Astro 3D using ChatGPT based on GAIA data.

发布于

After several attempts we (ie. me & Chat 🤣) arrived to quite useful command template. Here it is:

 

Create a wall-eye stereogram from the attached astronomical image.

Preserve exactly:

* original brightness,

* contrast,

* colors,

* natural noise,

* star halos,

* object scale,

* image proportions.

Do not perform:

* sharpening,

* noise reduction,

* HDR processing,

* Local Contrast Enhancement,

* saturation changes,

* generation of new structures.

Before generating the stereogram, remove 50–70% of the faintest background stars (the smallest stellar profiles), while preserving all stars that are important to the structure of the main object.

Star depth:

* apply GAIA Natural Depth,

* nearest stars receive the largest parallax,

* more distant stars receive proportionally smaller parallax,

* maintain a continuous depth gradient,

* avoid grouping stars into flat depth layers,

* preserve the intrinsic spatial structure of open clusters.

Nebula depth:

* treat nebulae as volumetric objects,

* bright Hα regions slightly projected forward,

* dark dust lanes slightly recessed,

* gas filaments distributed across multiple depth layers,

* preserve all faint structures and dust features,

* avoid the “cut-out sticker” appearance.

Stereogram composition:

* place the main object exactly at the center,

* no vertical displacement between left and right images,

* identical photometry in both panels,

* side-by-side stereo pair,

* optimized for parallel viewing (wall-eye).

Parameters:

* type: GAIA Natural Depth,

* strength: 140% of natural parallax,

* stellar depth: high,

* nebula depth: medium,

* depth-gradient smoothness: maximum,

* viewing comfort: high.

Priority:

Astronomical realism > 3D effect strength.

If there is a conflict between stereoscopic impact and realism, choose realism.

Avoid CGI appearance, artificial 3D modeling, and flat 2D layers.

The final result should resemble a realistic three-dimensional observation of the night sky while preserving full photometric fidelity to the original image.

讨论 5
收藏