Since AI technologies starting popping up all over the place, I’ve been really interested in seeing where the art generation aspect can take me. Can it unintentionally open a window into hell? Can it be usedul for illustrating stories? Can I easily generate components to use in collages?
I swear at one point I posted a comparison of several different AI art generators, but now I can’t find it.
Anyway, last time, I looked at some of the settings in the StarryAI app, (the preferred app coming out of the multi-app comparison post that I can’t seem to find)…but used a fairly complicated prompt, so today I wanted to play around with something simpler to see ehat else we could learn.
I’ve been logging into the app as often as I can remember, racking up the free points to allow myself a better look at how settings and prompts impact the output from this particular AI art generator…today I finally felt confident that I’d accumulated enough to allow for a comprehensive glimpse…if such a thing can in fact exist.
For today’s subject (which did morph some as the experiment unfolded), I decided I would try generating a still piece, featuring essentially just an acoustic guitar (plus whatever the AI wanted to add).
For the first round, very basic start; no exclusions and very little artistic dorection…I wanted to start by seeing how the engine would interpret a very low resolution prompt such as this and, i think the results were pretty cool, but not what I was after.
…so I stayed in the Anime style setting, added “still piece” to the prompt but kept all other elements the same, then added exclusions that I thought would get rid of the people from the images:
…but the girls were still there, which meant i would need to modify my request.
This one taught me something, but not until just now. As I was gathering screenshots and writing this section - when you evolve a previously generated image, the image is now referenced in addition to the prompts, so next time I should copy my prompts and paste them into a fresh generation; I had not realized I was carrying over so much of the previous image, but now the results for Rounds 3-5 make more sense.
For this round, I kept prompts the same but switched styles, from Anime to Abstract World.
Despite my written exclusion of all things human, I keep getting people. As mentioned above, we now have a good idea how that happened.
ROUND 4 was the same pretty much. From 4 to 5 i changed “Salvador Dali” to “Style of Salvador Dali.”
This was because, first i changed from Alphonse Mucha to Dali, just trying to get the people out of the image. Inknow Mucha usually had women in his work, so I thought maybe that was it. But then (in Round 4), it looked like the generator was trying to produce images of Dali playing the guitar, instead of a guitar painted in his style…again, we now know how that happened.
ROUND 6 was the same as 7 pretty much, but the new image was poorly cropped.
For round seven, I found a picture of a guitar by itself and added this to the prompt. Since I was starting from a new base image, at this point I brought Alphonse Mucha back into the prompt.
Round 8 was all the same, but “Acoustic Guitar” was changed out for “Baritone Ukulele”. Round 8 was still rendered in the same style as Round 7.
So for Round 9, we changed the style from Abstract World back to Anime…I wanted to see what the ukulele would look like in some of the various styles, now that I had the base where I wanted it.
Style change from Anime to Argo.
Style change from Argo to Inkpunk.
Style change from Inkpunk to Portrait.
For Rounds 13 & 14, I played with previously used styles, but also tried to add a blackbird.
In Round 15, I kept the blackbird (although it didn’t really show up much), but tried a new style.
Style change from Portrait to Lyra, “blackbird perched on […]” added to prompt.
StarryAI gives (free/non-subscription) users 5 free credits daily, but like Hive, you have to log in to claim them. Each Round of this experiment cost one credit. As far as the free generators go, I think this model gives me enough freedom to play and experiment.
I learned a few things in the process of writing up this post, and also of course during the experiment itself. For now, I will go back to saving up credits to build up a big enough stack to support another investigation.
…what’s your opinion on AI generated art? If you’re not completely against it, do you have a preferred generator? I make my own art as well, but ever since this became so easily accessible, I must say it is FUN to play with!
Thanks for checking out some more of my work! As always, I hope you enjoyed witnessing as much as I enjoyed creating!
© Photos and words by 
If you enjoy my content, check out the Albus Index!
Here you’ll find links to various post-groupings, so you can find all posts related to a specific topic quickly. Topics include lists of brewing recipes, flutes, crypto talks, rucking posts and an index of all acoustic covers. The lists are updated as new content is added, so they can always be used to quickly navigate my content.
I created a second account (
I’ll be using this as my account to share with bars and venues etc., so it will be a home for acoustic covers, live performances and auditions…pretty much, if I’m playing a stringed instrument amd singing, this is where I’ll be doing it…unless I’m doing it while operating a motor vehicle ;)
Click or tap the profile banner below to check it out!
