Offline Speech to Text 0.8.0

Version 0.8.0 makes long dictation smooth again while keeping speech private and on-device.

Ordinary recordings now stay in one continuous, contextual model run. Natural pauses no longer create artificial sentence boundaries. For very long input to non-Whisper models, OSTT uses bounded overlapping windows and removes overlap only when shared words align safely; if they do not, it preserves both hypotheses instead of risking lost speech.

Troubleshooting is easier too. Diagnostic logging remains opt-in, bounded, and content-free. When enabled, Settings can save the log as a text file, share a temporary read-only snapshot through Android's Sharesheet, or clear it immediately. Spoken audio, transcripts, partial hypotheses, prompts, dictionary entries, and editor text are never written to diagnostics.

The APK remains compact and does not bundle a large speech model. Existing models survive normal updates; a fresh install can download the recommended model or import one manually. Recognition remains offline after setup.

• Install or update from Zapstore
• Download the verified arm64 APK
• Read the source and integration guide

APK SHA-256: d144b928e95869dd557fea79cb391fdc6367cec972b0c1ff2dada57e3c70bb64