-side comparison of all transcription providers in Advanced settings. Shows local/cloud badges, batch/streaming mode, speed, cost, and transcription text for every provider — including unavailable ones with reasons why.
  • Apple on-device skills: On macOS 26+, skills (Cleanup, Formalize, Summarize) work out of the box with zero API keys using Apple's Foundation Models — free, private, offline. Includes defensive pre-check for the 4096 token context window limit to avoid app hangs.
  • Apple Transcription provider: On macOS 26+, users can select "Apple" as a transcription provider using the new SpeechAnalyzer framework — free, on-device, zero dependencies
  • Fix app hangs: Moved CoreAudio device queries and SMAppService checks off the main thread to prevent UI freezes during Bluetooth device changes and settings rendering
  • "All Apps" default skill chain: Set a default skill chain that runs on every transcription. App-specific skills override it. Replaces the old auto-run skill with a unified, more powerful approach.
  • App Skills: Automatically run skills based on the target app (e.g., Formalize for Slack, Cleanup for Ghostty). Configure per-app skill chains in Settings; skill-processed text is pasted instead of raw transcription.
  • Optional auto-paste: Toggle auto-paste on/off in Settings > General; when off, transcriptions copy to clipboard only. Accessibility permission is only needed when auto-paste is enabled.
  • Increased free tier from 26 to 100 transcriptions before a license is required
  • Parakeet attribution: Settings now credits NVIDIA Parakeet TDT model with links to model and CC-BY-4.0 license
  • Streamlined onboarding: Reduced from 8 steps to 7 by removing redundant final screen; trial and purchase both complete onboarding immediately
  • Lemon Squeezy checkout: Buy Toucan button opens hosted checkout for one-time purchase with license key delivery
  • Lemon Squeezy licensing: 7-day free trial with license key activation for $39 one-time purchase (3 devices)
  • Smooth volume ducking: Volume fades out/in smoothly instead of abrupt changes when recording
  • Auto-quiet media: Automatically lowers system volume while recording so music/videos don't interfere, restores when done (Settings > Audio)
  • Inline mic picker: click the StatusCard in ready state to switch input devices without opening System Preferences
  • Addeed in-app link to changelog
  • AssemblyAI LLM Gateway: Skills can use Claude Sonnet via AssemblyAI's LLM Gateway with your existing API key (requires billing)
  • Parakeet transcription: New default local provider using NVIDIA's Parakeet TDT model (~10x faster than WhisperKit, lower WER)
  • Fun loading messages: Playful rotating phrases while the model wakes up ("Folding ears", "Herding floating point sheep", etc.)
  • Browse Previous Days: Button to open your transcripts folder directly in Finder
  • First-run companion window: Practice window opens after onboarding to help new users get started
  • Redesigned onboarding: Streamlined first-run experience with reordered steps and Parakeet status indicator
  • Auto-return: Option to automatically press Return after pasting transcription (Settings > General)
  • Mic picker: Switch input devices from the top bar when multiple mics are connected
  • Mouse pip: Small floating indicator follows your cursor during recording/transcription, hides only when main window is on top
  • Bold visual redesign: New colorful interface with 8 selectable themes (System, Coral, Cream, Sky, Mint, Coffee, Aqua, Charcoal)
  • In-app feedback: Send feedback directly from within Toucan — type or speak your thoughts
  • Filler word removal: Automatically strips um, uh, er, ah from transcriptions — customize the list in Settings > Transcription
  • Crash reporting: Automatic crash reports via Sentry help us fix issues faster (no personal data collected)
  • Re-running setup shows test window: "Try It Out" companion window appears again after re-running onboarding
  • Corrected Parakeet download size: ~480 MB (was incorrectly shown as ~2.6 GB)
  • Separate model storage: Parakeet and WhisperKit models shown separately in Settings > Advanced > Local Data
  • Clickable provider indicator: Click the provider name in the main window to open Transcription settings

About 0.1.0

  • Voice transcription with push-to-talk — hold fn to record, release to transcribe
  • Three transcription engines: WhisperKit (local/private), OpenAI, or Gemini
  • Skills system for post-processing with LLMs (cleanup, formalize, summarize, emoji)
  • Daily transcript history saved as Markdown files
  • Pick your recording trigger: fn, Option+Space, or Right Command
  • Automatic updates via Sparkle