This is a 4-vector embedding of Billie Eilish made with a dataset of 90 images and 10000 steps.

I tested the embedding with different models. The sample images were generated with the following models.

Realistic Vision V1.2

HassanBlend 1.5

MoistMix V1

Protogen v2.2

Protogen x3.4

Protogen x5.3

Protogen x5.8

Protogen Nova

Dreamlike Photoreal 2.0

Art & Eros (aEros)

Analog Diffusion

Dreamlike Diffusion 1.0

Elldreth's Lucid Mix

DreamShaper

It seems to perform particularly well with DreamShaper , Realistic Vision 1.2 and the different Protogen models.

It performs worse with Dreamlike Diffusion 1.0 and Analog Diffusion

at least according to the tests I've done.