debpalash/OmniVoice-Studio v0.3.0

OmniVoice Studio v0.3.0

on GitHub

Added

• Frameless dictation widget. Global dictation upgraded from an in-app FAB to a true OS-level floating widget that hovers over any application. Transparent, decorations-free, always-on-top secondary Tauri window activated by ⌘+⇧+Space. Auto-hides 2.5 s after a successful paste.
• Standalone CaptureWidget component. Refactored CaptureButton into CaptureWidget, running on an isolated route (/?window=widget).
• Social preview image. Added social-preview.png for GitHub SEO.

Changed

• README overhaul. Compact 3-column feature grid, reorganized Quickstart (one-command install, Docker, Desktop App tips), updated comparison table, roadmap, and footer CTA.
• Docker Compose profiles are mutually exclusive. CPU service now requires --profile cpu (was the implicit default). Prevents port 3900 conflict when running --profile gpu. Usage: docker compose --profile cpu up or docker compose --profile gpu up.

Fixed

• Docker GPU detection false negative. Preflight reported "No compatible GPU detected" inside Docker containers because nvidia-smi isn't present in the PyTorch base image. The GPU probe now falls back to torch.cuda.is_available() and torch.cuda.get_device_name(), correctly showing CUDA as available in containerized deployments.

Windows x64 artifacts

9e3b598726efb19c3382d6ec6f7faf3b31beb0ca09bd451e3b60b3fef10b38ea *OmniVoice Studio_0.3.0_x64_en-US.msi
62268e015315c6bd362faad37537f38325f0d8f571b64713c9e94eeeefc45949 *OmniVoice Studio_0.3.0_x64_en-US.msi.sig