This model creates awesome space marines star field images I’m still learning how to create sd1.5 Lora’s fast on low end device iOS I used draw things and it only took 1hour and 35 minutes to train in 567 steps u can create nsfw potraits as I did train 1 topless girl image so it’s rare u really would have to try hard to get nsfw art but anyway have fun;)

Starfield_Galactica

Space battles! 🥹🤓👍🏾 clip skip 0 fixes deformation issues sd1.5 doesn’t need clip skip it ruins the images

Stealth joycap.zip

starfield_galatica_567_ema_0.135_lora_f32.safetensors.zip

Starfield_Galactic

I’m now creating sd1.5 Lora’s I figured out how to train a Lora with under 15 images and 50 steps it takes only 9 minutes no need to wait for 1000 steps it’s awesome!! All done on iOS 13 pro 6gb ram

Stealth_Echelon

This version took s couple retries but I finnaly got the style I relized I need to add trigger word for style Lora’s as sd1.5 uses the first tokens that u put in your images as the triggers for each image style so to ge the rainy out door style for example put this at beginning of your prompt:A dimly lit, rain-soaked night scene from a sci-fi film features aThis is not a trigger word but an image style trigger from the training data so download the training data and copy the first tokens from what ever images style you want in your images if I had of put a trigger word I can just put the trigger word and it would use all the training data images style because it puts the trigger word as first token in each image caption this is how sd1.5 works next time if I make a version 3 I’ll put a trigger word😅👍🏾

Files with SHA256: 7b8aaeec80155b5d9bf1b7eae3080e7f1c212060e16664d1fc43e3ba714bae98