Private prototype · hosting planned
Kalimni — Egyptian Arabic voice to text
A voice-input prototype that transcribes Egyptian Arabic and rewrites the transcript into a chosen style.
The problem
Speaking naturally in Egyptian Arabic should produce text that can be copied into a note, message or task.
My contribution
Built the RTL recording interface and a backend pipeline for transcription, cleanup and reformatting an existing transcript.
Engineering decisions
- Keep ASR and text cleanup as separate stages.
- Reformat the existing transcript without transcribing the recording again.
- Keep provider calls on the backend and expose a simple browser interface.
Results & evidence
- Implemented audio processing, style discovery and transcript reformatting endpoints.
- The source remains private. Public hosting and live provider inference still need verification before an open demo.
