Just compared two AI art generators side by side... wow the difference was huge

I spent last night running the same prompt through Midjourney v6 and DALL-E 3. The prompt was a simple one about a cat reading a newspaper in a 1920s diner. Midjourney got the lighting and mood right but DALL-E actually understood the newspaper text and made it look real. Has anyone else noticed one model is way better at handling text in images?

3 comments

3 Comments

black.pat20d ago

Yeah the text thing is a huge difference maker. I tried DALL-E 3 a few weeks ago on a sign in a coffee shop scene and it actually spelled out "Fresh Brew" correctly. Midjourney v6 is way better at lighting and detail but it still messes up letters half the time. The 1920s diner comparison is a good one because that lighting is where Midjourney shines. If you need readable text in your image DALL-E is the only real choice right now until Midjourney figures out that part.

ben_nguyen20d agoMost Upvoted

Right? It's like how autocorrect kills words in real life too, @julia_miller24 gets the short end there.

julia_miller2420d ago

Does anyone here actually get DALL-E to spell things right for them, or am I the only one who ends up with signs that say "Freh Bred"? @black.pat you got lucky with your coffee shop.