V.2 now available, Trained on 300+ images caption pairs from a selection of 5 Stanley Kubrick films (Clockwork Orange, The Shining, Full Metal Jacket, 2001, Eyes Wide Shut), this LoRA attempts to capture some of his iconic aesthetic in lighting and composition.
Description
Dataset expanded 3x to 300+ image caption pairs, trained using prodigy optimizer and a lightly modified version of the ai-toolkit script. Using wide aspect ratio's like 1408 x 768 that are similar to the film you may be trying to recreate images from is useful.
FAQ
Comments (10)
Beautiful. God loves Kubrick!
thanks god
great work - please consider using a shrinking script to get the size down considerably.
You mean the size of the Lora? No thanks, I take quality before filesize, also it's like 300mb only?
@olivetty definitely but it can still be shrunk considerably without any reduction in quality, look around it's a common comment/update here with the Flux LoRas...you can use the 'Resize LoRa' option in the Koyha utilities
@vibrantthang For styles I do prefer higher ranks, low size / 16rank loras are for clothing, poses, or light character training.
looking into other options, but for now, this is what i got
Nice work ! Results are quite consistent. Why have you choosen to go with prodigy optimizer rather than adafactor ?
out of all my tests i preferred the prodigy results, i know other people have had the same experience, this was the same case with SDXL as well
@blairesilver13 interesting, will try. Thanks
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.