# Paper edit (stage 03): "الصاروخ الباليستي ولا الثانويات"

Source: `ice/sources/episode/episode.mp4` (25 fps). Clock: SOURCE seconds. Word times from faster-whisper large-v3-turbo on short windows (`ice/transcripts/`, `ice/transcripts/seams/`); every in and out point snapped to the quietest 20 ms frame of the speech envelope (`tools/env.py`). No cut falls inside a word. Pauses inside segments are natural (0.3 to 0.7 s, `tools/gaps.py`); none reaches the 1.0 s podcast trim threshold, so none is trimmed. No speech is sped up.

Speakers: HOST = عثمان آي فرح (presenter). GUEST = د. عبد الكريم أمنكاي, assistant professor of international relations, Doha Institute for Graduate Studies (episode description; identified on screen by the episode's own quote card at 00:28:00).

| # | Speaker | Source in | Source out | Length | Spoken words (verbatim, fillers in brackets) | Role | Cut reason |
|---|---|---|---|---|---|---|---|
| 1 | HOST | 1699.72 | 1703.72 | 4.00 | وفيه صورة لماكرون وهو [يعني] مع إطلاق صاروخ باليستي | hook, part 1: names the picture | in: dip at 1699.70 after the guest's "لا"; out (r6): 1703.72, a frame boundary, after the host's /t/ release (unvoiced 1703.64 to 1703.70, after the closure at 1703.60) and where the guest's voiced "نعم" begins (158 to 163 Hz; host median 122, guest 166); a cut test on the source hears "باليستي" whole at 1703.70 and 1703.72 and "باليست" at 1703.62 and 1703.68 (D-38, `ice/transcripts/row1check/row1-out.json`, `cut-test.json`) |
| 2 | HOST | 1704.40 | 1709.04 | 4.64 | فبالتالي فيه ناس تقول لك [يعني]: إنت قادر تصرف على الصاروخ الباليستي؟ طب اصرف على الثانويات الأول! | hook, part 2: the contrast | in: silence before "فبالتالي" (1704.06 to 1704.86 under -45 dB); out: gap at 1709.0 to 1709.14 |
| - | (music) | | | 1.92 | | proof montage in a pause: the high schools and the street | the references put archive in speech pauses |
| 3 | GUEST | 1116.94 | 1124.00 | 7.06 | يعني فرنسا إذا استخدمت كل ما تنتجه في سنة واحدة، ستعجز عن سد مديونية | the debt | in: gap 1116.92 to 1116.98; out: gap 1123.98 to 1124.04 before "يعني هو معيار" (dropped) |
| 4 | GUEST | 1129.22 | 1134.00 | 4.78 | الحكومة منذ أسابيع ترسل إشارات أنه الدولة ليس لديها أموال | the claim "no money" | in: silence 1129.08 to 1129.28; out: silence from 1133.92 |
| 5 | GUEST | 1149.74 | 1159.64 | 9.90 | وهذا هو أحد المحركات لخروج أولئك الشباب. الحكومة تقول لهم منذ أسابيع: ليس عندنا أموال نخصصها أكثر لقطاع التعليم | the effect on the young and on schools | in: dip 1149.72 to 1149.78; out: dip 1159.62 to 1159.64 before "في حين" (seam check: whisper on 1155.5 to 1159.26 hears only "لقطاع", on 1155.5 to 1159.64 hears "لقطاع التعليم"; the dip at 1159.22 is inside the word) |
| - | (music) | | | 0.36 | | turn beat between two moments of the conversation | the two passages are about 8.5 minutes apart |
| 6 | GUEST | 1667.16 | 1695.60 | 28.44 | ما أغضب قسمًا من الشارع الفرنسي هو أنه لا يقبل أصلًا بفكرة أن الدولة عاجزة عن تحصيل الموارد، بكل بساطة. بعض الهيئات أو بعض أجزاء من المجتمع الفرنسي تقول: هذه أطروحة كاذبة. لماذا؟ لأنه عندما بدأت حرب أوكرانيا، وأرادت الحكومة الفرنسية أن تزيد من نفقات الجيش، وجدت موارد. | the turn and the payoff | in: silence 1666.72 to 1667.20; out: end of the held "موارد" (whisper on 1693.0 to 1695.56 hears the whole word; 1693.0 to 1694.84 does not), with a 60 ms tail fade |
| - | (music) | | | 0.40 | | payoff breath: the title cools, the music lifts | critic 09 r1 item 10: the payoff had 0.10 s before the sign-off |
| 7 | HOST | 325.28 | 328.76 | 3.48 | نحاول أن نفهم ذلك في هذه الحلقة من محاولة فهم | sign-off: the programme names itself | in (r3): the pause 324.84 to 325.36 after the whole of "المحتملة" (its tail runs 324.54 to 324.82; whisper on 324.30 to 326.2 hears it to 324.86, on 324.86 to 326.2 and 325.30 to 326.2 it hears only "نحاول أن نفهم"); out: dip 328.72 to 328.80 after "فهم" (seam check: whisper on 327.0 to 328.70 hears the whole "من محاولة فهم"); "أنا" starts at 328.82 |
| - | (music) | | | 1.40 | | title hold | the references hold the title after the name is said |

