Big 3 AI Image Gen

September 19, 2022

This is a quick visual comparison of various settings for the same prompt between Midjourney and Stable Diffusion via DiffusionBee and DALL-E 2. For every prompt, the top row of the gallery is Midjourney, the middle row is Stable Diffusion, and the bottom row is DALL-E 2. High resolution versions are available by clicking on the image.

I’m not covering the image-to-image abilities of Stable Diffusion, the image sourcing of Midjourney, or the outpainting capabilities of DALL-E 2.

Prompt 1

a robot dressed as a medieval knight drinking a goblet of wine

v3, medium style, max upscale v3, medium style, max upscale
 v3, very high style. max upscale v3, very high style. max upscale
test, medium style, max upscale test, medium style, max upscale
25 steps, 7.5 guidance 25 steps, 7.5 guidance
25 steps, 2 guidance 25 steps, 2 guidance
50 steps, 7.5 guidance 50 steps, 7.5 guidance
first first
variation variation
variation variation

Prompt 2

quirky underground comic style illustration, pen and ink and watercolor and collage in the style of Marc Bell, bathing cats, bath house onsen cats

This prompt was taken from this entry that I found on the Midjourney community feed. I just liked it.

v3, medium style, max upscale v3, medium style, max upscale
v3, medium style, max upscale v3, medium style, max upscale
v3, medium style, max upscale v3, medium style, max upscale
40 steps, 7 guidance 40 steps, 7 guidance
40 steps, 7 guidance 40 steps, 7 guidance
40 steps, 7 guidance 40 steps, 7 guidance
first first
variation variation
variation variation
art ai midjourney dall-e2 stable-diffusion