A small update that answers a question several of you asked in reviews and on the forum: "Can I run OCR on the CPU so the GPU stays free for the game, and how do I know which one is being used?"
NEW: The settings screen now shows which device is actually running text recognition. Go to Settings > Translation > Options: the new line "OCR device" reads "GPU (DirectML)" when the GPU is in use, "CPU (Low-VRAM mode)" when you have turned that mode on, or "CPU" if the GPU could not be initialized and the app fell back automatically. It shows the real, measured state, not just the setting.
IMPROVED: To be clear about the CPU option: it already exists. The switch "Low-VRAM mode (run text recognition on the CPU)" in Settings > Translation > Options moves OCR off the GPU entirely. We named it after its measured benefit: it frees almost all GPU memory, which helps on PCs that crash or stutter from running out of VRAM. In our own tests it did not make games run faster (CPU load rises and frame rate was slightly lower than the default), and translation takes longer (about 0.3s to 1.3s), so we recommend it for VRAM problems rather than as a general "game priority" mode. When it is on, the control bar now also shows "Text recognition running on CPU" once at startup so you always know.
FIXED: Traditional Chinese interface texts used two different words for the control bar; they now consistently use 操作列.
DLC: The free "High-Quality Chinese Pack" (a local LLM tier for Chinese) is still under Steam store review. We resubmitted it after updating the required AI-content disclosure, and we expect it to become available soon; we will post here as soon as it is live. The pack is already included in this build, so once the store page is approved you can add it to your library and use it immediately. Sorry for the wait.