Zirek has officially joined the Open Source Collective as our fiscal host.

If you are new here: Zirek builds local, offline neural text-to-speech and screen-reader support for languages commercial vendors overlook—starting with Kurmanji Kurdish, a language spoken by 15–20 million people without production-ready accessibility tooling.

Today, this is a solo-maintainer project built on evenings and weekends. An NLnet Foundation grant application is underway to target the core Kurmanji milestones—finishing the translation through a structured retention tool, refining our existing voice model, and keeping pace with NVDA’s release cadence.

This Collective funds the operational baseline around that work. At this early stage, small recurring contributions have an outsized impact: a few backers each month make the difference between paying for model-training GPU compute out-of-pocket or keeping the pipeline running continuously.

If this work matters to you or an under-served community you know, consider backing us or sharing this page.