The installer path stays drag-and-drop.
No Xcode. No Swift toolchain. Download the DMG, drag Scrivora.app into Applications, and grant macOS permissions when prompted.
v0.4.1 release notes · updated June 20, 2026
Realtime voice. Private by default.
Scrivora turns speech into text locally and writes back into the app you are using. Download the preview DMG, drag Scrivora into Applications, and start from the menu bar.
What changed
If macOS says the app is damaged.
Run sudo xattr -rd com.apple.quarantine "/Applications/Scrivora.app", then open Scrivora again. This removes quarantine from Scrivora only.
Check the downloaded DMG.
Download SHA256SUMS.txt from the release and compare it with shasum -a 256 ~/Downloads/Scrivora-0.4.1-preview-unnotarized.dmg.
Speech lands as text in under a second.
The homepage now says the product loop plainly: hold Control, speak naturally, and Scrivora writes back into the focused app. Longer notes are processed continuously instead of waiting for a final cloud job.
Voice speed is now explained with real math.
The homepage compares average typing around 40 WPM with normal English speech around 150 WPM, including the 500-word time saved calculation and a clear note that 250 WPM typists are outliers.
No account. No card. No cloud speech API.
The site now leads with the real promise: everything private, nothing leaves your device for core dictation.
The landing animation now shows voice becoming text.
The old static signal was replaced with a live local path: pulsing waveform, moving voice-to-text line, and word-by-word transcript output.
Menu bar and copy fallback are release-critical.
Quick access to latest transcript copy, model switching, History, update checks, and clipboard fallback remain part of the release scope.
Release foundations
Third-party credits live beside the release.
Model credits, notices, and source-release notes stay visible instead of being buried in app internals.
Edit it if you want to.
The source is available for review, changes, and local builds. The prebuilt DMG is the normal install path.
Model downloads stay explicit.
FluidAudio Parakeet V3 is credited clearly, and model downloads remain user-controlled.