Camera terms work when they make one visible decision at a time. This protocol keeps the fictional adult identity, wardrobe, setting, style, and aspect ratio fixed while testing four framing changes.
What this four-shot test adds
Most camera-angle lists are vocabulary dumps. The useful question is whether a term produces a frame that can be evaluated and repeated. The matrix below turns camera language into four isolated prompt versions. It also separates shot size from viewpoint: “full-body” describes how much of the subject is visible, while “low angle” describes where the camera appears to be.
Keep the model, seed or variation method, base identity, wardrobe, location, lighting, and style unchanged. Use one subject first. A second subject creates extra occlusion and interaction variables that make a framing test harder to diagnose.
The controlled framing matrix
| Version | Only variable added | Visible pass condition | Common failure |
|---|---|---|---|
| A — close portrait | close portrait, eye-level camera, shoulders visible, centered face | Both eyes, jawline, hair boundary, and shoulders remain readable | The model also tries to include the full body, shrinking or distorting the face |
| B — three-quarter | three-quarter shot, eye level, framed from head to mid-thigh | Hands and wardrobe silhouette fit without an awkward crop | Hands fall exactly on the frame edge or the image becomes a close portrait |
| C — full body | full-body shot, camera at waist height, head and feet fully in frame | Headroom and floor space are visible; neither feet nor hair are clipped | The prompt says full body but a portrait aspect ratio is not selected |
| D — over shoulder | over-the-shoulder view, subject looking toward a mirror, face partly visible | The shoulder establishes viewpoint and the reflected face remains coherent | The model creates two unrelated faces or treats the reflection as another person |
Write camera instructions in the right order
1. Shot size
Choose one: close portrait, medium shot, three-quarter shot, or full-body shot. Shot size determines the information budget. A close portrait can prioritize expression and facial detail; a full-body frame must distribute attention across anatomy, wardrobe, pose, and floor contact. Adding portrait-level facial detail to a distant full-body frame can create competing priorities.
2. Camera height and tilt
Add eye level, high angle, or low angle only after shot size is stable. Extreme low angles can enlarge foreground limbs; extreme high angles can compress the torso and hide floor contact. The editorial pass condition is not “dramatic.” It is whether the perspective supports the intended composition without creating a new anatomy or age-appearance problem.
3. Subject placement
Use centered, left third, right third, or negative space on one side. Placement is especially important when the image will carry a headline or be cropped into a card. If the model ignores placement, simplify the environment and remove secondary props before repeating the instruction.
4. Lens language
Lens terms can communicate a visual intention, but they are not physical guarantees. “Wide-angle environmental portrait” asks for more context; “telephoto portrait with compressed background” asks for less apparent depth. Judge the visible perspective instead of assuming a numeric focal length was simulated accurately.
POV is not a substitute for composition
“POV” is ambiguous unless the prompt identifies whose viewpoint is represented, what is visible in the foreground, and where the fictional adult is located. For a safe benchmark, use a neutral action such as entering a studio, reviewing a wardrobe rack, or looking into a mirror. Avoid unexplained body fragments in the foreground; they often create duplicated hands or unclear ownership.
A better structure is: viewer position + subject position + action + frame boundary. Example: “viewer standing at the studio doorway, fictional adult subject beside a wardrobe rack, three-quarter body visible, hands unobstructed.”
Crop diagnosis: fix the cause, not the symptom
| Symptom | Likely conflict | Smallest useful repair |
|---|---|---|
| Feet cut off | Full-body wording conflicts with a tight crop or square output | Choose a taller ratio, add floor visible and feet fully in frame, remove close-up language |
| Face too small | The prompt asks for full body, a large environment, and portrait detail equally | Decide whether face detail or environmental scale is primary; create separate versions |
| Hands at edge | Three-quarter crop ends at the wrist | Change the boundary to mid-thigh and ask for both hands inside frame |
| Distorted foreground limb | Extreme perspective or unclear POV foreground | Return to eye level, remove foreground body parts, then reintroduce perspective |
| Two faces in a mirror | Reflection and person are interpreted as separate identities | State one fictional adult and one coherent reflection; simplify the background |
A six-point framing score
- Boundary: the requested head, hands, feet, and wardrobe edges are inside the frame.
- Viewpoint: camera height and direction can be inferred without guessing.
- Identity: one original fictional adult aged 25+ remains visually consistent.
- Silhouette: pose and clothing read at thumbnail size.
- Environment: background detail supports rather than competes with the subject.
- Repair cost: the frame does not require extensive inpainting or cropping to become usable.
Score each item 0, 1, or 2. Keep the prompt only when it improves the target framing score without reducing the safety or identity checks. Save the rejected version and the reason; otherwise the next revision tends to repeat the same conflict.
Primary sources and limitations
Product controls and terminology change. Check the current documentation for your model before copying a parameter literally. The sources below support the general workflow; they do not guarantee identical output across services.
Frequently asked questions
Should I put several camera angles in one prompt?
No. Use one shot size and one viewpoint per version so you can tell which instruction changed the frame.
Why does a full-body prompt still crop the feet?
The output ratio, tight portrait language, or competing detail requests may conflict with the full-body instruction. Use a taller ratio and make floor contact visible.
Is POV always a first-person image?
No. Models interpret POV inconsistently. State the viewer position, subject position, foreground elements, and frame boundary explicitly.
Reviewed July 2026 by the AI Porn XXX Editorial Team. Recheck product controls and documentation before repeating the protocol.