v1's my first 3.5 medium lora. The samples during training were doodooass, but it works 'okay' during inference.
FinallyGoodEdition is finally good, and was trained using sd-scripts with immiscible noise and many other adjustments because holy hell things seem messy unless you dig a lot.
Trigger words
You can use mspaint style to trigger the image generation. It may not be needed, however.
Download model
Weights for this model are available in Safetensors format. I recommend mspaint-style-sd3-FinallyGoodEdition.safetensors
