## v1.3.004 — 2026-07-08
- feature: **"System Voice · Windows" — a new voice option that speaks with the voices built into Windows.** It runs as a separate line from the other local voices and lists the Windows voices installed on your PC, grouped by language. A refresh button re-scans for voices you add through Windows Settings, and if a selected voice cannot speak the current text's language the app reports it instead of playing silence. (Windows only.)
- feature: **GPU layer control for the local language model.** A per-model slider sets how many model layers run on the GPU, with an option to remove the VRAM budget limit so a model can be loaded even when it exceeds the estimated budget.
- feature: **An 8K context-length option for the local language model,** for lower memory use on smaller GPUs.
- feature: **A screenshot resolution setting for the vision feature,** to balance detail against speed.