None of my art attempts using Artificial Intelligence, or generative art, has ever really been fully explored or fully expressed. Each run or attempt has been a test run or series of 'sketches' of sorts.
Generating images is something i have been doing to have fun and destress a little. I haven't been going at it with any real plan. That will change soon.
So for now, it's always fun to experiment with some of the old images i have already done before. I tend to revisit some of the old prompts and renders from time to time. There have been so many upgrades and additional add-ons that using the same prompts and seeing the differences is pretty epic.
Before i get going, i wanted to share with everyone the opportunity to do this yourself for free!
Come join Ausbit's Discord and illustrate your imagination via text prompts using the bot with a zillion settings for free by clicking this invitation link. I highly recommend coming and giving it a try! You can tag me on discord @ castleberry#6859 and i will help you out the best i can. Or, if i see you there, i might just tweak your prompts with you and save you some credits while helping you refine your concepts.
Now, here are a few examples where i used the same prompt for both images. I went back and made most of all the settings the same. The main difference is the model. I used the same sampler and specs in the prompts but, the model is a newer custom hybrid we came up with, and it supercharges some of the imagery. I think i can turn the model off too to try and get the best comparisons. Either ways, enjoy the spectacle.
This image below is one of the first runs that i did once i joined Ausbit's discord to try out his bot back in 2022's early October.
1a Original Result
I ran the prompt below for the Original Result and the following images in this run.
1b New Result
I tried to run the same and remove the model from the settings, since i didn't originally use one on the original. Now, the custom model is part of the default settings, which is awesome!
1c Retry Result
I realized that i could not remove the model after running # 1c. So, i decided to tweak the prompt a little and adjust some settings for higher quality and better sized images. Below, are a few of the final images i ended up with, just to see what could be produced. I tried not to get too lost in the process, since this was just one of many intended prompts and renders to revisit!
1d Tweaked Prompt and Oriented to Portrait
1e Tweaked Prompt and Oriented to Portrait
Of course, one of my first things to try and get some AI Art of was Bob Ross! I did quite a few Bob Ross runs with all kinds of results. Some results were ok, some were quite terrible... others were glorious. Below, is an image that i was most pleased with. It wasn't my first try... but, it was one of the best outputs i have achieved so far.
2a Original Result
The Golden Perm of the Autumnal foliage hair was epic. Bob Painting a self-portrait couldn't have turned out any more rad than that! That didn't stop me from revisiting this render! You can never have too much Bob Ross and anything less cool would just be chalked up as Happy Accidents!
Here is the prompt for the original image and these new variants.
2b New Result
2c New Result
2d New Result
We will see some more of Bob here in a bit... but for the next run i wanted to remind you that when i joined the discord there it was around Halloween time. So, a lot of the earliest things i experimented with was pumpkins, jack-o-lanterns, and other assorted Halloween themed things.
The results were mostly less than desirable so it took a lot of runs and tweaking back then to get anything worthwhile. Every now and then a gem would pop out at random... but mostly, the AI hadn't been trained enough for those specifics at the time and therefore, had too many options to draw from.
This created a lot of extra 'noise' in trying to generate a decent image. This is where prompt-crafting would come in. Sometimes, even that wouldn't help.
3a Original Result
As you can see, that turned out super weird when you compare the image above to the prompt below. Here is the prompt for the new results and the original image.
3b New Result
3c New Result
3d New Result
These new results were a bit more of what i originally had in mind. That first one looked like it was making Halloween propoganda and got confused. I could have spent more time on this tweaking the prompt and newer settings, but i thought this would suffice for now as just an example.
I haven't shared a vast majority of the images i have generated. Fortunately, i never really lose the images or have to get rid of them since they are in Discord's interwebs.
Sometimes, i just generate images in the background for hours. It's really fun until i realize how disorganized i am with my generated artworks. Making this post, over a few sitdowns in free time and sessions, has been a nightmire. Hahaha
For this post, i wanted to share a variety of the images but not go into so much detail about the settings and prompts. However, that didn't really work, as most of my intended directions of post making do not.
In this case, i blame it on a few happy accidents. Now, i am just trying to Surrender to the Render and share some cool stuff... old and new.
4a Original Result
The image above was one of the first images that i found extremely stunning. I used the prompt below and ran it a few different times before getting the one that led to this original result. I think the curls of the hair and the lighting of the background, skin, and hair is what really drew me to this image... along with the contours.
Here is the prompt for the original image and the inspriation for this next run.
4b New Results
For the image above, 4b, i used the exact same prompt with the settings as close to the same as i could get. The difference was interesting. Rather than doing a bunch more to try and get something closer or more similar to the original, i decided to spend my time just furthering the concept with settings and prompt tweaking.
4c Updated Inspirations
4d Portrait & Prompt
I usually don't switch to portrait or landscape orientation until i feel i have something worth working more on. I will tweak on more runs trying to refine the image from there. Fortunately, for this run some of the artists i used in the updated prompts do (or did) fantasy art portraits.
I did many more of these variations and they kept getting better and better with minimal refinement of the prompt. I could have gotten really creative and switched up a few things to change the whole design... but tried to stay with a spectrum of consistency, which is very tough to do with AI art.
Maybe i will share more of those another time. Besides, i didn't want to have too many titties in one post and i'm already over in titties and words, than my original intentions for this post.
Early on, i tried H. R. Giger style with an easy prompt to test out the bot. He was always one of my favorite personal art influences growing up. I figured an easy prompt would give more room for the style to be revealed.
5a Original Result
H. R. Giger has become a pretty popular AI Art influence for his biomechanical themes and elite level of detail. I tried to keep this second image as close to the original settings as possible with the exact same prompt. It came out similar but different as expected.
5b New Results
5c
5d
5e
It has taken a lot of time and a lot of practice mixed with keen observations to engineer prompts effectively. I am not claiming to be a pro or any kind of expert, but i would say that my results are to my liking and better than average more often than not! Some of this i attribute to my prompt craft... but a good portion is from the specs of the Ausbit Arty Bot.
Many of the runs are based on the subject matter and how it's detailed along the way of generating a composition. There are different ways to take advantage of both options in their myriad of varieties of how you direct them to be manifested visually.
Now, for the Grand Finale! I promised more Bob Ross for you and i didn't forget. I assure you, this probably wasn't the Bob you were expecting. Enjoy this last run before we call this post finished.
Thanks for stopping by! I really do appreciate your time and attention. I will be giving a little extra curation to more meaningful comments. I will also be giving even more curation for those that engage within this comment section with each other.
I would love to know which images are your favorites and why! It would also be super cool to hear any suggestions or ideas you have for me to try out later. I would give even more curation for anyone that shares something of their own from Ausbit's bot in the comments!!!
Until next time, y'all take care, and thanks again for dropping by!
