One of the fun things about Web 3 is getting to experiment around with different things. Ice Planet is now my third interactive fiction. It is my second one fully on Hive in the form of posts. There is however a bit of an issue with doing it that way.
I have already talked a bit about my past of making interactive fictions and the level of organization among other things in creating them in Ice Planet An Interactive Fiction Now Available On Hive which was posted on my main blogging account.
I also touched a bit on AI artwork and a bit of an internal struggle in even wanting to try using it in the first place. Doing so I feel has allowed me to take my interactive fiction creation to the next level. I also feel it might also solve another issue.
The biggest issue is this project itself was quite time-consuming. It’s over 70k words in total and includes 59 images in the story itself. Along with some other ones.
This means I more or less wrote an entire novel and just released it all over a couple of days through around 169 posts. In a week those posts will close and that is it. Whatever I earn if anything at all that is about it. That is an insane amount of work just for a window of a week of earning potential.
It is true people can choose to tip. Tipping however on Hive has not been a huge thing. I more than get it. People can just upvote a post to show their support if the post is still open. I have never been a big tipper when it comes to content.
In the past, I’ve given leaving an open comment weekly on some very long-term things. My results of doing that in the past were not worth the time and many times they even closed getting a vote but under were dust.
What people tend to want for giving up the red Hive stuff is something further in exchange. That is where I feel NFT artwork from the story can come into place. It allows people to further show support if they wish and they get something out of it for doing so.
Keep in mind the artwork is AI created. Before going into this creating AI artwork I knew very little. I know some people that have done their own AI artwork projects in the past on other blockchains. I however wanted to get down and dirty exploring this avenue further myself.
Many of the amazing sites to create AI artwork on now have gone into a fully paid model. Some of them are also getting quite pricy. I wanted to see if I could create AI artwork without paying anything extra and still retain commercial rights to that work.
Platform I Used For Making AI Artwork
That is where the platform Playgroundai came into play. At the time of writing and using them they allowed me to generate 1k images per day for free. Once you hit that daily limit they use an increasing time delay to throttle you back but still allow you to create further artwork if needed.
The biggest thing is they do not have a daily reset set time. Instead, it’s a rolling limit. So once you go over 1k images and exceed that by continuing to keep going it could be a while before you fall under that 1k limit and can start making further artwork without getting time limited.
The free version is further restricted in quality and size among other things. The other big kicker I ended up finding out is they limit you to 30 x4 scales per day. Originally when I was just making the artwork for the posts that was not an issue.
However, when I decided to upscale those pieces of work to sell as NFTs that daily limit was not the only issue I ran into. AI upscaling is not always perfect. A couple of pieces that I ended up swapping out of the posts as they ended up having strange glitching in the artwork itself in the upscale version. Thankfully it was not more
Making The AI Artwork Itself
The biggest issue going into this that I thought I had with AI artwork is the lack of coherence between the pieces. They really can be all over the place. That is why I spent a while testing out filters (presets) along with trying out many of my own.
I then found one that was going to work decent enough along with my prompt. That I would attempt to use for the remainder of the project.
The biggest thing to note is I first wrote around 95% of this interactive fiction before generating more than two of the artworks. I did not want to create artwork and then base the story around them. I wanted to write the story and then create artwork based on the story.
It turns out that is a lot harder and took way more time than I was expecting. In many instances, the more details I gave the AI prompt the less of what I wanted occurred.
Cranking up the prompt guidance literately gave me total unusable nonsense such as the piece featured above which was supposed to be an underwater tunnel. I don’t see it at all.
Another big thing I wanted to use AI for over using stock photos is the theme of this interactive fiction has a lot of stuff you are not going to get a stock photo of. While the AI was able to do some of the stuff I wanted. You will notice the stuff I was able to do best with it were landscapes of forests, mountains, and beaches. That makes sense with how AI learns to create artwork.
There was some stuff I wanted to have artwork for. After far too many attempts it was clear I was not going to get what I wanted with my understanding of AI. A lot of the time I was aiming for close enough if I was more than fine using it or that I could finish it off from there.
There were quite a few pieces of artwork I ended up using in the story that took well over 50 attempts to get to where it was. I would not be shocked if there were some closing in on 100 attempts. It was quite rare even after using the site for a while to get on the first few attempts something I would end up using.
After a while, I ended up developing my own method. I would first start with a rather generic prompt of kind of what I wanted. I would then start to add further details in the prompt or exclusion words to see how that would start to shape things.
Sometimes that would start to get me closer to what I wanted. Other times I seemed to be drifting a bit further away from the target. So I’d make what adjustments I needed and would continue. Many times it felt like I was on the right track to the point I just needed to find something I could drill down further on.
Once I got to that point it was time to switch over to having a variant creation based on the image I wanted to use. I would then go through a few of those.
Till I felt I was getting even closer to what I wanted. However, there would always be something not quite right. Sometimes adding further exclusion words or changing the percentage of the variant would be based on the original was good enough.
Many other times it required a bit more hands-on approach. That is where the inpainting mask tool came into play. I could simply remove the parts of the artwork I did not want. I then could continue from there creating variants.
After a lot of trial and error, changing my prompt, adding exclusion words, variant generation, and inpainting I’d get a result I was happy with sometimes.
Even with all that, there would times be due to part of the prompt I needed to keep to try and maintain some level of coherence across all the artworks it would not be quite right. That is where I caved slightly and was willing to edit up to five words in a post if that was what was keeping things from working.
If that was not enough then I’d just move on to the next post. There were a lot of failed attempts that lead to nothing I found useful. I found it kind of funny how I would think something would be so simple yet the AI would get it so wrong from what I wanted.
Pricing
As far as the pricing for the AI artwork I’ve set it to be more in line that you are supporting the interactive fiction story element more than the artwork itself. While the majority of pieces will be listed for sale some I’ll be holding back. Some might find out later why.
They will be as following:
- First edition: 100 Hive
- Second edition: 50 Hive
- Third edition: 25 Hive
- Fourth edition: 5 Hive
- Fifth edition: 5 Hive
I wanted there to be some rather reasonable prices for some of the editions. I feel 5 Hive is a decent spot on the lower end. There are also higher amounts for the first edition and so forth for those that want to support something like this further.
If this part and post payouts end up doing really well I’ll consider making another interactive fiction for next year. At this point, I’m able to make one of these a year. I might also consider going back on my other interactive fiction on Hive and making some AI artwork for that as well and listing them for sale.
Final Thoughts
Creating AI artwork ended up being a lot more time-consuming than I thought it would be. It was quite the learning experience regardless of the outcome that ends up coming of this. I love writing this interactive fiction and feel like this one has dramatically increased over the last one I created. Adding a couple of new elements to it including AI artwork was certainly interesting to give a try.
Please note you are NOT required to hold any of these NFT artworks to read the story. They also do NOT provide any benefits from doing so.
If you would like to check out the AI artwork for sale you can do so by looking at the Ice Planet gallery on NFT Showroom.
Information
Content is written by @Enjar of @Enjargames. Screenshot and AI artwork from Playgroundai.