v0.5.8

What's New in v0.5.8

RustTtsWrapper 0.3.16

• Online model voice support — OpenAI, Google, and other cloud engines now support voice listing (GetVoices), previewing, and promotion to SAPI
• All 22 engines confirmed working with dynamic credential discovery

Dynamic Engine Discovery

• No more hardcoded engine lists — EngineDefinition.DiscoverAll() calls TtsClient.ListEngines() to discover all available engines with their credential requirements at runtime
• Adding a new engine in rust-tts-wrapper automatically makes it appear in VoiceGarden UI
• Credentials built dynamically from each engine's credential_keys — no engine-specific switch statements

Credential Management Rework

• Single source of truth — API keys entered only in Configure Credentials (removed from Configure Voices)
• Verify button — each provider has a Verify button that tests credentials and shows ✓/✗ with voice count
• Configure Voices reads from registry — no duplicate key entry; dropdown shows only enabled providers

Screen Reader Accessibility

• Live regions — status updates, download progress, and verification results announced automatically via LiveSetting=Polite
• View change announcements — switching between Configure Voices, Configure Credentials, etc. is announced
• Onboarding page announcements — "Page 2 of 3: Getting Started" announced on navigation
• Password field descriptions — HelpText on all API key fields ("Enter your Azure API key")
• Control naming — all buttons, checkboxes, textboxes have descriptive accessible names
• ListBox items — model/voice items describe name, language, and status

Manage Models Crash Fix

• Async catalog loading — 1300+ model catalog now loads off the UI thread with Parallel.ForEach, preventing UI freeze and crash

Single Instance Guard

• Mutex prevents duplicate instances — launching a second VoiceGarden.UI brings the existing window to the foreground

Avalonia 12.1.0

• Upgraded from 11.2.7 to 12.1.0 — compiled bindings, performance, accessibility improvements
• Removed deprecated Avalonia.Diagnostics package

CI Fixes

• Stable download URL — https://github.com/AACTools/VoiceGarden-SAPI/releases/latest/download/VoiceGarden-release-layout.zip

---

What's New in v0.5.2

Grid3 & System.Speech Compatibility

• Cloud voices only via registry promotion — Azure/Edge voices are no longer enumerated dynamically. Only voices promoted via "Install Selected" (HKLM registry tokens) appear in SAPI
• Voice promotion fixed — PromoteElevated now uses reg.exe import with UAC elevation
• Voice enumerator deadlock fixed — double mutex lock caused crash on first enumeration
• No more voice flood — only promoted voices appear in SAPI

SSML & Speech Markdown via Web Speech API

• PUA sentinel — embed SSML or Speech Markdown via Unicode Private Use Area characters
• SSML normalization — adapter injects required <speak> attributes when missing
• Speech Markdown syntax — use (text)[rate:"slow"] format

Chrome & Edge Support

• Speech_OneCore registration — voices appear in Chrome and Edge when promoted

RustTtsWrapper 0.3.14

• SSML pass-through, Speech Markdown converter, voice selection, GetVoices crash all fixed

MMS Model Data Fix

• 1143 MMS models now searchable with proper language names

Localization (36 Languages)

• Full UI translation, automatic OS language detection, RTL support

Onboarding & Installer

• 3-page wizard, version-gated, installer error fixes

---

What's New in v0.4.0

Major Architecture Changes

• RustTtsWrapper Integration — complete transition to Rust-based TTS wrapper
• Removed legacy DotNetTtsWrapper and EngineConfig components

Breaking Changes

• Removed Components — DotNetTtsWrapper and EngineConfig no longer supported
• Minimum Requirements — Windows 10+