# EP01 Presenter Cut Plan (R3) + Card Drop Plan (R4)
**Authored:** SA-COMP-EP01R2, 2026-09-02. Spec source: `C:/DandyDon/smm/receipts/QC_S1_repub_socialism_EP01_2026-09-02.md` remediation table R3 (presenter cap) and R4 (b-roll cards).
**Timeline basis:** `EP01_build_spec.json` (145 shots, total 1181.77s — matches master) + `visuals/EP01/*_presenter.mp4` asset inventory (56 files, 437.99s — matches the QC receipt's build-side 438.1s) + `qc_ep01_ocr.jsonl` frames.
**Applies to:** the trimmed master `EP01_master_qc.mp4` (959.90s). All timings below are on the ORIGINAL master grid (pre-trim); the trim removes everything ≥949.88s except a 10s end card (959.88s master tail → 949.88+10s).

---

## 1. The presenter-cap problem, measured (not estimated)

Presenter material in the build = 56 rendered clips (`*_presenter.mp4`, sum 437.99s), mapped 1:1 to
56 spec shots (beat files are 1:1 with spec shots: `.tmp_beat_126.mp4` = 13.633s = spec S1-EP01-126's 13.62s).

| Bucket | Shots | Duration | Where |
|---|---|---|---|
| Body presenter shots (tc < 949.88) | 39 | **286.58s** | Inside narration — THIS is the R3 defect |
| Tail presenter shots 127–143 (tc ≥ 949.88) | 17 | 225.54s | Already removed by the R1 trim (no action) |

After R1, presenter exposure on the 959.90s master = **286.58s = 29.87%** (the QC receipt's 46.1%
figure was exposure-vs-narration-including-the-uncut-tail; on the trimmed master the exposure is
286.58/959.9). Cap = 18% = 172.78s → **must remove ≥ 113.80s of presenter-visible material** from the body.

Note the render composition (from OCR evidence): presenter appears inside composite cards
(`framing: split/pip_left` shots like 005, 014, 023, 078) and as full-frame cards (`framing: full`).
Full-frame shots are total presenter exposure; split shots are partial. The plan below cuts
full-frame cards first — this is both the largest, safest reduction AND the composition fix the
QC demanded (presenter beside docs, never full-frame takeover).

## 2. Cut list — Tier 1: full-frame presenter cards (remove ≥114s)

Cut these 12 body shots entirely (replace each with the adjacent receipt/ledger/b-roll card of the
same segment — the beat's narration continues under the replacement; re-time to the SRT cues):

| Shot | tc_in | dur | card | Replacement guidance | SRT cue anchor |
|---|---|---|---|---|---|
| S1-EP01-006 | 43.08 | 11.58 | act | adjacent receipt card (007/008 beat) | cue ~14 |
| S1-EP01-066 | 487.63 | 11.00 | act | 067 receipt | cue ~152 |
| S1-EP01-090 | 656.59 | 8.69 | act | 092/093 receipt wall | cue ~202 |
| S1-EP01-031 | 238.56 | 8.10 | act | 029/030 ledger | cue ~72 |
| S1-EP01-009 | 75.73 | 7.18 | act | 010/011 chart | cue ~25 |
| S1-EP01-012 | 97.27 | 7.18 | act | 013→ replaced too (see Tier 2) | cue ~31 |
| S1-EP01-013 | 104.45 | 7.18 | act | 014 split chart | cue ~34 |
| S1-EP01-024 | 188.88 | 7.18 | act | 022/023 chart | cue ~61 |
| S1-EP01-082 | 594.98 | 7.18 | act | 084 broll/085 receipt | cue ~183 |
| S1-EP01-083 | 602.16 | 7.18 | act | 084/085 | cue ~186 |
| S1-EP01-091 | 665.28 | 7.18 | act | 090→(also cut)/092 wall | cue ~204 |
| S1-EP01-075 | 543.91 | 6.95 | act | 074 act→cut / 077 ledger | cue ~170 |

Tier-1 sum = **96.58s**. Exposure after Tier 1 = 286.58 − 96.58 = 190.00s = **19.8%** — still over cap.

## 3. Cut list — Tier 2: split/pip composites (convert, don't cut)

The remaining 190.00 − 172.78 = **17.2s** overage comes out of the composite (split/pip) shots,
where the presenter is a PANEL, not the frame. Converting one `split` shot to a doc-only framing
removes only its presenter-panel share (~40–50% of shot area). Options, in order:

1. **S1-EP01-005** (35.90, 7.18, chart/split) — chart carries the beat alone; drop the presenter panel.
2. **S1-EP01-023** (175.56, 13.32, chart/split) — longest split shot in the body; doc-only.
3. **S1-EP01-014** (111.63, 9.85, flow/split) — doc-only.
4. **S1-EP01-078** (558.04, 7.18, ledger/pip_left) — ledger alone.
5. Keep 012/064/090/065 etc. only if still needed after 1–4 land.

