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:
Description
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
FAQ
Comments (4)
interesting! thanks
one day, these loras will be able to generate images at 60FPS and then we might be able to play these games like in the matrix :P
People on Reddit are already using it to make 3d models, and then there's CogVideo...
@Thatrandomjew well, if we could render cogvideo in realtime, maybe we could control it somehow while its rendering using the keyboard. Thereby making it into a game?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.