stereogramification

Prompt requirements are defined in lib/mask_context.MD. Paste into your image model, export a square grayscale PNG, then use Upload here.

You are generating a DEPTH MASK for an autostereogram (Magic Eye image).

IMPORTANT: This is NOT a visible illustration. This is a grayscale depth map.

Definition:
- Black (#000000) = farthest depth (background)
- White (#FFFFFF) = closest depth (foreground)
- Gray values = intermediate depth levels

Core Requirements:
1. The image must be strictly grayscale (no color).
2. The background should be pure black or near-black.
3. The subject should be represented using smooth depth gradients:
   - Avoid flat white regions unless intentional
   - Use soft transitions between depth levels
4. The shape must be clearly readable as a silhouette AND in depth.
5. Avoid noisy textures — this is NOT a pattern.
6. No outlines or strokes — depth should be implied through shading only.
7. No text, no symbols, no visible overlays.
8. The object should be centered and occupy ~60–80% of the frame.
9. Use soft lighting logic:
   - Simulate depth as if lit from the front
   - Brighter = closer to viewer
10. Edges should be slightly softened to avoid harsh depth discontinuities.

Style Guidelines:
- Smooth, sculpted, almost 3D “heightmap” look
- Think “clay render” or “depth pass” from a 3D engine
- Clean, minimal, high contrast between subject and background

Output:
- A single grayscale image representing the depth mask
- Square aspect ratio (e.g., 1024x1024)

---

User Input:
"your subject"

---

Generate a depth mask of the subject above, following all rules.
Do NOT render the subject visibly — only encode it in depth using grayscale.

Prompt requirements are defined in lib/tile_context.MD. Paste into your image model, export a square PNG, then use Upload here.

Generate a seamless square tile image designed specifically for use as an autostereogram base pattern.

SUBJECT: your subject  (e.g. "dragon", "horse", "jungle", "abstract shapes")

STRICT REQUIREMENTS:
- The image must be a single tile (NOT a repeated pattern preview).
- No hidden images, silhouettes, or depth shapes — pattern only.
- High-frequency detail everywhere (no large flat or smooth regions).
- Absolutely NO horizontal bands or rows of similar color or texture.
- Color variation must be chaotic and evenly distributed across the entire tile.
- Avoid gradients that run horizontally.
- Avoid repeating stripes, lines, or directional artifacts.
- Texture should feel noisy, organic, and complex at the pixel level.

STYLE GUIDELINES:
- Dense, cluttered composition (like a zoomed-in chaotic texture).
- Overlapping elements, randomized scale and rotation.
- Mix micro and macro detail (tiny + medium shapes).
- Use strong contrast in hue, brightness, and saturation.
- Prefer organic randomness over symmetry.

TILING:
- Edges must tile seamlessly on all sides.
- No visible seams when repeated horizontally or vertically.

CAMERA / FRAMING:
- Extremely zoomed-in view (macro texture feel).
- No clear focal point or subject center.
- Fill the entire frame edge-to-edge.

OUTPUT:
- Square image (e.g. 1024x1024).
- Crisp, high-detail, no blur or depth of field.

GOAL:
The pattern must work well for stereogram generation, meaning every horizontal slice of the image should contain varied pixel information so horizontal pixel shifting produces a visible depth illusion.