The one I tortured extensively back in 2024. History matters.
OpenAI's video-to-image darling. Finally taking visuals seriously.
Google's offering, apparently named by someone who lost a bet.
Best known for making Elon Musk memes. We'll give it a shot anyway.
The model that every AI enthusiast insists will "totally nail medical accuracy this time."
Ghost's own AI medical image generator, trained on thousands of Ghost-created illustrations. Domain-tuned, not vibes-tuned.
Made by actual humans using Maya, Redshift, Photoshop, Houdini, Substance, and about a dozen other tools that reward anatomical knowledge over clever prompting.
We excluded Stable Diffusion from this test because it constantly refused to render requested images, flagging them as "too graphic." Medically accurate imagery is not offensive or shameful. We don't have time for pearl-clutching algorithms. Until this AI matures enough to realize that nothing is offensive about medically accurate imagery, we won't even consider it.