Now — active
These are items currently in active development or finishing work.
- Bug fix: consecutive transcription spacing — automatic leading space between back-to-back dictations in the same text field (evoglyph#1713).
- Bug fix: compound name digit merging — extending LCP compound-name licensing to cover teens, tens, hundreds, and thousands (evoglyph#1331).
- Bug fix: injection routing — capability-first element routing to replace the bundle-ID table for apps with mixed native and web content (evoglyph#1465).
Next — planned
These are items planned for the near term and have active tracking issues.
-
Voice profile — speaker-personalization feature that lets
evoglyph ignore other speakers nearby during dictation. The storage and enrollment
layers are built; the pipeline gate (
PersonalizedVadGate) ships post-launch. README-documented as in development (evoglyph#1724). - Custom command layer — user-defined spoken phrases that expand to saved text snippets, generalizing personal-values insertion (evoglyph#1516).
- Per-app cleanup prompts — different AI cleanup behavior per target application or mode (evoglyph#1153).
- Corrections feedback loop — the corrections table collects your manual edits to feed back into vocabulary and cleanup tuning (evoglyph#1439).
Later — investigation stage
These are items under investigation. They are not committed and may not ship.
- Multi-language support — investigation into non-English language support (evoglyph#1789). The underlying speech engine has a 25-European-language variant; evoglyph’s full pipeline — transcription and AI cleanup — is validated on English only. Non-English support requires end-to-end validation before we can claim it.
- Grammatical-correction mode — a toggleable mode in the cleanup layer to apply or suppress grammatical corrections (evoglyph#1774).
- Faster pasteboard injection — reducing the pasteboard-restore ceiling from 700 ms to 300–500 ms for apps that use the pasteboard path (evoglyph#1780).