This is the first model I've ever trained and it came out 100x better than I thought it would, though you should probably expect janky results.

If someone could peer inside the code and let me know what I could have done better, please do so I can make a better follow-up.

UPDATE: So it turns out this model works a lot better if you're using it without other LoRAs. If you're using it with other LoRAs you'd be better off turning the strength down to 0.4-0.5, especially if it's a character with the proportions of, say, a Sonic character.

Also it turns out you're supposed to remove the background from the source images before you start training. Wish I'd found out before I blew my 500 credits.