v0.3.2

What's New

VoiceGarden.UI — Avalonia Configuration App

• Replaced the old C++ Installer with a modern Avalonia UI app
• Download and manage SherpaOnnx offline models
• Configure 20+ cloud TTS engines (Azure, Google, OpenAI, ElevenLabs, Polly, and more)
• Register/unregister 32-bit and 64-bit SAPI adapter DLLs
• Preview voices before installing
• Full CLI mode for automation

SherpaOnnx Offline TTS

• Supports Kokoro, Piper, MMS, VITS, and Matcha model types
• Auto-detection of model type from directory contents (voices.bin, vocoder.onnx)
• MMS models download individual files from HuggingFace directories
• Auto-extraction of orphaned .tar.bz2 archives on rescan
• Download progress with file size and percentage

Cloud Engine Support

• Azure Cognitive Services (REST + Speech SDK)
• Edge browser voices (WebSocket)
• Google Cloud TTS, OpenAI, ElevenLabs, AWS Polly, Cartesia, Deepgram
• 15+ additional cloud engines via generic HTTP TTS

SAPI Voice Management

• Model type-aware promotion to HKLM (Kokoro=type 2, Matcha=type 1, VITS=type 0)
• .reg file import for fast, reliable elevation (no slow exe relaunch)
• 32-bit and 64-bit COM registration (WOW6432Node aware)
• Grid 3 / System.Speech compatibility (HKLM token promotion)

Build & CI

• Full GitHub Actions CI pipeline (12 jobs, all green)
• MSI setup + setup.exe bootstrapper
• Per-platform release ZIPs (x86, x64, ARM64)