All the best AI models in one place
Flux, Stable Diffusion, DALL-E, Ideogram, RunwayML and more. Try it now for free!
20s on 20 - Dall-E 3
Created 2 years ago · 2 comments· 0 likes
Uploaded
The 20s on 20 Project
Comparing images made by twenty different AI models, all using the same prompt... well, except for this one. Dall-E 3 is notorious for refusing certain prompts, and this was no exception. I had to move the word "voluptuous" from near the front of the prompt for it to make any images at all.
I don't think Dall-E 3 did so well on this one. It has a tendency to make portraits look like plastic surgery victims. This was the best out of 12 images, not including all the completely failed attempts.
One slight thing in its favor is that it was the only AI model to have anything even remotely like the "red" I had specified near the end of the prompt. In general, it does seem like it reads more of the overall prompt instead of just focusing on bits and pieces.
The biggest point in its favor is that, while most models here can look similar to each other, which can get boring, Dall-E 3 is unique.
All the images: /tag/the-20s-on-20
Creation Settings
This creation was not made on NightCafe.
- Model
Dall-E 3
- Prompts
- A captivating 1920s-inspired portrait painting of a young, French ingenue with pale skin, dressed in an elegant, hyperfeminine outfit. She has long, dark hair adorned with an art-deco barrette, and her expression is a combination of innocence and allure. The background features a cozy Left Bank cafe, ...
- Workflow
This is the same prompt used with the other models as a test EXCEPT that I had to move "voluptuous" to the end of the prompt before it would generate any images. The infamous "Unsafe image content detected" kept coming up otherwise.
Thank you for doing this project, it provides very useful information!