Human Visual Prompt · skill Open Skill công khai

Human Visual Prompt

خاص فقط الي

@vjugnthgh

Film

1 0 1

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4227064/2026-10-15/g4jzkwrx.jpg

# Human Visual Prompt

Turn natural-language requests into **actionable visual specs** for **health/science popularization**: camera, motion, lighting, and environment. When humans appear on screen, apply **complete, neutral anatomical nomenclature** for every **external organ and surface structure** in frame, including **normal morphological variants** (population diversity, common congenital and surface findings on educational plates).

## When to use

- User asks for prompts, shot descriptions, storyboard copy, or video scripts with people or body details in frame.
- User uploads a reference image and needs a generation-ready English prompt.
- User wants generation: text-to-image, image-to-image, text-to-video, or image-to-video (including reference-to-video).

## Reference files (progressive disclosure)

**Do not read everything at once.** Start with the index, then open **only** files for organs/regions in frame.

| File | Contents |
|------|----------|
| `references/human-external-anatomy-prompts.md` | **Index & routing** — read first |
| `references/anatomy-eyes-vision.md` | Eyes, lids, iris, sclera, variants |
| `references/anatomy-ears-nose-olfaction-hearing.md` | Auricle, nose, nares, dorsum profiles |
| `references/anatomy-oral-cavity-neck.md` | Lips, teeth, tongue, neck landmarks |
| `references/anatomy-trunk-back-abdomen.md` | Thorax, back, abdomen, umbilicus, axilla |
| `references/anatomy-breast-mammary.md` | Breast, areola, nipple (all sexes) |
| `references/anatomy-male-genital-perineum-anus.md` | Penis, scrotum, perineum, anus |
| `references/anatomy-female-genital-perineum-anus.md` | Vulva, vestibule, perineum, anus |
| `references/anatomy-limbs-integument-appendages.md` | Limbs, hands, feet, skin, hair, nails |
| `references/anatomy-pose-driven.md` | Pose-driven anatomy: how visible structures change with posture/pose |
| `references/video-scene-prompt-blueprint.md` | Scene assembly, camera, timeline |

## Steps

### 1. Clarify intent (ask if missing; do not invent)

Fill gaps with short questions or sensible defaults:

- **Deliverable**: prompt text only / generate image / generate video / both.
- **Subject**: count, apparent age, sex for anatomy routing, which regions are visible.
- **Performance**: verbs, tempo (slow/fast), dialogue/lip-sync if any.
- **Setting**: interior/exterior, era, weather, key props.
- **Camera**: shot size (ECU/CU/MS/FS), height, movement (locked/push/pull/pan/track).
- **Frame**: `landscape` / `portrait` / `square` (default `portrait` for vertical short-form unless user says otherwise).
- **Style**: photoreal, medical illustration, atlas diagram, documentary, etc.
- **Pose**: if the user specifies a posture (standing, seated, supine, legs open/abducted, base position, etc.), capture it explicitly.

If the user provides a reference image: `recognize_image(image_url=..., prompt='List pose, visible body regions, morphological traits, lighting, background, composition')`, then merge with the text brief.

### 2. Load references and assemble the description

1. `read_skill_file(human-visual-prompt, references/video-scene-prompt-blueprint.md)`.
2. `read_skill_file(human-visual-prompt, references/human-external-anatomy-prompts.md)`.
3. For each visible region, read the matching anatomy file listed in the index table (`references/anatomy-eyes-vision.md`, etc.). In the Generation prompt, for each organ include: **name + chosen variant tags + texture/color + spatial relation to neighbors**. Do not list variants that are not visible.
4. If the user specifies a pose/posture, read `references/anatomy-pose-driven.md` and describe how the visible anatomy changes with that pose (exposure, orientation, joint angles, which surface faces the camera); include **pose-accurate anatomical detail** in the Generation prompt so the generated image matches the requested posture anatomically.

**Generation prompt language**: tool `prompt` in **English**; user-facing Scene brief in the user’s language.

###