Cutting the presenter panel from 005+023 (20.50s of shot time ≈ 8–10s of presenter area,
conservatively) brings exposure to ~180–182s ≈ **18.8–19.0%**. If still >18%, ALSO convert 014 and
078 → exposure ≈ 172–176s ≈ **17.9–18.3%** — at the cap line; the remaining 1–3s of overage is
removed in the Tier-3 trim below.

## 4. Cut list — Tier 3: trim the longest remaining exposure, don't cut narration

- S1-EP01-136/135/134/133/… are TAIL shots (tc ≥ 949.88): **already removed by R1. No work.**
- Body: shorten S1-EP01-014's flow hold by 1–2s (SRT slack cue 34→35) and S1-EP01-023's chart hold
  by 1.5s (SRT slack after cue 57) — frame-accurate trims at the next cut boundary.

## 5. Verification protocol for the rebuild (SA-COMP re-render)

After the cut rebuild, re-measure presenter share on the RENDER (not the build inventory):
1. Extract 1 fps across the full master; run the presenter detector used by the build (`*_presenter.mp4` asset match by frame hash at shot tc_in boundaries).
2. PASS = sum of presenter-visible seconds ÷ 959.9 ≤ 0.18 AND no presenter full-frame shot >4s.
3. Receipt the number in the rebuild's QC addendum.

## 6. R4 — card drops (factbase-cited replacements)

The two findings (QC frames f_0085 / f_0086, both inside **S1-EP01-055**, 422.32–429.50, spec
`card: broll`, `receipt_ref: 23`, provenance E&E News `inside-trumps-foray-into-mineral-ownership`):

| Frame | t | Rendered content (OCR) | Defect | Fix |
|---|---|---|---|---|
| f_0085 | 422.5 | E&E News archive page (bright screenshot, lum 145) | off-thesis interstitial, not the cited doc | **DROP.** Rebuild 055_broll from the correct local capture: `visuals/EP01/vst/023_www_eenews_net_articles_inside_trumps_foray_into_min.png` (the manifest's OWN provenance path — it was never wired into the render). Ken-Burns it for 7.18s per shot spec. |
| f_0086 | 427.5 | Headline "US, Australia advance mineral projects under $3.5B agreement" | **number NOT in factbase** (no $3.5B entry) + wrong-subject for the equity-stakes beat | **DROP.** Replace with a factbase-backed equity-stakes card. |

**Replacement content (verified factbase lines, local file line numbers):**
- `factbase.md:23` (S1_repub_socialism) — "2025-2026 US equity stakes in critical-minerals companies: DOD took 10% equity stake in Trilogy Metals (announced Oct 6 2025, $35.6M + warrants for 7.5%); DOE took 5% in Lithium Americas (Oct 2025, renegotiated $2.3B loan); DOD largest shareholder in MP Materials ($400M preferred, 15% if warrants exercised) — E&E News by POLITICO, Oct 8 2025 — VERIFIED." Card shows the three stakes + the E&E citation. This is EXACTLY what S1-EP01-055/056's spec headline already says (`"2025-2026 US equity stakes in critical-minerals companies: DOD took 10%"`, `receipt_ref: 23`) — the spec was right; the b-roll render wired the wrong screenshot.
- `factbase.md:24` — CNBC full-list card (MP 15%/$400M, Lithium Americas 5%+warrants, Trilogy 10%/$35.6M, Intel 10%/$8.9B CHIPS, Vulcan $50M-in-$750M, ReElement $80M, Korea Zinc 10%, xLight $150M CHIPS, Atlantic Alumina $150M) — use as 056_receipt's content if the receipt card needs the full list.
- **Narration needs no edit:** SRT cues 84–88 (7:02–7:09) speak equity stakes/customer-of-first-resort — the replacement cards ARE that beat's subject. No VO re-record required.

**Downstream guard for the rebuild:** after swapping 055's visual, run the OCR scan pass
(`C:\tmp\qc_ep01_ocr_scan.py` pattern) on the rebuilt section and grep frames 422–430 for
`Australia|\$3\.5B` — must be zero hits.

## 7. What this plan does NOT do (per scope)

- Does not touch the R6 GAO-itemized-framing card (factbase:14 mandatory line) — separate SA-RES→SA-COMP item.
- Does not modify narration/SRT — the presenter cuts ride under existing cues with cut-point re-timing only.
- Presenter discipline verification on the RENDER (≤18% measured frame-side) is owed by the re-run of SA-QC-VIS (R9) together with the vision-capable review.