With the very busy preparations of exams, I figured I could try some new things while taking a break from studying. Most of the time, humor works best for me to refreshen the brain.
I noticed that a lot of my peers already tried Dall-E, but I was wondering if there are also other tools worth using to generate images, and what are the differences between these tools.
In my search for AI-driven image generators, I noticed that there are tools that use external platforms, require the user to use code or are very specific in theme. For example, Midjourney (in its beta version) currently uses Discord and requires its users to do so as well (for now). Whereas, This Beach Does Not Exist has the theme of, surprise surprise, beaches.
However, the tools I found relatively easy to use and are free were Picsart, Stable Diffusion and Neural.love.
In my quest to generate some random, but funny images, I used the prompt “Human-like Capybara at a victorian-era fashion show in the renaissance period”. Please don’t ask me how I came up with it, half the inspiration came from TikTok. Please find below the results:
Picsarts – Best capybara, more in Bridgerton style rather than a fashion show
Benefits
(+) Tool is very intuitive to use
(+) Generation is quite fast and best carried out the human-like capybara
Disadvantages
(-) Overall accuracy (including setting) of carrying out the prompt could be better
(-) Output always has a certain style (when comparing to the second prompt below)
Stable Diffusion – Spiky haired bird with ghosts
Benefits
(+) Image looks more historic
(+) Output is relatively easy to customise in terms of dimensions (not sure it’s too relevant though)
Disadvantages
(-) Accuracy of output is incorrect – rather looks like a bird surrounded by dressed-up ghosts
(-) Generates a bit slower
Neural.love – Best setting with a glitched capybara
Benefits
(+) Generates 4 images at once and lets you choose
(+) Easy to fill in prompt (and adjust via prompt engineering) and suggesting the style of image (e.g. fantasy, cyberpunk, etc.)
Disadvantages
(-) More complex prompts tend to still be difficult for the tool (e.g. with layers)
(-) Small errors in the image (the mouth was placed in the neck, but extra points for the sassy gentleman)
I attempted a second prompt: “Snoop Dogg being chased by a demon”, – the output was most definitely interesting. Any thoughts? And do you have a favourite tool I should be aware of for my next exam period?