v1 trained on unedited cogVLM captions, this means the captions are verbose and natural language like descriptions as opposed to a list of tags.









This SDXL LoRA attempts to capture the stylistic elements of the aesthetic that has recently been branded as Frutiger Aero. Because this branding of the style is fairly new I noticed base models don't really know this concept. This is a somewhat hyper version of that aesthetic focusing on grassy hills, cloudy blue skies, water splashing, iridescent translucent plastic, dolphins, hummingbirds, flowers, rainbows and lens flares. This was trained on 28 images, a selection of the classic frutiger aero style wallpapers, new images creating by blending the classics in MidJourney to create new variations and a handful of dalle3 and MJ prompted images. Captions on this version are unedited, raw cogvlm outputs prepended with the tag 'FA Style', it was trained with a batch size 4 & gradient accumulation 2 with the prodigy optimizer for 40 epochs or 1400 steps. After testing I chose epoch 35 to share.
v1 trained on unedited cogVLM captions, this means the captions are verbose and natural language like descriptions as opposed to a list of tags.
1. 재게시된 모델의 권리는 원 제작자에게 있습니다.
2. 모델 원작자가 모델을 인증받으려면 공식 채널을 통해 SeaArt.AI 직원에게 문의하세요. 저희는 모든 창작자의 권리를 보호하기 위해 노력합니다. 인증하러 이동
