From my first successfully trained LoRA of Sawyer.
Captured images from the source material (207), upscaled them very cautiously (to 2048x2048), then curated and manually corrected errors. There's more work that could be done, but this is fine for now.
6 months ago