v2.14.0: the reel preview shows the facecam and the frame; ChatGPT's match-video rules

The timeline preview draws the facecam where the render puts it and frames the game in the area left; match-video stretches start at most 8 s (15 for a stand or clutch, 20 with talk) before the player's first play; no CAM under 3 s; a 2-6 s cold open.