My Second Lora: Mattew Mercer

TL;DR:

1. This is my second attempt at training a LoRA - Would love CONSTRUCTIVE feedback

2. Trained using rMadArt (SD1.5) (v4.0 | 2BD42F315F specifically)

3. 13 High Res images used, hand captioned

4. 1,300 total Steps

5. Should I make more LoRAs?

6. There will like likely be a second version of this with more training data provided

Settings:

• CFG: Seems pretty flexible, but thoroughly tested on 5 - 9 CFG

• Weight: LoRA weight works fine at 1.0, but as per most LoRA, 0.7 gives you the best results

• Model Checkpoints: Although trained on rMadArt, I've tested with a number of other models and it seems to play very nicely

• Trigger Word: No trigger word needed, but if you have issues, use 'Matt Mercer', especially at the front.

Usage:

• Non-commercial

• Don't be a dick with your renders

• The intention of this LoRA if for creating tasteful art and fan art

• Any request by Matt Mercer to remove this will be honoured immediately

Contact/Questions:

Send through to contact<dot>hotdog800<at>gmail<dot>com