[Tested] How Far Can You Push the Horror? Escalating Fear with VHS Grain, a Crawling Pose, and Corpse-Pale Skin

Conclusions First

• The single most effective fear-boosting phrase is unnaturally pale bluish-white skin. In 3 of 3 images the skin turned blue-gray — a bloodless, corpse-like complexion. As a bonus, the off-shoulder look that had failed 0-for-3 last time finally appeared under this condition (1 of 3).
• Ghost-story underlighting is achieved through pose, not lighting keywords. With crawling on the floor toward the camera, the flashlight landed on the floor or in hand in 3 of 3 images, lighting the face from below as a result. What the lighting condition (h02) failed to do last time happened naturally via pose.
• We could not hide the face with hair. completely covering her face reproduced in 0 of 3 images — the face stays clearly visible in all three. Together with last time’s h02, this strengthens the observation that this model steers toward showing the subject’s face.
• grainy VHS footage, analog video noise adds grain and an analog color cast (3 of 3), but scanlines and tracking artifacts — the signature VHS glitches — appeared in 0 of 3 images.

Previously on This Experiment

In our 4-axis J-horror staging test, we decomposed horror into location, lighting, costume, and vocabulary, and the combined condition (h05: horror movie still + abandoned hospital + white kimono) came out scariest.

This time, h05 is our baseline, and we add four fear-escalation axes one at a time, then stack the winners. See the previous article for baseline images (same prompt, same seeds — not regenerated).

Experiment Setup

Base Prompt (Last Time’s h05 = This Time’s Baseline)

Conditions

Note: g01–g04 each change exactly one axis relative to baseline. g05 is a combined condition, so we make no claims about individual elements there.

Results by Condition

g01: VHS Grain

seed 42	seed 43	seed 44

Observations: All 3 images picked up coarse film grain, further desaturation, and an analog color cast — the “something caught on tape” texture. However, true VHS artifacts (scanlines, tracking noise, chroma bleed) appeared in 0 of 3. In seed 42, a horizontal band of light falls across the mouth with hair draping both sides of the face, dialing up the creep factor. In seed 44, a red decorative obi appeared on what should be a plain kimono.

g02: Crawling Pose

seed 42	seed 43	seed 44

Observations: All 3 images show her on all fours, leaning forward, eyes up at the camera. The lighting is the headline: in 3 of 3 images the flashlight sits on the floor or in her hand, lighting the face from below as a consequence. The classic ghost-story underlighting that direct keywords failed to produce last time emerges naturally once the pose drops to floor level. Seed 42 has heavy shadows around the eyes and an intense stare. Seed 44 shows dark smears running from the eyes, with debris scattered across the floor.

Going from “standing still” to “coming toward you” changes the species of fear — and the forward-leaning pose preserves more allure than any other condition this round.

g03: Face-Covering Hair

seed 42	seed 43	seed 44

Observations: completely covering her face reproduced in 0 of 3 — the face is clearly visible in every image. Strengthening partly to completely only increased hair volume around the face. Instead, dark circles and sunken eyes intensified in all 3, and seed 43 rendered reddish-black swollen skin around the eyes.

Combined with last time’s failed deep shadows on face, the pattern is consistent: this model resists instructions to hide the subject’s face. “Scary because you can’t see it” is hard to build with prompts alone.

g04: Pale Skin

seed 42	seed 43	seed 44

Observations: The biggest fear jump of any single change this round. All 3 images render blue-gray, bloodless skin. Seed 43’s face is fully blue-gray — it does not read as living. Seeds 43 and 44 add redness around the eyes, like they’ve been crying or are bloodshot.

An intriguing byproduct: the off-shoulder finally appeared in seed 44. white burial kimono slipping off one shoulder had failed across 6 conditions and 18 images since the last experiment; under this condition it worked in 1 of 3.

g05: Final Form (Everything Stacked)

We stacked the three effective elements — VHS grain, crawling pose, pale skin — onto the baseline.

seed 42	seed 43	seed 44

Observations: Pale blue skin × all-fours crawl × floor-level flashlight underlighting × grain all landed in 3 of 3 images — the scariest condition of the entire series. Seed 44 adds unprompted decay details: rust-brown stains on the kimono and a bandaged foot.

One unexpected shift: the kimono, nearly plain in the baseline and g04, switched to a red obi in 3 of 3 images here. Being a combined condition we can’t attribute it to any single element, but g01 (VHS) seed 44 also produced a red obi — a plain white burial kimono may become harder to hold as conditions stack up.

Cross-Condition Comparison (seed 42)

VHS grain	Crawling pose	Face-covering hair

Pale skin	Final form

For the baseline (last time’s h05, seed 42), see the previous article. Fear gain ranks texture (VHS) < composition (crawl) < skin color (pale), and stacking compounds them.

Summary: The Fear Escalation Recipe

These results are specific to z-image-turbo (CFG=1.0, 8 steps); other models and parameters may differ. For fundamentals, see our verified best practices.

Related Articles

[Tested] Can Horror and Allure Coexist? Breaking J-Horror Staging into 4 Axes — Location, Lighting, Costume, Vocabulary — Across 18 Images

[Verified] Image Generation Prompt Best Practices

[Experiment] Prompts for Goth and Dark Makeup | 8 Conditions, 24 Images