← Back to homeChangelog
Release notes and version history for WalkAITalkie.
v2.0.0
- Complete v2 redesign: new 5-tab settings window, streamlined menu bar UI, and modern design language
- Parakeet speech engine: ANE-native STT with 110x faster processing for 25 European languages
- App Store readiness: Pro feature gating, subscription JWS verification, dual-environment support
- Subscription JWS reporting to backend for immediate credit grants
- Removed hardcoded prices from App Store texts across 11 locales
- Apple trademark compliance: removed forbidden keywords from all locales
- Pro Lifetime IAP removed from App Store build (direct distribution only)
- Onboarding step 3 redesign: parallel model loading with try-it textfield
- Language mismatch detection with automatic switch suggestions
- Release automation with Sparkle integration and CI pipeline
Download v2.0.0 →v1.2.0
- Smart memory management: RAM-aware model unloading with automatic recovery and lightweight model recommendations
- Translation quality improvements: dedicated translator prompt, quality indicators, and cloud recommendation UX
- Qwen3.5 thinking mode fix: eliminates slow generation caused by unnecessary reasoning tokens
- Dictionary term filtering to prevent hallucinated word insertions
- LLM download confirmation sheet when switching processing styles
- Email verification required for credit purchases (security hardening)
- Managed cloud credits end-to-end improvements with action banners and pre-recording gates
- App Store review prompt at usage milestones
Download v1.2.0 →v1.1.5
- LLM download confirmation sheet: choose model before downloading when switching to Cleanup or Style modes
- Auto-switch to already-downloaded LLM model when switching styles
- RAM-aware model recommendation in download sheet
Download v1.1.5 →v1.1.4
- Unified account auth: single email field for register & login
- Hide style sections in Raw mode (Pinned, All, Custom Styles)
- Contextual performance card adapts to on-device vs cloud backend
- Style suggestions hidden in menu bar when using Raw mode
- Memory warning only shows for on-device processing
Download v1.1.4 →v1.1.3
- Offline detection: instant fallback to on-device when cloud is unavailable
- Daily credit limit reset time shown in error messages
- Account device tracking for multi-device management
- API mode localization completed for all 11 supported languages
- 51 new integration tests for backend API endpoints
- Graceful fallback for cloud unavailability with user-friendly toasts
Download v1.1.3 →v1.1.2
- LLM fallback indicator: preview shows '⚠ raw' when model was unavailable
- Stronger output language prompts for GPT-4.1 Nano/Mini translation compliance
- Swift 6 Sendable compliance for animation handlers and onboarding closures
- Pin MLXLLM to main branch for latest Qwen3.5 support
- Fix #if APPSTORE conditional compilation block structure
- App Store entitlements fix: sandbox + audio + network + file access
Download v1.1.2 →v1.1.1
- Fix Carbon hotkey registration with improved F5 fallback (App Store)
- Fix VAD auto-stop on silence not triggering
- Suppress Whisper hallucinations on VAD silence detection
- Credits UI: clearer zero-balance state with 3-tier coloring
- Fix BYOK backend picker flicker when switching providers
- Improved API key observer for seamless engine switching
Download v1.1.1 →v1.1.0
- Cloud API Mode: Use OpenAI or Anthropic APIs as transcription and LLM backends
- Credit system for managed cloud usage (no API key needed)
- Network connectivity monitor with inline warnings
- Security hardening: API key validation, response size guards, JWT expiry checks
- Shared retry logic with exponential backoff across all API engines
- IAP credit purchases via App Store
- App Store Server Notification v2 webhook support
Download v1.1.0 →v1.0.2
- Fix App Store hotkey permission flow
- International keyboard layout support for hotkeys
- Badge readability improvements
- Localized onboarding, paywall, and permissions strings for all 11 locales
- Full IAP text localization
Download v1.0.2 →v1.0.0
- Push-to-talk speech recognition with WhisperKit (on-device)
- LLM post-processing with MLX Swift (grammar, punctuation, formatting)
- Custom dictionary for domain-specific vocabulary
- Hold and toggle hotkey modes
- Transcription history with search
- 11 interface languages
- Free tier with Whisper Tiny + Pro unlock for larger models and LLM
- Animated hero demo on the landing page
- Direct download (DMG) and Stripe checkout for Pro