May 7, 2023


This week I read:

Roleplaying Games

The Halls of Arden Vul

Well, we’ve given up on finding new players for now.

So it’s just me, two regular players, and a couple of people who are occasionally around but can’t commit to anything regular. So Impossible Landscapes is out. It’s a shame, we were all really looking forward to it, but I’d rather wait and run it with four players in the future than run it with two players now.

This week we didn’t play an RPG. We played Spirit Island instead, which was pretty fun. Afterwards I raised the elephant in the room, we talked about possible alternative campaigns, and have decided that next week will be session 0 for The Halls of Arden Vul.

I think this is a good fit because when the flaky players are around, or if someone new wants to try things out, they can just take over an NPC retainer (since it won’t just be two PCs delving the dungeon on their own). There’s no overarching plot, just an overarching location, so it’s easy for people to drop in and out according to their availability. And if someone new who can’t commit to a regular game joins the group to play occasionally, they might end up liking it enough that they make time for a regular game.

Impossible Landscapes, we’ll get to you one day!

So, uh, now I just need to prepare for a huge megadungeon I’d not expected to run until next year, in the next fortnight. So I’ve been reading volume 1 (which covers the lower-level areas and is likely to be the subject of months of play), volume 4 (which covers all the peoples, places, items, and lore), and volume 5 (the maps) and hope to have at least a broad understanding of all that content by the time we start. Week to week I’ll read through the material I expect the next session to cover, but I’d like at least a rough overview before we start.

Of course, one of the advantages of a megadungeon is that prep is pretty light: the author’s done all that for you. You just need to run it as written and, after every session, think about how the players have changed the status quo. But mapping of locations, finding of monster stats, creation of major factions and NPCs, is largely done for you. It does look like I’ll need to make maps for some of the adventure sites outside of the dungeon proper, but many of those sites are hidden and will take time for the players to find, so if I don’t have it all ready by session 1 it’s no big deal.

It’ll be fun!

Stable Diffusion

I’ve been playing around with Stable Diffusion (via Easy Diffusion) this week and it is pretty amazing. Back in March, I generated some character portraits with DALL-E and they’re alright, it’s cool to see a computer making art, but they’re all pretty basic.

However, I had seen really good AI art, so I assumed I just didn’t know how to use DALL-E properly. After all, DALL-E is OpenAI’s cutting edge proprietary model, it costs actual money to use (beyond a small monthly allowance), so it must be near the top, right?

Reader, I was wrong. So, so, wrong. Stable Diffusion is so much better than DALL-E.

A mushroom person villager, generated with Stable Diffusion.

The images are just… well, go check out my Twitter thread sharing a bunch.

It’s free. It runs locally. It takes mere minutes. I’m no “prompt engineer”, I’m sure there are tricks I don’t know to generate the really good stuff, but just what I’m able to puzzle out myself is leagues ahead of what I was getting out of DALL-E.

The age of bad NPC portraits is over!