RPG Maps: v4.0: TTRPG-000001.safetensors

Began captioning images manually, resulting in really good quality after only one epoch. If this model ends up undercooked, I have later epochs from this training run saved, but they appear to have equal quality in my testing

Update 10/11/2024:

Since Civitai doesn't allow DoRAs in their image generator, I've opted to continue making LoRA updates, however for anyone running locally, I HIGHLY recommend using the DoRA instead, most UIs that support LoRAs also support DoRAs and are much higher quality.

Next update will have a complete revamp of the dataset, separating battle maps from dungeon maps, and greater variety for greater usability, though preventing overtraining can be difficult.

The DoRA can be found here:

https://civarchive.com/models/840900