The real-time subtitle translation feature is ready!
In the previous user guide, we mentioned that role-playing games are recommended for learning English only when the user's vocabulary is roughly equivalent to the vocabulary used in the game's dialogues.
This is because if the game's dialogue contains too many new words, users would need to pause frequently to look up words, making it difficult to stay immersed in the game. However, this situation has now changed.
With real-time subtitle translation, users can focus on the main story of the game while selectively learning some of the vocabulary.
To support this feature, we have also upgraded the local LLM models used for translation. The models now run faster and translate more accurately.
We have also optimized our text detection algorithm, as a result the real-time text detection consumes about the same amount of CPU resources as before. Please note that text translation still requires a substantial amount of CPU resources, so we recommend that users choose the appropriate model option (medium or small) based on their computer's performance.
By default, the subtitle area is located at the bottom of the screen. While the game is paused, users can adjust the subtitle region. Setting accurate boundaries for the subtitle region can greatly improve recognition accuracy.
At present, this real-time subtitle translation feature is experimental, and it only supports white and cyan subtitles. It is compatible with games running in borderless windowed or windowed mode. If you encounter any issues during use, please feel free to give feedback in the Steam community.
In addition to real-time subtitle translation, the following new features have been added:
Instant Pause: When the real-time subtitle translation feature is enabled, pressing the pause key no longer requires waiting 1–2 seconds for text detection to complete. Instead, the system immediately reuses the results from the real-time text detection module.
Custom Area Text Recognition: While the game is paused, users can specify a rectangular area using Ctrl + left mouse button. Text recognition will then be performed within the selected region.