<h1 align="center">
<a href="https://prompts.chat">
Before writing any prompt, mentally construct the scene in a 3D application like Blender. Place the camera, the subjects, the lights, the environment — then describe what the camera SEES, not what you KNOW about the scene.
Loading actions...
<a href="https://prompts.chat">
TypeScript and ESLint rules that MUST be followed when creating, modifying, or reviewing any file under apps/frontend/, including .ts, .tsx, .js, and .jsx files. Also apply when discussing frontend linting, type safety, or ESLint configuration.
risks
Before writing any prompt, mentally construct the scene in a 3D application like Blender. Place the camera, the subjects, the lights, the environment — then describe what the camera SEES, not what you KNOW about the scene.
Image generation models build an internal 3D representation of the scene before rendering. If your prompt describes things that are physically impossible from the specified camera position, the model gets confused and produces inconsistent results. Thinking in 3D first eliminates:
Writers describe what HAPPENS. Directors describe what the CAMERA SEES. This skill makes you a director.
Before writing any prompt, mentally construct the room/environment:
HOSPITAL ROOM (mental 3D model):
- Room: ~4m x 5m, door on SHORT wall (south)
- Bed: against NORTH wall, head toward EAST, foot toward WEST
- Window: NORTH wall, behind bed headboard
- Bedside cabinet: EAST side of bed
- IV stand: EAST side, behind cabinet
- Floor: white linoleum
- Walls: white/cream
- Door: SOUTH wall, opens inward to the LEFT (west)
Define the camera as a physical object in the 3D space:
CAMERA (Panel 01):
- Position: SOUTH wall, in doorway, center
- Height: 1.78m (5'10" eye height)
- Direction: facing NORTH (into room)
- Lens: ~35mm equivalent (natural FOV)
- Tilt: 0° (looking straight ahead)
From the camera position, trace rays outward. What do they hit?
VISIBLE FROM CAMERA (Panel 01):
- LEFT edge: doorframe (wood), then west wall
- RIGHT edge: doorframe (wood), then east wall with cabinet
- CENTER-NEAR: floor stretching ahead (linoleum, ~4m to bed)
- CENTER-MID: Ovi running (her back/side, 2m ahead)
- CENTER-FAR: bed (3.5m away), Nikita propped up, window behind
- TOP: ceiling (but crop — 16:9 won't show much ceiling)
- BOTTOM: floor in foreground (close, detailed)
Before finalizing the prompt, verify:
When the camera moves between panels, think of it as a Blender camera keyframe:
Panel 01: Camera at (0, 0, 1.78) — doorway, looking north
Panel 02: Camera at (0, 1.5, 1.78) — stepped 1.5m into room
Panel 03: Camera at (0, 2.5, 1.78) — halfway across room
What CHANGES in the render:
- Doorframe: goes from visible edges → behind camera (gone)
- Bed: grows from ~20% of frame → ~50% of frame
- Ovi: was 2m ahead → now 1m ahead (larger in frame)
- Floor: less visible (objects fill more of frame)
- Cabinet: shifts from right edge toward center-right
Now write the prompt as if describing a render from this exact camera:
BAD (describing the scene):
"A hospital room with a bed, a girl running, and a mother holding a baby.
Warm lighting."
GOOD (describing what the camera sees):
"First-person POV from a doorway, 5'10" eye height. Wooden doorframe
edges visible at left and right margins. White linoleum floor stretches
ahead 4 meters to a hospital bed against the far wall. A toddler in
purple pajamas is 2 meters ahead, mid-stride, running AWAY from camera,
her back partially visible. At the far wall, a woman lies propped in
the bed, backlit by warm golden window light, face softly lit, looking
toward the camera."
| Angle | Blender Equivalent | Prompt Language |
|---|---|---|
| Eye level, forward | Camera at eye height, 0° tilt | "standing height, looking straight ahead" |
| Looking down at floor | Camera at eye height, -60° tilt | "looking down at the floor from standing height, steep downward angle" |
| Looking down at bed | Camera at eye height, -30° tilt | "looking down at the bed surface from above" |
| Looking down at baby | Camera at eye height, leaning forward, -45° tilt | "leaning over, looking STRAIGHT DOWN at the baby" |
| Looking at own body | Camera at eye height, -70° tilt | "looking down at own legs, foreshortened thighs, feet far below" |
| Tilted to side | Camera at eye height, 15° roll | "head tilted slightly, the room at a gentle angle" |
Every prompt should include at least ONE spatial anchor — a fixed object whose position relative to the camera tells the model WHERE the camera is:
| Camera Position | Spatial Anchor |
|---|---|
| In doorway | "doorframe edges visible at sides of frame" |
| 2m into room | "doorframe behind, no longer visible. Cabinet to the right." |
| At bedside | "white bedsheet fills the lower third of frame" |
| Leaning over bed | "pillow and headboard frame the subject" |
| Looking at own body | "hospital bed edge visible to the right side of frame" |
To sell depth and 3D space in a 2D image, include: