LexaWrite Changelog - February 2026

Here’s what’s new in LexaWrite this month.

What’s New

Multi-Step Onboarding with Model Download

New users now get a guided onboarding experience that walks through microphone permissions, model selection, and a first test dictation. The model download happens during onboarding so LexaWrite is ready to use the moment setup completes.

The floating recording indicator has been refined - smaller, subtler, and better positioned. It stays out of your way while clearly showing when recording is active.

Analytics Foundation

We’ve added lightweight, privacy-respecting analytics to understand which features are used most. No personal data is collected. No audio is ever transmitted. This helps us prioritize what to build next.

What’s Improved

Faster model loading - the Whisper model now loads into memory faster on launch, reducing the delay before first dictation
Better audio format conversion - improved the 48kHz → 16kHz resampling pipeline for cleaner audio input to Whisper
Transcript history search - searching past transcripts is now instant, even with hundreds of entries

What’s Next

We’re working on:

Style matching improvements - better context detection for adapting transcription style
Keyboard shortcut customization - for users who prefer a different key than Fn
iOS improvements - bringing the best of the macOS experience to iPhone and iPad

Numbers

Current Whisper.cpp version: 1.7.5
Supported models: Tiny, Base, Small, Medium, Large
Supported languages: 99
macOS requirement: 12+
Architectures: Apple Silicon (native) + Intel (Rosetta)

Have feedback? We’d love to hear it - get in touch.

Download LexaWrite if you haven’t already.