Offline Speech to Text 0.13.1

This maintenance release simplifies diagnostic validation and marker writes. Speech recognition, supported models, settings and privacy defaults keep their existing behavior. No new model is included.

It includes the features introduced in 0.13.0: optional early processing for long voice-keyboard dictation, clearer explanations for empty results and separately warned, one-hour troubleshooting. Early processing uses the existing overlapping decode windows, retains the recording and delivers one assembled result after Stop. The setting starts off. Any reduction in waiting after Stop remains unmeasured, and processing during recording may increase heat.

Detailed troubleshooting records timing, process CPU and thermal/battery context. Audio, transcripts and dictionary contents stay outside diagnostics. It requires explicit consent, expires after one hour and is exported only when you choose to share it.

Existing model downloads and settings are preserved when updating the production app.

• Update from Zapstorezapstore.dev
• Sourcegitworkshop.dev

• Verified APKblossom.primal.net