Speech total 62.32 s; with the four music gaps the reel runs 66.36 s (inside D-04, 55 to 70 s). Revision r6 (2026-10-29, critic final check): row 1 out 1703.72, after the /t/ release, the montage pause 1.92 s (every time after 10.56 s unchanged). Revision r5 (approver 06 r1): row 1 out 1703.62 at the closure, which cut the /t/ release. Revision r4 (2026-10-29, critic 09 r1): row 1 out 1703.76, the montage pause 1.88 s, row 5 out 1159.64 (in the dip, not on the tail of "التعليم"), the turn beat 0.36 s, a 0.40 s payoff breath; the loaded word written as said, "مديونية" (D-31). Revision r3 (2026-10-29): row 1 out 1703.64 (the last vowel of "باليستي"), row 7 in 325.28, after the round 2 approver. Revision r2 (2026-10-29): rows 2, 5 and 7 corrected after the stage 03 approver's return (`ice/review/stages/03-approver.md`) and a seam check by re-transcription (`ice/transcripts/seamcheck/`).

## Retakes and silences decided on paper
- Dropped between 3 and 4 (1124.00 to 1129.22): "يعني هو معيار، إنه معيار ضخم، بكل بساطة": a self-correction ("معيار... إنه معيار") that adds no fact.
- Dropped inside 4 to 5 (1134.00 to 1149.74): "في نفس السياق أنت ترسل إشارات إلى القطاعات ... لهذه القطاعات": 15.7 s that the kept sentence "ليس عندنا أموال نخصصها أكثر لقطاع التعليم" says in 4 s, in the guest's own later words.
- Dropped between 1 and 2 (1703.72 to 1704.40): the guest's "نعم" (1703.72 to 1704.06, identified by pitch in `ice/transcripts/row1check/row1-out.json`) and the breath after it.
- Fillers "يعني" inside sentences 1 and 2 stay in the audio (no cut inside a phrase) and are left off the screen.
- Kept: "بعض الهيئات أو بعض أجزاء من المجتمع الفرنسي تقول: هذه أطروحة كاذبة. لماذا؟" The pause after "كاذبة" and before "لماذا" is too short (no gap of 60 ms under the speech floor) for a clean cut, and the line is the turn of the argument.

- Word truth inside the kept rows (r3): whisper drops fillers and repeats. The retimed transcript now merges the YouTube auto-caption words that whisper dropped, aligned as sequences (`tools/build_edl.py`, marked `asr: youtube-asr`), and takes the caption onset where whisper stretched a word back over a dropped filler. Kept in the audio, never on screen: "يعني" at source 1117.00, 1152.76, 1675.52, 1678.52, 1680.60, 1684.36; the repeated "بعض" at 1678.80 ("بعض، يعني، بعض الهيئات"); the repeated "وجدت" at 1694.44 ("وجدت ... وجدت موارد", held vowel of "موارد" to 1695.54). None has a clean gap of 60 ms on both sides, so cutting any of them would cut inside a phrase.
- Repetitions inside row 6, decided on paper and kept: "بعض الهيئات أو بعض أجزاء" (a self-correction that is one phrase; no gap of 60 ms under the floor between the two "بعض"), and "وجدت ... موارد" with the held vowel of "موارد" (the auto-captions write "وجدت وجدت"; the word-level whisper pass hears one "وجدت" at 1693.06 to 1693.86, then "موارد" to 1695.54). The fillers "يعني" in rows 1, 2 and 6 sit inside phrases with no clean gap; cutting them would cut inside a phrase, so they stay in the audio and stay off the screen.

## First three seconds, cut from the real footage (short-form-hooks.md section 2)
Each opening of `hooks.md` was cut as a 3 s test from the episode and compared (`ice/plan/hook-tests/`, made with `tools/hook_tests.py`); the comparison is in `hooks.md`.

## Meaning check
- The host voices critics ("فيه ناس تقول لك"); the screen keeps that frame and never presents the line as his own view (COMPLIANCE RC4).
- Sentence 5's "وهذا" refers to sentence 4's "no money" signals, as in the source, where it follows the same idea one sentence later.
- Sentence 7's "ذلك" refers in the source to how the movement began and spread; in the reel it refers to the argument just made. Both are what the episode tries to understand; the line is used only as the programme's sign-off.