Changelog

Purchases and AI Text Correction

• Purchases run in the background: the app stays fully responsive while the App Store completes a transaction. Windows still move, and the interface does not wait.
• Redesigned purchase screen: it follows the app theme, is available in all 18 supported languages, and describes each step of the App Store sequence as it happens.
• AI Text Correction favors your wording: it concentrates on punctuation and spelling rather than rewriting what you said.
• AI Text Correction in Chinese, Japanese and Thai: languages written without spaces between words are now supported.
• Voice Activation reports its state: it watches the detection engine and tells you what it is doing.
• Performance and stability improvements.

Files, Video and Hands-Free Recording

• Transcribe Several Files at Once: Pick a set of recordings and VocoSafe AI works through them one after another, showing which file it is on and letting you stop at any point.
• Video and iPhone Voice Memos: M4A and AAC now open alongside WAV, MP3, FLAC and OGG, and the audio is pulled straight out of MP4, MOV, AVI and MKV. No converting first.
• Save Transcriptions as .txt: Tick the box on the Files tab, choose a folder once, and every transcription is written there as its own text file. Off by default, and dictation is never written to disk.
• Long Recordings Fixed: Interviews, lectures and meetings past ten minutes previously stopped partway through with an unclear message. They now transcribe end to end.
• Global Shortcut: Optional Option-Shift-D starts and stops recording from any app, without bringing the window forward. No extra system permissions are needed.
• Menu Bar Record Button: Optional one-click record button next to the VocoSafe AI icon, with the mode chosen from the menu.
• Help Section: A built-in guide covering dictation, files, purchases and activation, translated into all 18 interface languages.
• Rate the App: Available from the menu bar and the About screen.
• Queued Start: Pressing record while a model is still loading now queues the recording instead of reporting that no model is selected.

Accuracy and Performance Update

• Fewer Invented Phrases: Reworked filtering catches the stock phrases speech models produce from silence, breathing and background noise, so they no longer land in your text.
• Much Lower CPU Use: Background voice activation dropped from around 66% of a core to roughly 1%, which means a quieter fan and longer battery life.
• Window Resizing: Fixed a case where the window could not be resized properly.

Pricing & Convenience Update

• Local Pricing: Prices are now shown in your local currency and storefront.
• Launch at Login: New option in Settings to automatically start VocoSafe AI when you log in.
• Detailed Logging: New optional detailed logging mode to help diagnose issues faster when contacting support.
• Stability: Minor bug fixes and improvements.

Initial Public Release

• Private On-Device AI: Complete offline speech-to-text processing for total privacy.
• Smart AI Post-Processing: Integrated local Qwen 2.5 model, running on your Mac, for real-time grammar and punctuation correction.
• System-Wide Integration: Automatically types text directly into any active application.
• Advanced VAD: Professional Voice Activity Detection to filter background noise and pauses.
• Multilingual Support: Automatic detection and transcription for 18+ major languages.
• Localized UI: Full interface translations for 18 languages.
• Modern Design: Liquid Glass UI optimized for macOS 15+ and Apple Silicon.
• Transparent Pricing: The trial call-to-action and fine print clearly show the post-trial price and billing period before you subscribe.
• Smarter Purchase Flow: Users with an active subscription no longer see the purchase screen unnecessarily.

First Beta Build

Kept here for the record. This build went to TestFlight testers only and was never released on the Mac App Store. It carried the core of what shipped a month later in 1.1: on-device transcription, the local grammar model, system-wide typing, voice activity detection and the 18 language interface.