F-captions: auto subtitles

— NEW FEATURES — • New option: Caption split. In Model settings, you can now set how often the next transcription cuts speech into separate captions. Default keeps each model's own segmentation. Set a lower value to cut more often, all the way to captions of a single word. Set a higher value to join neighboring captions into longer lines. A pause still starts a new caption, and lines that are already shorter stay shorter. • New caption animation: Karaoke slide. With this animation, the subtitle text stays visible while a background fill slides from word to word as each word is spoken. You can set the fill color, roundedness, size, and opacity. As always, the default look can be set in settings. — OTHER IMPROVEMENTS — • Visual cues added to the video editor, as well as encyclopedia-style Docs section. • Various UI improvements.