Improved

• Multi-line dialogue — longer dialogue that wraps across several lines now stays together more reliably before translation, instead of breaking into separate pieces. This continues the line-grouping improvements from recent updates.
• Overlay in recordings and captures — the translation overlay now shows up in screen recordings and captures, so it stays visible when you record.
• Reading still scenes — on static screens, translations stay steadier and react faster, and in cursor-follow mode they re-appear reliably when you move back over text you read a moment ago.
• Word readings and pronunciation — word splitting, readings, and read-aloud pronunciation now share an improved Japanese analyzer across the app, so saved words and pronunciation come out more accurate.

Known issue

• Dense panels and stacked text — when several panels or labels are captured at once, text can still be grouped incorrectly. Drawing a tighter region around just the text you want gives the best result. We're continuing to improve this.