Comparison of Generative AI Tools
a look at prompt writing for generative AI tools and how the various tools compare
PROMPT WRITING
I spent time learning how to write prompts for generative AI tools, including aspect ratios, style, and formatting. I also completed the Linkedin Learning course, Midjourney: Tips and Techniques for Creating Images.
COMPARISON OF AI TOOLS
I inputed the same exact prompt into four different industry leading AI tools to observe the different outputs each gave. Scroll down to see the results and read my thoughts!
THE PROMPT
An Avant-Garde Hotel Lobby in Blue, Red and Yellow
DALLE-3
The newest model of by OpenAI created the most futuristic, whimsical, intricate, dreamscape versions of the brief. The images are rich in color, texture, and furnishings. The light and reflections on the shiny floor are better than expected, though the tile patterns aren’t quite right. And while the images are not entirely realistic, it might take a few looks to understand they were generated.
MIDJOURNEY
Midjourney’s images blew me away! For the most part, they look like they could be actual hotels, with a reasonable level of whimsy. The use of sunlight, shadow, textures, reflections, plant life and even Gaudi-like inspiration is all way more advanced than the one sentence prompt I inputed. Call me impressed!
I also love the ability to view what other creators prompt and the images outputted in Midjourney’s public discord channels.
ADOBE FIREFLY
These images are almost there…but not quite. The lobby settings are mostly realistic. But the lines are too wavy, the images are oversaturated and overexposed. In some places, it looks like items were cut, pasted and not blended well into the image.
CANVA AI
Canva’s AI tool is built into their web-based design software, which impressed me. While the colors are one-note and the furniture is odd or unrealistic in many places, some of the compositions are great. And the use of natural light and plant life in some of the images is better than what we saw with Adobe’s tool. I think with some toggling of the prompt, the images could be improved.
© 2019–2024 Ruthie Greenblatt