Quick Translate Engine · published · build 23612933
The entire interface has been migrated from Tkinter to modern PySide6.
Win11-style dark theme (Graphite • Blue): rounded cards, shadows, modern combobox/buttons/scrollbars.
8 setting pages rewritten: Quick Translation Modes, OCR, Full Screen, Subtitle, Audio, Offline, API Settings, Logs.
Logo support added to the top-left corner.
New Themes page: Graphite, Nord, Dracula, Midnight, Carbon, Light.
Theme is applied instantly and saved (palette-based system).
User Manual page added inside the app.
Scope: API setup, Quick Translation, OCR, Instant Subtitles, Full Screen, Audio, Offline, Appearance/Themes, Logs + tips.
Latency reduced: idle scanning max 100ms, active scanning 30ms→20ms, fast mode duration 5s→8s (fixed the slow loading issue after long intervals).
Loss of meaning fixed: blur-cache threshold 0.85→0.96, similarity skip 0.85→0.93 (prevented incorrect/outdated translations from showing).
"Show previous sentence on top" feature added: previous translation is displayed in a separate, faded box above the main box. (Disabled by default; enable in Settings → Instant Subtitles.)
List → paragraph issue fixed: lines are merged only if they are a continuation of a real sentence (list items remain separate).
Single region → 2 bubbles / overlap fixed: now displayed in a single bubble, preserving the line structure.
Bubble flickering/overlapping in frequent scans resolved (overlay is not redrawn unless content changes).
OpenAI + OCR was completely broken, fixed: backward-compatible context parameter added to translate_text, None error in batch resolved.
OpenAI errors are now written to the Logs page (shows quota/model/key reasons).
Clients automatically restart and refresh language lists when the API key changes.
Overlay sub-process no longer loads unnecessary libraries.
OCR thread count limited (default core/2, max 4) — reduced CPU pressure during gaming.
Fixed pool (ThreadPoolExecutor) instead of spawning a new thread per translation + early stale-skip: eliminated wasted API calls and thread churn (retaining quality and speed).