I'm amazed by the rapid progress of creative AI. Midjourney recently went into open beta, giving anyone the opportunity to hop on and create art from text prompts. (They have a pretty generous free trial; seriously, go check it out!) They've trained an AI model on a huge portion of the internet's art, essentially creating an "artist-in-a-box" that taps into the collective creative conscious.
There are many big (and troubling) philosophical questions here, but I think technology like this opens up many doors for worldbuilders. In minutes, you can make concept art for your world or D&D campaign. People, places, monsters, maps, items... And the joy of this is that there's some mystery for you as the creator. One of the problems with worldbuilding is sometimes we know too much about our creation! But an AI creative assistant can riff on familiar tropes, while still surprising us with ideas stemming from randomness.
The process of creating AI art is not perfect. The final result is almost always a compromise between you and the AI. But go look up Midjourney examples from two months ago, and then look up examples from today. The progress is kinda scary. We're at the beginning of a revolution for every type of creator: Imagine combining GPT-3, Midjourney, and your own creative direction to manifest entire worlds in only a few minutes.
Midjourney, Dall-e 2, and the hundreds of other AI creative tools that are no-doubt coming, will serve as an extension to your imagination. The mechanical barriers to materializing the realms in your mind are dissolving, for better or for worse.
Prompts for Worldbuilding with Midjourney
I'll share a few prompt formats I've been using for riffing on concept art. You'll notice little keyword tricks you can use to influence quality and style. Stuff like "4K", "trending on artstation"*, and "moody lighting" goes a long way.
4k of barren rocky wasteland devoid of life with mountains in the distance maximum detail dark fantasy photorealistic moody lightin
4k of a foreboding door into a dark dungeon that goes deep below concept art maximum detail houdini-render photorealistic moody lighting
Your campaign's magic MacGuffin just got a new makeover!
4k of concept art D&D magic item sword of eternal ice dof maximum detail dark fantasy photorealistic moody lighting, glowing blue
Midjourney is getting surprisingly good at creating maps, though the consistency requirements of maps makes it a harder task than the abstract or natural stuff above. They're not "ready-to-go", but I can see a lot of potential here.
top-down view of a dungeon map on parchment paper, ink, intricately detailed 4k
What are your thoughts?
Is AI art the democratization of visual creativity? *Or is it a fancy way to launder copyright and the effort of the world's artists into zero-marginal-cost assets? What will you do with this capability? What if LegendKeeper could look at an article you wrote and then generate a preview image for it automatically? Or vice versa; look at your image and then write an article? What if you dropped a pin on the map and a town map and its inhabitants, prose and pictures for all, suddenly sprung to life? Stop by the Discord and let me know what you think.
In addition to taking part in the AI art revolution, I got a lot done this week! I hope to have another patch out soon for 0.12.x to reintroduce preview images and few other quality of life things. Progress on the new editor, mobile UI, and public worlds is also going really well; honestly a lot faster than I thought it would. I think it's likely the first mobile-ready, public world will launch before the end of August. This will perhaps be LK's most important milestone ever.
Justin and I are going to GenCon next week! We look forward to wandering the convention floor, and maybe even playing a game or two. This is our first GenCon so I'm pretty excited! PAX Unplugged was a blast and apparently GenCon is even bigger. If you happen to be going, let me know and we can say hello there!
There will be no weekly next week!