358 - Seeing Leonardo

In blog post 356 - Visual Gook I was thinking about descriptions of visuals and how AI might learn about images.

Here is the alt text description for 'Ginevra de' Benci' (by Leonard Da Vinci) on the US National Gallery website.

alt="This square portrait shows the head and shoulders of a young woman in front of a spiky bush that fills much of the background except for a landscape view that extends into the deep distance to our right.

The woman's body is angled to our right but her face turns to us. She has chalk-white, smooth skin with heavily lidded, light brown eyes, and her pale pink lips are closed. Pale blush highlights her cheeks, and she looks either at us or very slightly away from our eyes. Her brown hair is parted down the middle and pulled back, but tight, lively curls frame her face. Her hair turns gold where the light shines on it.

She wears a brown dress, trimmed along the square neckline with gold. The front of the bodice is tied with a blue ribbon, and the lacing holes are also edged with gold. A sheer white veil covers her chest and is pinned at the center with a small gold ball.

The bush fills the space around her head with copper-brown, spiky leaves. A river winds between trees and rolling hills in the distance to our right. Trees and a town along the horizon, which comes about halfway up the painting, is pale blue under an ice-blue sky."

I turned the image upside down and asked Gemini to describe it for me. It seems to have learn the concept of 'upside-down' for this context.

Gemini's description

This image is a Renaissance-style portrait of a young woman, shown from the chest up, but it is oriented upside down. The woman is positioned against a landscape background with a prominent tree.

Succinct Summary

A 15th-century oil portrait of a pale woman with curly golden-brown hair, set against a dark evergreen tree and a distant water landscape. The entire image is inverted, so the woman's head is at the bottom and her shoulders are at the top.

Detailed Description

The Subject

The woman has a very pale, luminous complexion with a smooth, almost marble-like texture. Her hair is styled in tight, intricate curls that frame her face and forehead. Because the image is upside down, her chin points towards the top edge and her forehead towards the bottom. Her expression is neutral and reserved, with a steady gaze and a small, closed mouth.

Clothing and Attire

She wears a dark, possibly black, gown over a white undergarment.
The white chemise features a high, straight neckline with delicate gold embroidery or trim.
At the center of her chest, there is a small blue lace-up detail.
The fabric shows fine horizontal creases, suggesting a thin, delicate material like silk or linen.

The Background

Immediately behind her head is a dark, dense juniper tree with sharp, needle-like leaves.
To the left side of the frame, the background opens up to a serene landscape.
This landscape features a calm body of water reflecting the blue of the sky and several thin, tall trees on the far bank.
The sky transitions from a soft blue at the horizon to a slightly darker tone.

Artistic Style

The painting is highly detailed and uses a technique called sfumato, which creates soft, smoky transitions between colours and tones. This is particularly visible in the gentle shadows around her eyes and the corners of her mouth. There are also visible fine cracks across the surface of the paint, known as craquelure, which is typical for an aged oil-on-wood panel.

Would you like me to identify the specific historical painting and artist this image represents?

Fail

I asked Gemini to turn the image upside down and it failed. I got this.

It left it upside down but modernised the face.

Her nose is a little wider at the top, her eyes look into your eyes and her lips are fuller and rosier.