In my travels across the multiverse, the most magical places I find are forests. It does not matter which star system you are in, which galaxy or even which Universe. Nothing bewitches the mind like the heart of a forest. It beats with a rhythm unique to every forest and the surrounding flora and fauna dance to it's beat.
Not only are these enchanting forests a haven for a variety of local beings, but they are also a home to many rare and magical organisms seen very rarely elsewhere. The famed Magi of Girolden are known to hone their skills in the deep forest caves. The Brunicorn of Sinyter has only been spotted deep in its forests. The First One, when seen in the physical dimension, often chooses to be seen meditating in the forests of the Sentients....
Some of them are rumoured to have portals and wormholes within them through which The Watchers can pass undetected.
Hello everyone, hope you are doing fine. Thank you for your continued support and encouragement.
Here, I continue to delve deeper into the mind of the mysterious AI art generation program. I have used some of my artbreeder generated landscapes as initial images in the modified MSE VQGAN+CLIP script by @jotakrevs . I have noticed that we can generate consistent shapes and as a result more predictable results because of this.
I have not yet uploaded any of my images to artbreeder but I have read that the uploaded images are converted to an AI recognisable form. Therefore I infer that the images generated by artbreeder must also be of a form recognisable by AI. So, we should be able to get predictable results using AI generated images as initial images. Looks like that is the case.
Also, I tried a different aspect ratio for the images this time - the exact size being 680 x 382 pixels. This is almost a 16:9 ratio and looks good for landscapes and this pixel size does not give any out of memory issues while using the free version of Google Colab.
This is my artbreeder profile if you want to see what images I have used as initial images : https://www.artbreeder.com/atv8802.
All the images in this post are AI generated using the script mentioned above. The images have been upscaled on https://bigjpg.com/. There are no other edits to the images.
All the writing is mine, inspired by the images generated, my previous posts and my socio-cultural influences.
Thank you for reading and seeing these images. I hope you liked them. See you in the next one!
This is awesome! Both art and writing, and I like a lot how you are weaving the different posts, building a little world there!
For what I know about this, The AI doesn't care how the input image was created, any image gonna get checked at pixel level, shapes and other stuff, and converted to a bunch of numbers hehe
Use Artbreeder to create the init images sound very good. I did it once with Dall-E but it didn't bring me any improvement in the final results.
Regards and get this !PIZZA!
Thank you, am glad you liked the art and writing. Yeah, trying to create a lore kind of thing by linking all the posts….
Hmm, what you say is right technically…. Let’s keep experimenting!
And thanks for the Pizza too:)
This post is badass! Thank you for putting all the hot links to the other posts of yours. I like the creative theme you are working with... it seems to lend itself to this creative process. I have been using 640x480 in my sizes... but mostly for the video at the end.
I have played around with the initial image starting point... but never thought to use an AI generated image from the get go. Unless you count when i would use the last image of a run to be the initial image of my next run. =) Thanks for sharing with us and keep up the great work!
Thank you so much Castleberry :). Yeah, I am trying to link all characters and posts together to create a lore kind of thing instead of isolated posts. I am enjoying the writing part as much as the image generation.
Are you using Google Colab pro? Because I will get an out of memory error if I tried using 640x480 size for the images.
Hi @lavista, cool post and pictures. I find the 16:9 especially polished. Usually free colab and K80 should support 640X480 size (with VQGAN at least). Number of iterations should not influence memory consumption. Did you adapt any other parameter ? Otherwise i can only think about maybe the size of your init picture. Did you try the notebook with 640X480 without any init picture and just a prompt ?
Hi @dbddv01. Thank you for checking it out. I was just seeing your Broozel City guide series, pretty cool, what you did there.
Yeah regarding the image size, castleberry also said the same and I tried it. This time there was no error. Maybe I got a P4 GPU back when I tried and had gotten an error....
I love that you have your own lore going and that it all coordinates. @dbddv01 had a really cool series with a music-inspired theme. You may have seen that series he had a while back... if not you should definitely go check it out on his blog. =)
I find it so satisfying that everyone is getting inspired in their own ways about their own creativity and processes over these AI endeavors. It is almost as if we are being prompted rather than the other way around! hahhaa
I haven't splurged yet for the Google Colab Pro. I got that error when i tried to do the original size i wanted to do the art and video in which was 720x480. I usually end up with the K80 gpu when i connect... so i don't know if the gpu has anything to do with it or maybe how many iterations you set it for? I definitely don't get 12hrs so maybe it just cuts my time back to compensate? i have no idea really.... hahhaha
I will surely check out his music inspired series.
You are so right about it seeming as if the AI is prompting us, I had mentioned a similar thing in one of my posts too. I most definitely started writing for the public to see after I started generating these images with AI :).
I just tried again to test and indeed there were no errors with the 640x 480 size. Maybe I had got a P4 GPU and I might not have paid attention when I got the error....
Thank you so much for your support :)
PIZZA Holders sent $PIZZA tips in this post's comments:
@jotakrevs(5/10) tipped @lavista (x1)
Learn more at https://hive.pizza.
Congratulations @lavista! You have completed the following achievement on the Hive blockchain and have been rewarded with new badge(s) :
Your next target is to reach 300 replies.
You can view your badges on your board and compare yourself to others in the Ranking
If you no longer want to receive notifications, reply to this comment with the word
STOPCheck out the last post from @hivebuzz:
Fantasy and exciting. I like these illustrations of this hidden and magical world. I see that they work very well for you from an AI perspective. Also very good writing, it goes perfectly with the illustrations. Greetings! @lavista
PS: I invite you to use the #creativecoin tag in your publications, so that you can generate rewards in CCC in your publications.
Hi, thank you so much and am glad that you liked the writing too :)
I will remember to use the creativecoin tag from my next posts, thank you.
I have joined your community too, thanks again for the support :)
Hello, this is beautiful, delicate, I like it.
I follow you on Artbreeder
This is my profile: https://www.artbreeder.com/eveira
When one discovers AI, it becomes a vice, I have my computer full of images I have created, I also use Artbreeder and the images I generate there, sometimes I use them in other AI.
Hi. Thank you so much. I am glad that you liked it. I have been seeing your art on Alien art too, they are fantastic.
I see that you have been on Artbreeder for quite some time. Really beautiful art there too :)
Yeah so true regarding AI, I am just unable to pull myself away from it. All my experiments with it are taking up space on my computer too; I soon will have to think about deleting some stuff!