v1.76 — Faster GPU OCR & Text-to-Speech

ゲーム画面オーバーレイ翻訳 · published · build 26393993

Faster OCR on GPU: full-screen recognition is now about 9-16% faster (output slimming on the DirectML path; identical recognition results, verified bit-exact on 3,000+ screenshots).

New: Text-to-speech — the app can read finished translations aloud using Windows voices (off by default; turn it on in Settings. A voice matching your target language is selected automatically when installed).