Bring the iconic nostalgia of classic 80s and 90s fantasy art to your generations! This LoRA captures the unique oil-painting aesthetic of legendary artist Jeff Easley, famous for his timeless Dungeons & Dragons (AD&D) rulebook covers, fierce dragons, and gritty, detailed fantasy characters.
Welcome to Version 2 of the Jeff Easley Style LoRA! This version has been completely overhauled with a massively expanded dataset.
It perfectly replicates the legendary oil-painting aesthetic famous for Dungeons & Dragons (AD&D) rulebook covers, fierce dragons, gritty heroes, and atmospheric sword-and-sorcery illustrations.
🚀 What's New in V2?
Massive Dataset: 107 high-quality curated images.
⚙️ Training Details & Dataset Info:
Dataset Size: 107 Images (fully captioned and tagged)
Training Resolution: 512 x 512 pixels
Total Training Steps: 3,585 Steps
Total Epochs: 67 Epochs (Batch Size 2 in OneTrainer)
🔧 Recommended Generation Settings:
Trigger Phrase:
vintage fantasy oil painting in the style of Jeff EasleyLoRA Weight:
0.5 - 0.7Sweetspot:
0.6(All showcase/example images were generated exactly at0.6weight for the perfect balance of style and flexibility!)
CFG Scale:
3.5 - 5.5(Keep it lower for that authentic, analog hand-painted texture)
📝 Prompting Tips:
To get the absolute best out of this V2, mix the trigger phrase with classic dark fantasy keywords:
oil on canvas, cinematic chiaroscuro, heavy shadows, dramatic lighting, detailed armor, dynamic combat stance, retro fantasy book cover illustration.
If you like this version, please leave a review and share your generations below! I’d love to see your retro fantasy worlds.
Description
FAQ
Comments (3)
yeah this one slaps
LoRA seems to get best image when you using "A highly detailed digital matte painting in a cinematic concept art style" although I don't know that's it's fully indicative of Jeff's actual style in terms of fine detail. I won't wonder if the his name was even used in the captions and if other stylization words were excluded from the captions?
The dataset I used was old; it still contained prompts tagged with Florence. It was also one of the first LoRAs I made for Krea2. I checked again, and it contains things like 'A dark fantasy illustration shoot, A vibrant and detailed depiction, A vibrant, fantasy digital illustration' etc. 'A highly detailed digital matte painting in a cinematic concept art style' doesn't appear at all. I didn't use any trigger words either, but I'm working on a v2 LoRA with better tagging, trigger words, and a 1024x image dataset. However, I was pleasantly surprised by the results, otherwise I wouldn't have uploaded the LoRA.