The timeline's scroll box is the flex row itself, so its reach is its
shots; a focused shot brings itself whole into view through
layout::bring_into_view, which table headers now share; Extend sits
with the caption under the strip. A voice's tags cap at the row's
width and wrap inside it.
AudioGenerationPlayer shows a generated sound's prompt, model and
length, a bar while it is made, then chat::AudioMessage to play it;
Regenerate and Download wait for it. TTSVoicePicker lists voices with a
sample to play, the chosen one washed with a check, tags dropping below
when narrow. VideoGenerationTimeline lays shots along their time under
a ruler, with posters, placeholders, a playhead and Extend.
AudioMessage's waveform is one Tab stop where Left and Right seek, and
it gives way at 280px. debug::ruler serves the timeline's ruler. A mask
shown to read no longer follows the pointer.
ZacharyZhang-NY