SYNC TEST CASE 01 — "Countries That Vanished From The Map"
Total duration: 45.0s | 5 clips x 9.0s (10s generator ceiling)
==========================================================

SEGMENT 1 — CLIP 001 — 0:00.0 to 0:09.0
Mode: ON-CAMERA DIALOGUE (lip-sync required, audio native to clip)
Speech window: 0:01.0 - 0:07.4 (6.4s)
Text:
"In 1990, this country had a flag, an army, and a seat at the
United Nations."
Silence: 0:00.0-0:01.0 and 0:07.4-0:09.0 (room tone only, no VO)

----------------------------------------------------------

SEGMENT 2 — CLIP 002 — 0:09.0 to 0:18.0
Mode: DOCUMENTARY VOICE-OVER (ElevenLabs, no speaker on screen)
Speech window: 0:09.0 - 0:17.8 (8.8s)
Text:
"Twelve months later it was gone. The Soviet Union did not collapse
in a single afternoon - it came apart republic by republic."

----------------------------------------------------------

SEGMENT 3 — CLIP 003 — 0:18.0 to 0:27.0
Mode: DOCUMENTARY VOICE-OVER (ElevenLabs, no speaker on screen)
Speech window: 0:18.0 - 0:26.0 (8.0s)
Text:
"Through the winter of 1991, the red flag over the Kremlin was
lowered for the last time on December twenty-fifth."

----------------------------------------------------------

SEGMENT 4A — CLIP 004 (first 5s) — 0:27.0 to 0:32.0
Mode: ON-CAMERA DIALOGUE (lip-sync required, audio native to clip)
Speech window: 0:27.5 - 0:31.5 (4.0s)
Text:
"Fifteen new countries appeared overnight, and every map was wrong."

SEGMENT 4B — CLIP 004 (last 4s) — 0:32.0 to 0:36.0
Mode: DOCUMENTARY VOICE-OVER (cut to B-roll inside the same file)
Speech window: 0:32.2 - 0:35.4 (3.2s)
Text:
"Not slightly wrong. Obsolete, from the first day."

----------------------------------------------------------

SEGMENT 5 — CLIP 005 — 0:36.0 to 0:45.0
Mode: DOCUMENTARY VOICE-OVER (ElevenLabs, no speaker on screen)
Speech window: 0:36.2 - 0:44.6 (8.4s)
Text:
"Cartographers spent a decade redrawing borders that had held for
seventy years. Some of those lines are still argued over today."

==========================================================
WHAT THIS TESTS
1. Lip-sync clips (1, 4A) whose native audio must NOT be ducked,
   replaced, or trimmed to VO length.
2. Pure VO clips (2, 3, 5) where narration drives the cut point.
3. A mid-clip handoff (4A -> 4B) inside a single 9s file, with only
   a 0.2s gap. A renderer keying off file boundaries will start the
   4B voice-over at 0:27 and talk over the on-camera line.
4. Cumulative drift across five cuts instead of three - four
   opportunities for per-segment error to compound.
5. Segment 1 and 4A speech windows are shorter than their clips.
   Trimming to VO length will cut dialogue mid-sentence.
