Screen translation beside your microphone
The screen translation button now sits with the microphone and desktop audio buttons in the main window, and works like them: off is red and struck through. The first click opens its setup, where you choose the area to read. After that, a click turns it on or off, and holding the button opens its settings.
New installs start by asking What do you want to translate?: voice chat, on-screen text, or both. You can change it any time from the same buttons.
Screen translation also puts less load on your computer while the text on screen stays the same.
Turn inputs on and off while translating
The microphone, desktop audio and screen translation can each be turned on or off while translation is running. An input you turn on joins the running translation and shows Preparing... until it is ready. Turning off the last one stops translation.
A taller overlay
Settings > Overlay > Overlay Height has three sizes: Standard, Half and Full screen. A taller overlay keeps more lines on screen.
The translation limit stays on screen
When your daily cloud translation allowance runs out, a notice now stays on screen until the allowance refills, instead of a popup that disappeared after a few seconds. Subtitles keep showing what was said, in its original language, and screen translation pauses until the refill.
On the Free plan, speech recognition then runs on your computer only, without cloud assist, until the allowance refills.
Starting while speech recognition gets ready
If speech recognition for one of your languages is not ready yet, because cloud assist is still connecting or the local speech model is still downloading, Start translation now waits for it instead of failing, and Unitra AI tells you which languages are waiting.
The local speech model now always downloads in the background on a new install. The switch for it in Settings > Privacy is gone.
Voice output joins cloud assist
Voice output now uses cloud assist too, like speech recognition already does. Part of its work runs in the cloud, which frees more of your graphics card for your game and gets your translations spoken sooner. Your voice profile stays on your computer, and if cloud assist does not answer, that sentence is spoken on your computer as before.
Fixes
- Game audio subtitles that caught only fragments while voice output was on.
- Lines left untranslated when a sentence was taken for the wrong language, for example English spoken in a conversation set up for Chinese and English.
- A stray fragment after a finished line, made of its last words and a closing quotation mark, in Chinese and Japanese.
- Every sound on your computer being translated after a game went quiet for a while, for example while you were alt-tabbed, instead of just the game.
- Short stutters that kept coming back on computers with little free graphics memory while speech recognition ran on cloud assist.
- Long English sentences that voice output took several seconds longer to speak.
- Console windows that flashed up during installation.
Plus a number of smaller fixes to audio capture, cloud assist, voice output and installation.
If something does not work as expected after updating, let us know in Discord.