Timing Constants

Defined at the top of video/hf/build.py:

PAD_LEAD  = 0.6   # silence before voice starts (seconds)
PAD_TRAIL = 0.9   # silence after voice ends (seconds)
OVERLAP   = 0.4   # crossfade between adjacent scenes (seconds)

How they affect the timeline

Scene N:   |← PAD_LEAD →|←── audio ──→|← PAD_TRAIL →|
           0            0.6           0.6+dur        0.6+dur+0.9

Scene N+1 starts at:
  scene_start[N] + scene_duration[N] - OVERLAP

The OVERLAP creates a 0.4 s window where the previous scene is still visible while the next scene fades in. Because scene blocks simply have opacity: 0 until their GSAP entrance fires, the overlap is handled naturally by the separate block timelines.

Audio timing

Within a scene block, audio starts at data-start="{PAD_LEAD}" (0.600 s).
In the host index.html, the absolute audio start is scene_start[i] + PAD_LEAD.

Adjusting timing

To change the silence padding, edit the constants and re-run build.py. The changed values will propagate to all scene blocks and the host automatically.

To change timing for a single scene only, you would need to override that scene's computation in main() — there is currently no per-scene override mechanism.

Scene duration summary

With the defaults above and the current narration durations:

Scene	Audio dur	Scene dur	Start
00 intro	14.45 s	15.95 s	0.00 s
01 llm	28.87 s	30.37 s	15.55 s
02 ai	31.24 s	32.74 s	45.52 s
03 mcpgw	27.43 s	28.93 s	77.86 s
04 mcpreg	29.86 s	31.36 s	106.39 s
05 skill	27.59 s	29.09 s	137.35 s
06 outro	15.05 s	16.55 s	166.04 s
Total			182.57 s