Thank you for your valuable feedback during the Early Access period. This update is focused entirely on improving performance, speed, and overall user experience.
Faster Live Captions
Speech recognition now uses GPU acceleration more consistently. On some systems, the workload was silently falling back to the CPU, causing delays and interruptions during live translation.
Live caption performance has been improved by up to approximately 4.5x on many systems.
Change Language and Model Directly from the Overlay
Source language, target language, and AI model selection can now be changed directly from the subtitle overlay toolbar.
You can instantly adjust these settings without opening the main application window.
Lower VRAM Usage and Smart Model Loading
Live caption mode now starts with a lighter configuration, providing faster startup times and reduced memory usage.
The image and screen translation model is now loaded automatically only when it is used for the first time.
Settings Saving Improvements
The "AI Memory" (context size) setting is now properly saved and no longer resets after restarting the application.
Additional Improvements
This update also includes general stability improvements and various reliability fixes.
Thank you for continuing to share your feedback and helping us improve Live Translate during Early Access.