THE PROMPT
An illustration to be a blog post thumbnail cover that represents a blind vote on ai art. Text that says "Blind Vote for best Image Model"
EDIT: Didn't release this as a poll type post so you'll need to vote here: @jarvie/re-peaksnaps-snwys9
Choose top 3 images that reflect the PROMPT above
A lot of my tests are seeing how well the models handle text because a lot of my needs are images with text.
I could get a subscription to Canva or even use free version or i could use some other tool... but is it better to just run each model twice essentially doing 16 image model runs resulting in 22 images and get something potentially pretty awesome and spend less than $1
Sadly right now https://peakd.com/AI doesn't have a fast way to run every model with one click... but if it did then it would ALSO be the fastest option.
Interestingly looks like about 6 models got pretty close to the exact text (which is not always the case) and sometimes it's easy to take some of the smaller mistakes like when there is an extra word like in image 7 and remove that word in something like lightroom. Some others have the wording but very strange layout of the wording.
THE IMAGES
IMAGE 1 (Below)
IMAGE 2 (Below)
IMAGE 3 (Below)
IMAGE 4 (Below)
IMAGE 5 (Below)
IMAGE 6 (Below)
IMAGE 7 (Below)
IMAGE 8 (Below)
IMAGE 9 (Below)
THE METHEDOLOGY
STEP 1. Ran each model twice with Same Aspect ratio
NOTE: Schell gives 3 images at a time and DEV gives 2 by default
STEP 2. I Chose the best image (Favoring correct spelling)
STEP 3. I added in midjourney 1 run = 4 images (raw 6.1, stylization 400, 20 weirdness, 200 variety)
YOU NOW VOTE
- Pick your top 3 results in the poll
- I won't share the models right now...
- Later I'll share which models ranked highest
The cost of the test is 0.88HBD (does not include midjourney)
Published with PeakVault
