What's New
- Added Balanced dictation profiles for English and multilingual use.
- Clarified recognition profile tiers across speed, balance, realtime, and accuracy modes.
Improvements
- Improved performance readiness on more supported Windows PCs.
- Reduced setup friction for devices that can use enhanced performance paths.
Fixes
- Improved guidance for recommended recognition profiles.
- Made fallback behavior more conservative on weaker or unsupported systems.
Known issues
- Performance improvements still depend on device compatibility and system readiness.
- Some Windows PCs may continue using the standard path when that is the more reliable option.
What's New
- Upgraded live dictation overlay.
- Added a richer 3D voice sphere status view.
- Added a privacy-focused recording option.
Improvements
- Faster spoken-language switching.
- Quicker access to favorite languages.
- Clearer model-match guidance and overlay feedback.
Fixes
- Validated the 1.0.16 to 1.0.17 update path.
- Improved post-update recording and text insertion.
- Promoted 1.0.17 as the stable Windows x64 release.
Known issues
- This release is for Windows x64.
- Some features, including update checks, require the installed app experience.
What's New
- Added clearer model setup diagnostics.
- Added support bundle export in Diagnostics.
- Improved optional GPU acceleration setup.
Improvements
- Simplified the Private Mode toggle.
- Expanded download diagnostics and validation coverage.
- Refreshed Windows update feed and installer metadata.
Fixes
- Hardened checks for model files and acceleration packs.
- Added deeper acceleration startup and fallback diagnostics.
- Promoted 1.0.16 as the stable Windows release.
Known issues
- GPU acceleration still depends on supported hardware, drivers, and runtime components.
- Support bundles exclude audio, transcript text, and raw telemetry payloads.
What's New
- Added clearer recommended-model context.
- Clarified fast vs. quality-focused model details.
- Expanded diagnostics for model choice and performance.
Improvements
- Improved acceleration selection and startup validation.
- Made cross-session dictation performance easier to interpret.
- Reduced stuck model download states in the Models view.
Fixes
- Improved diagnostic completeness for support reviews.
- Reduced duplicate audio stop details.
- Promoted 1.0.15 as the stable Windows release.
Known issues
- Larger models can still finalize more slowly on lower-power hardware.
What's New
- Improved the update path for older Windows builds.
- Aligned update prompts with the latest stable release.
- Clarified the recommended Windows installer path.
Improvements
- Strengthened supported-version availability signals.
- Improved version guidance across app, update flow, and website.
- Prepared the update path for the next stable release.
Fixes
- Reduced outdated update guidance cases.
- Improved restart and update flow from older builds.
What's New
- Improved ready-to-dictate startup.
- Better automatic performance-path selection.
- Clearer readiness and support signals.
- Strengthened the update path for future releases.
Improvements
- Reduced startup friction from network and device checks.
- Made dictation readiness more resilient.
- Smoothed performance across Windows devices.
- Improved support information clarity without exposing private content.
Fixes
- Reduced startup delays before dictation.
- Improved transcription mode selection consistency.
- Aligned app, downloads, and release notes with the current build.
What's New
- Improved support for compatible Windows hardware.
- Expanded out-of-box setup coverage.
- Added a readiness view for transcription status.
- Made support details easier to share privately.
Improvements
- Reached ready-to-transcribe state faster.
- Improved fallback to standard transcription mode.
- Clarified system readiness status messages.
- Simplified copying support details.
Fixes
- Reduced startup slowdowns before transcription.
- Improved switching reliability between transcription modes.
- Smoothed installation into the current release.
What's New
- Improved new-version delivery reliability.
- Made the update experience more consistent.
- Aligned app, download page, and release note version info.
Improvements
- Improved update availability detection.
- Stabilized public downloads for early-access users.
- Strengthened the release flow for future updates.
Fixes
- Fixed incorrect update availability guidance.
- Improved product guidance for installed versions.
- Cleaned up the update notification layout.
- Refined the app control screen layout.
- Improved anonymous usage insight reliability when enabled.
- Reduced duplicate first-run signals.
- Hardened queued improvement signal handling.
- Added privacy-conscious usage insights without collecting dictated text or audio.
- Improved dictation session measurement.
- Tracked successful and interrupted dictation outcomes more clearly.
- Kept privacy controls explicit around analytics, transcripts, and audio.
- Prepared the Windows installer and update package for 1.0.9.
- Smoothed push-to-talk between back-to-back dictations.
- Scoped Escape handling to active Sonavi cancellations.
- Clarified missing Windows microphone permission states.
- Reduced unnecessary reloads during model and language switching.
- Improved recording status feedback while processing finishes.
- Prepared the Windows installer and update package for 1.0.8.
- Improved the first-run experience.
- Strengthened privacy-first quality signals for opted-in analytics.
- Confirmed that transcripts, audio, and personal identifiers are not sent.
- Improved temporary offline handling for eligible quality data.
- Prepared a fresh Windows installer and update package.
- Added smarter spoken-language matching.
- Clarified preferred language vs. active language behavior.
- Improved fallback to automatic detection when a language cannot be applied.
- Improved spoken-language control accessibility.
- Strengthened update delivery checks.
- Reduced background audio while you speak.
- Added a simple control for the behavior.
- Restored audio smoothly after speaking ends.
- Cleaned up the overall dictation flow.
- Improved in-app guidance for important updates.
- Clarified update progress during prepare, download, and install steps.
- Added a clearer fallback download path when in-app updates are unavailable.
- Refined update suggestion visibility.
- Strengthened the overall release process.
Known issues
- Some uncommon or development installs still use the website download path.
- If an in-app update cannot be completed, Sonavi will guide you to the external download path.
- This release focuses on update reliability rather than major transcription features.
- Improved Windows release delivery reliability.
- Added stronger pre-publish release checks.
- Stabilized downloads and updates for larger Windows packages.
- Clarified Settings access to version, update status, and release notes.
- Clarified early-access policy and release access.
- Published the 1.0.3 desktop release metadata.
Known issues
- Early-access users can continue using the current public build while licensing evolves.
- Improved transcription UX, dashboard usability, and release infrastructure.
- Strengthened update and app-control handling.
- Expanded home, history, copy, scrolling, and layout refinements.
- Added recoverable Escape cancel during dictation.
Known issues
- Recoverable Escape restores captured transcript but does not resume recording.
- Dashboard and home-screen refinements may continue to evolve.
- Released the first tagged Sonavi desktop version.
- Established the Sonavi Windows app identity.
- Shipped the initial desktop UI and packaging baseline.
- Laid the foundation for future transcription and update work.
Known issues
- This early baseline did not yet include later usability, update, or transcription improvements.
- Launched Sonavi as a public early-access Windows app.
- Published the first public download path and support resources.
- Established Sonavi as a privacy-first dictation product.
- Positioned the installer as the starting public build.
Known issues
- The download endpoint may evolve during the current early-access cycle.
