Need prompts for this LoRa? Try these pre-made JoyCaption prompt templates (94 in total): https://imgur.com/gallery/outfit-prompts-w-o-images-fFfcXnu
You can also try these streetphoto prompt templates: https://imgur.com/gallery/streetphoto-captions-pt-1-v0E6UyK
FLUX is a tad lacking when it comes to fabrics and fashion so I made a LoRa for it.
It's a mix between modern goth style and cosplay fashion , using the composite image LoRa training method.
LoRa uses the keyword 'xxx photography' so that it is compatible with the other LoRas I have trained.
//--//
As usual , feed a reference image into Joycaption to get a prompt useful with this LoRa: https://huggingface.co/spaces/fancyfeast/joy-caption-alpha-two
A good custom Joycaption prompt is the phrase 'describe this image in 400 words' .
//---//
Training data: https://imgur.com/gallery/dollskills-fashion-lora-training-data-0wh0oSe
//---//
Training parameters:
{ "engine": "kohya", "unetLR": 0.00018, "clipSkip": 1, "loraType": "lora", "keepTokens": 0, "networkDim": 64, "numRepeats": 20, "resolution": 1024, "lrScheduler": "cosine_with_restarts", "minSnrGamma": 5, "noiseOffset": 0.15, "targetSteps": 2700, "enableBucket": true, "networkAlpha": 32, "optimizerType": "AdamW8Bit", "textEncoderLR": 0, "maxTrainEpochs": 20, "shuffleCaption": false, "trainBatchSize": 4, "flipAugmentation": false, "lrSchedulerNumCycles": 4 }