AI Art Prompt Engineering, The Merge and are NFTs done?
As we switched to much anticipated PoS Ethereum Network on the 14th, there was much speculation about the future of Ethereum. Now that the dust is settling, we can see interesting discussions about, is Eth now a security?, does it fall under US jurisdiction? Will eth prices skyrocket? Etc .. One thing obvious is, the energy consumption did not go away, most of it rather shifted from Eth to Eth Classic.
Days of selling animal cartoons for millions of dollars is probably over. OpenSea has seen a sales drop of nearly 90% within 7–8 months going from $5 Billion to $500 Million. That is a big drop!
This brings us back to Blockchain Tech. As I have quipped over the years, real value of NFTs is not in the monkeys but in utility. Monkeys were fun but there was no real utility or use of the tech to its full potential.
We are starting to see great companies building real solutions using Blockchain+NFTs to solve real world problems. For example, Anotherblock’s amazing solution for Music Royalties using NFTs, TravelX selling its first NFT Airticket from Madrid to Miami on Air Europa and Art Authentication services provided by PictureThis. These are true utilities of NFTs and the great value provided is totally immune to the ups and downs of the highly volatile Crypto and NFT markets.
2023 will sure be the year of Utility NFTs!
Now to the main topic, it has been quite a few days of fun with my new AI Art Toy Box. Thank you Stability AI and Gradio. Apart from generating endless images using complex text prompts, I tested out the img2img function which is designed to take one image and a prompt as input and generate an enhanced AI created output.
I uploaded a couple of rather crude computer drawings and what came out was quite impressive!
You can give it a go here:
Click on img2img tab, upload an image and click generate. You can use the CFG Scale & Denoising Strength sliders to fine tune the Artistic License of the AI. Once you see an image you like, you can feed it back to the AI using the Send to img2img button to loopback the image and continue developing it. It will take few iterations to see the magic at work.
Now the next interesting part is, how to talk to the AI in a way it understands you clearly. It is pretty good recognizing natural input (Raw Prompt), but, if you master “Prompt Engineering” you can make it work even more effectively.
Here goes a beginners guide to Prompt Engineering
A Raw Prompt is a natural language prompt, explaining something.
For example:
A warrior with a sword
People on the beach
Girl with an umbrella on a pier
You can add a style to this, for example, Realistic, Oil painting, Pencil drawing, Concept art
To get an oil painting, you can use:
Oil painting of a warrior with a sword
You can combine “Realistic” and “Photograph of” to get more realistic or photo quality output.
Next is of course the more controversial option of using an Artist’s style. If you like to use this, you just have to add the Artist’s name to the prompt. For example,
A warrior with a sword , Paul Gustav Fischer
You can add some amazing finishing touches by using detailing prompts. For example,
Cinematic lighting, Unreal engine, Highly detailed, trending on art station, dramatic lighting, octane render, Matte finish, washed colors, ambient lighting, epic composition
Busy street at night cinematic lighting
Busy street at night epic composition
You can test and experiment with more phrases on my SD test instance.
Have fun!!!
