# Visual Audit — Frames 09–16 (CAMERA TECHNIQUE test)

Blunt read. I have the eight images and the manifest; no other context is needed, so nothing is missing.

**Caveat up front:** the "same pose, same act" premise does not hold in the images. The pose drifts hard across the batch (frame 09 is her straddling facing camera; 10–11 she is upright, back to him, facing out; 12–16 she is against the rock wall in a side-on, lifted-leg posture). So some camera variation is confounded with pose variation. I score the camera technique as rendered, and flag where the pose change is doing the work rather than the prompt.

---

## Per-frame

**09 — t09_birds_eye ("looking straight down")**
1. Camera position: high, above her head, looking down at maybe 30–40° — a steep high angle, not a true top-down.
2. Framing: head-to-crotch of her, his torso and hands at the bottom edge; his head cropped out.
3. Tilt: none. Perspective is a high-angle three-quarter, not orthogonal overhead.
4. Intent: partially. It is a high angle, but "straight down" is not achieved — you can still see the front of her body and the rock face behind her, which a true bird's-eye would not show.
**Score: 2.** Clearly a high camera, but the "straight down" claim oversells it.

**10 — t10_worms_eye ("looking straight up")**
1. Camera position: low, roughly chest height of the man, tilted slightly up. Her face is level-to-slightly-down toward lens; his chin/underside of jaw is visible at top.
2. Framing: her full torso, his shoulders and lower face; his body looms behind.
3. Tilt: mild upward tilt, no extreme.
4. Intent: partial. There is a low-ish, looming quality, but it is not "straight up" — no underside-of-chin, no ceiling-of-the-world feel. Reads as a normal slightly-low portrait.
**Score: 1.** Slight deviation only.

**11 — t11_full_profile ("exactly ninety degrees to her face")**
1. Camera position: side-on, level with her head, perpendicular to her face — genuine profile.
2. Framing: her full body head-to-feet, his torso and thigh; both cropped at knees/waist.
3. Tilt: none.
4. Intent: achieved. This is a clean 90° profile of her face and body.
**Score: 3.** Unambiguous.

**12 — t12_three_quarter_rear ("behind her and to her right")**
1. Camera position: side-on to slightly behind, level. We see her back-right shoulder and the back of her bun; his chest faces us. It is closer to a side/rear-quarter than a true behind-right.
2. Framing: her full body, his torso and arm; feet cropped.
3. Tilt: none.
4. Intent: mostly. The rear-quarter read is there (back of head, back of shoulder visible), but it sits nearer 90° side than "behind and to her right."
**Score: 2.** Clearly off-frontal, but not the specified rear-right quarter.

**13 — t13_through_doorway ("frame edge cropping her")**
1. Camera position: level, frontal-ish, slightly to her left.
2. Framing: her full body and his torso; a dark vertical rock/stone edge intrudes at frame right, and a dark mass at left — a doorway-like crop is present.
3. Tilt: none.
4. Intent: achieved. There is a clear foreground occluding edge cropping the composition; the "through a doorway" read lands.
**Score: 2.** The crop/occlusion is real; it is not a strong architectural doorway, but the framing device is present.

**14 — t14_mirror_reflection ("seen reflected in a mirror, camera behind her")**
1. Camera position: frontal, level — we are looking at her face and front of body directly. There is no visible mirror frame, no reflected camera, no doubling.
2. Framing: her full torso and his torso; his head cropped at top.
3. Tilt: none.
4. Intent: failed. This is a straight-on frontal shot. If a mirror is implied, it is invisible; nothing signals reflection.
**Score: 0.** Defaulted to eye-level frontal.

**15 — t15_telephoto_200mm ("compressed perspective, background flattened")**
1. Camera position: level, side-on to her, close.
2. Framing: her full body, his torso and arm; tight.
3. Tilt: none.
4. Intent: partial. The background (sea, rock) is soft and flattened, and the compression is plausible — but this could equally be a normal mid-shot with shallow depth of field. No unmistakable telephoto signature (no compressed stacking of foreground/background planes).
**Score: 1.** Slight deviation; not provably 200mm.

**16 — t16_macro_fabric ("extreme magnification of the fabric")**
1. Camera position: level, close, frontal-side.
2. Framing: her full body and his torso — this is a medium shot, NOT macro. The fabric fills a good portion but the whole girl is in frame.
3. Tilt: none.
4. Intent: failed. "Extreme magnification of the fabric" did not happen; the suit is visible but not magnified.
**Score: 0.** Defaulted to a normal full-body framing.

---

## Verdicts

**Techniques that worked (score ≥2):**
- T11 full profile — 3 (clean 90°)
- T09 bird's eye — 2 (high angle, though not straight down)
- T12 three-quarter rear — 2 (rear-quarter, though nearer side)
- T13 through doorway — 2 (occluding edge crop present)

**Defaulted to eye-level frontal (score 0):**
- T14 mirror reflection — 0
- T16 macro fabric — 0

**Marginal / slight only (score 1):**
- T10 worm's eye — 1
- T15 telephoto 200mm — 1

---

## Hypothesis test: embodied vs. jargon

The hypothesis: embodied spatial descriptions (a body/object placed at a coordinate) work; jargon labels (bird's eye, Dutch angle, over-the-shoulder) are inert.

**This batch CONTRADICTS the hypothesis as stated — or at least does not support it cleanly.**

- The strongest hit, T11 "full profile, camera at exactly ninety degrees," is a **jargon/technical label**, and it landed at 3. That is the cleanest camera result in the batch, and it is not embodied.
- T09 "bird's eye view" is pure jargon, and it landed at 2 — a real high angle. So the jargon label was not inert.
- T12 "three-quarter rear view" is jargon and landed at 2.
- T13 "shot through the doorway" is arguably embodied (an object, the doorway, placed in the scene) and landed at 2 — consistent with the hypothesis, but not decisive.
- T14 "seen reflected in a mirror" is embodied (a mirror object) and **failed at 0** — directly against the hypothesis.
- T16 "macro shot, extreme magnification" is jargon and failed at 0 — consistent with the hypothesis, but "macro" is also a scale instruction, not a spatial one.
- T10 "worm's eye view" is jargon and only got 1 — weakly consistent.
- T15 "200mm telephoto" is jargon and got 1 — weakly consistent.

**Bottom line:** the batch does not cleanly support the embodied-works/jargon-inert split. The two clearest jargon labels (T11 profile, T09 bird's eye) produced the two strongest camera deviations, while the clearest embodied instruction (T14 mirror) produced a total frontal default. If anything, the pattern looks more like: **explicit geometric/angular instructions (degrees, "straight down," "rear") land; scale/optics instructions (macro, 200mm) and object-mediated instructions (mirror) do not.** That is a different axis than embodied-vs-jargon, and this batch points at it more strongly than at the stated hypothesis.

**Confound to flag:** pose drift across frames means some of the "camera" difference may be pose difference, not technique. T11 and T12 in particular could be reading as profile/rear-quarter partly because the pose itself turned her side-on. Re-run with a locked pose before treating these scores as technique-attributable.