# Critic round 4
Copy for the case-study exporter of the final check, `critic-09-r2.md` section "Final check"; the original stays the source.

Verdict: FIX row 1 out to source 1703.70 so "باليستي" is whole (P1), then SHIP

## Final check
Render checked: `ice/renders/final/mohawalat-fahm-france_reel_9x16_master.mp4`, sha256 89a6d15a...7dddf1d (checked), 66.36 s, 1659 frames; changes as in D-37 and D-38. Ranges checked: 0 to 11.8, 55.6 to 58.1, 61 to 66.36, plus every cut for doubled first frames. Against the fix-check render, frames changed only in 0.04 to 2.72, 4.32 to 11.80, 26.68 to 32.24 (start frames rounded) and 55.88 to 58.00; the dialogue stem changed only in 3.8 to 10.5. Evidence in `scratch/critic-r2/final/`.

| Item | Status | Evidence |
|---|---|---|
| Disc matte, frigate (0.44 to 2.72) | FIXED | Both heads sit in front of the disc; the disc shows only on the sky; hair, ears and faces of both men carry no red (crops at 1.20, 2.00, 2.48). |
| Disc matte, army budget (55.74 to 58.00) | FIXED | Macron's whole head is in front of the disc and the disc is whole in front of the soldiers behind him. The fix-check render had those soldiers cut out in front of the disc (visible at 56.80 to 57.60); I missed it in the fix check. |
| Row 1 out at source 1703.62 (seam now 3.88) | NOT FIXED: P1 regression | Whisper ends the hook's key noun as "باليست" (master, whole hook and 2.6 to 5.3 window), "باليستا" (master, 1.9 to 4.4) and "باليس" (dialogue stem, three windows): 0 of 6 passes hear "باليستي", against 6 of 6 on the round 2 render. The pitch track, mine and the one in `transcripts/row1check/row1-out.json`, is unvoiced from 1703.56 to 1703.68 and voiced from 1703.72 at 158 to 163 Hz (the guest's range; host median 115 to 122). So 1703.64 to 1703.70 is the host's /t/ release, and the guest's "نعم" starts at 1703.72. Cut-point test on the source audio: out at 1703.62 gives "باليست" (2 of 2), at 1703.66 mixed, at 1703.70 "باليستي" (2 of 2, p 0.87 to 0.91). No click at the seam: the largest sample step there (0.078) is the speech itself (the word before it has steps up to 0.237). |
| Hook moved 0.16 s earlier | PASS | Every hook word still shows after its sound ("فيه" on screen 4.73, spoken 4.58 to 4.61; "الأول" 8.26, spoken 8.14). The riser and the music are 10 to 17 dB under "الأول". Host lip sync is 0 frames (motion-signature match at 5.6, r 0.97). |
| Doubled first frames | FIXED | No first frame repeats at any of the 26 cuts. The only held frames left are the breath (61.08 to 61.44, intended) and the missile clip's own repeats. The hFNj0osPUqE clip shows 2 frames of another scene at 10.40 to 10.44 (a cut inside the source clip, also in the earlier renders): a strobe within the montage's grammar, accept. Lip sync on the rounded shots: 0 frames at 28.6. |
| Music after 61.1 and peak | FIXED | Music is -30 to -35 dB under the sign-off and the title hold, as in the round 2 render; the master over the title hold is -24 to -45 dB. -14.1 LUFS, LRA 1.8 LU, true peak -1.8 dBTP (D-27: -1.5). |

Smallest fix for the open item: row 1 out at source 1703.70 with a 10 ms fade (it keeps the /t/ release and stops before the guest's voicing at 1703.72), and take the 0.08 s from the montage pause (2.04 to 1.96 s), so everything from 10.56 on stays where it is.

Round verdict: FIX row 1 out to source 1703.70 so "باليستي" is whole (P1), then SHIP